提交 1cafa1b3 编写于 作者: H Hongze Cheng

Merge branch '3.0' into 3.0_refact

...@@ -990,6 +990,22 @@ typedef struct { ...@@ -990,6 +990,22 @@ typedef struct {
/* data */ /* data */
} SAlterTableRsp; } SAlterTableRsp;
typedef struct {
/* data */
} SDropStableReq;
typedef struct {
/* data */
} SDropStableRsp;
typedef struct {
/* data */
} SUpdateTagValReq;
typedef struct {
/* data */
} SUpdateTagValRsp;
#pragma pack(pop) #pragma pack(pop)
#ifdef __cplusplus #ifdef __cplusplus
......
...@@ -57,6 +57,7 @@ extern int32_t tsCompressMsgSize; ...@@ -57,6 +57,7 @@ extern int32_t tsCompressMsgSize;
extern int32_t tsCompressColData; extern int32_t tsCompressColData;
extern int32_t tsMaxNumOfDistinctResults; extern int32_t tsMaxNumOfDistinctResults;
extern char tsTempDir[]; extern char tsTempDir[];
extern int64_t tsMaxVnodeQueuedBytes;
//query buffer management //query buffer management
extern int32_t tsQueryBufferSize; // maximum allowed usage buffer size in MB for each data node during query processing extern int32_t tsQueryBufferSize; // maximum allowed usage buffer size in MB for each data node during query processing
......
...@@ -118,7 +118,7 @@ typedef struct { ...@@ -118,7 +118,7 @@ typedef struct {
int32_t mnodeGetStatistics(SMnodeStat *stat); int32_t mnodeGetStatistics(SMnodeStat *stat);
/** /**
* Get the auth information. * Get the auth information of Mnode.
* *
* @param user, username. * @param user, username.
* @param spi, security parameter index. * @param spi, security parameter index.
......
...@@ -49,7 +49,7 @@ typedef struct { ...@@ -49,7 +49,7 @@ typedef struct {
} SVnodeFp; } SVnodeFp;
typedef struct { typedef struct {
SVnodeFp fp; SVnodeFp fp;
} SVnodePara; } SVnodePara;
/** /**
......
...@@ -20,14 +20,14 @@ ...@@ -20,14 +20,14 @@
extern "C" { extern "C" {
#endif #endif
typedef int32_t (*InitFp)(void **obj); typedef int32_t (*InitFp)();
typedef void (*CleanupFp)(void **obj); typedef void (*CleanupFp)();
typedef void (*ReportFp)(char *name, char *desc); typedef void (*ReportFp)(char *name, char *desc);
struct SSteps *taosStepInit(int32_t maxsize, ReportFp fp); struct SSteps *taosStepInit(int32_t maxsize, ReportFp fp);
int32_t taosStepExec(struct SSteps *steps); int32_t taosStepExec(struct SSteps *steps);
void taosStepCleanup(struct SSteps *steps); void taosStepCleanup(struct SSteps *steps);
int32_t taosStepAdd(struct SSteps *steps, char *name, void **obj, InitFp initFp, CleanupFp cleanupFp); int32_t taosStepAdd(struct SSteps *steps, char *name, InitFp initFp, CleanupFp cleanupFp);
#ifdef __cplusplus #ifdef __cplusplus
} }
......
...@@ -20,14 +20,16 @@ ...@@ -20,14 +20,16 @@
extern "C" { extern "C" {
#endif #endif
typedef int32_t (*ProcessReqFp)(void *ahandle, void *msg); typedef int32_t (*ProcessStartFp)(void *ahandle, void *pMsg, int32_t qtype);
typedef void (*SendRspFp)(void *ahandle, void *msg, int32_t qtype, int32_t code); typedef void (*ProcessEndFp)(void *ahandle, void *pMsg, int32_t qtype, int32_t code);
struct SWorkerPool; typedef bool (*ProcessWriteStartFp)(void *ahandle, void *pMsg, int32_t qtype);
typedef void (*ProcessWriteSyncFp)(void *ahandle, int32_t code);
typedef void (*ProcessWriteEndFp)(void *ahandle, void *pMsg, int32_t qtype);
typedef struct { typedef struct SWorker {
pthread_t thread; // thread
int32_t id; // worker ID int32_t id; // worker ID
pthread_t thread; // thread
struct SWorkerPool *pool; struct SWorkerPool *pool;
} SWorker; } SWorker;
...@@ -35,18 +37,42 @@ typedef struct SWorkerPool { ...@@ -35,18 +37,42 @@ typedef struct SWorkerPool {
int32_t max; // max number of workers int32_t max; // max number of workers
int32_t min; // min number of workers int32_t min; // min number of workers
int32_t num; // current number of workers int32_t num; // current number of workers
void * qset; taos_qset qset;
const char * name; const char * name;
ProcessStartFp startFp;
ProcessEndFp endFp;
SWorker * workers; SWorker * workers;
ProcessReqFp reqFp;
SendRspFp rspFp;
pthread_mutex_t mutex; pthread_mutex_t mutex;
} SWorkerPool; } SWorkerPool;
int32_t tWorkerInit(SWorkerPool *pPool); typedef struct SWriteWorker {
void tWorkerCleanup(SWorkerPool *pPool); int32_t id; // worker id
void * tWorkerAllocQueue(SWorkerPool *pPool, void *ahandle); pthread_t thread; // thread
void tWorkerFreeQueue(SWorkerPool *pPool, void *pQueue); taos_qall qall;
taos_qset qset; // queue set
struct SWriteWorkerPool *pool;
} SWriteWorker;
typedef struct SWriteWorkerPool {
int32_t max; // max number of workers
int32_t nextId; // from 0 to max-1, cyclic
const char * name;
ProcessWriteStartFp startFp;
ProcessWriteSyncFp syncFp;
ProcessWriteEndFp endFp;
SWriteWorker * workers;
pthread_mutex_t mutex;
} SWriteWorkerPool;
int32_t tWorkerInit(SWorkerPool *pool);
void tWorkerCleanup(SWorkerPool *pool);
taos_queue tWorkerAllocQueue(SWorkerPool *pool, void *ahandle);
void tWorkerFreeQueue(SWorkerPool *pool, taos_queue queue);
int32_t tWriteWorkerInit(SWriteWorkerPool *pool);
void tWriteWorkerCleanup(SWriteWorkerPool *pool);
taos_queue tWriteWorkerAllocQueue(SWriteWorkerPool *pool, void *ahandle);
void tWriteWorkerFreeQueue(SWriteWorkerPool *pool, taos_queue queue);
#ifdef __cplusplus #ifdef __cplusplus
} }
......
...@@ -60,6 +60,7 @@ float tsRatioOfQueryCores = 1.0f; ...@@ -60,6 +60,7 @@ float tsRatioOfQueryCores = 1.0f;
int8_t tsDaylight = 0; int8_t tsDaylight = 0;
int8_t tsEnableCoreFile = 0; int8_t tsEnableCoreFile = 0;
int32_t tsMaxBinaryDisplayWidth = 30; int32_t tsMaxBinaryDisplayWidth = 30;
int64_t tsMaxVnodeQueuedBytes = 1024*1024*1024; //1GB
/* /*
* denote if the server needs to compress response message at the application layer to client, including query rsp, * denote if the server needs to compress response message at the application layer to client, including query rsp,
......
...@@ -21,21 +21,14 @@ extern "C" { ...@@ -21,21 +21,14 @@ extern "C" {
#endif #endif
#include "dnodeInt.h" #include "dnodeInt.h"
typedef struct SDnCfg {
int32_t dnodeId;
int32_t dropped;
char clusterId[TSDB_CLUSTER_ID_LEN];
char file[PATH_MAX + 20];
pthread_mutex_t mutex;
} SDnCfg;
int32_t dnodeInitCfg(SDnCfg **cfg); int32_t dnodeInitCfg();
void dnodeCleanupCfg(SDnCfg **cfg); void dnodeCleanupCfg();
void dnodeUpdateCfg(SDnCfg *cfg, SDnodeCfg *data); void dnodeUpdateCfg(SDnodeCfg *data);
int32_t dnodeGetDnodeId(SDnCfg *cfg); int32_t dnodeGetDnodeId();
void dnodeGetClusterId(SDnCfg *cfg, char *clusterId); void dnodeGetClusterId(char *clusterId);
void dnodeGetCfg(SDnCfg *cfg, int32_t *dnodeId, char *clusterId); void dnodeGetCfg(int32_t *dnodeId, char *clusterId);
void dnodeSetDropped(SDnCfg *cfg); void dnodeSetDropped();
#ifdef __cplusplus #ifdef __cplusplus
} }
......
...@@ -21,11 +21,9 @@ extern "C" { ...@@ -21,11 +21,9 @@ extern "C" {
#endif #endif
#include "dnodeInt.h" #include "dnodeInt.h"
typedef struct SDnCheck {
} SDnCheck;
int32_t dnodeInitCheck(SDnCheck **check); int32_t dnodeInitCheck();
void dnodeCleanupCheck(SDnCheck **check); void dnodeCleanupCheck();
#ifdef __cplusplus #ifdef __cplusplus
} }
......
...@@ -19,22 +19,12 @@ ...@@ -19,22 +19,12 @@
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "thash.h"
#include "dnodeInt.h" #include "dnodeInt.h"
typedef struct SDnEps { int32_t dnodeInitEps();
int32_t dnodeId; void dnodeCleanupEps();
int32_t dnodeNum; void dnodeUpdateEps(SDnodeEps *data);
SDnodeEp * dnodeList; bool dnodeIsDnodeEpChanged(int32_t dnodeId, char *epstr);
SHashObj * dnodeHash;
char file[PATH_MAX + 20];
pthread_mutex_t mutex;
} SDnEps;
int32_t dnodeInitEps(SDnEps **eps);
void dnodeCleanupEps(SDnEps **eps);
void dnodeUpdateEps(SDnEps *eps, SDnodeEps *data);
bool dnodeIsDnodeEpChanged(SDnEps *eps, int32_t dnodeId, char *epstr);
void dnodeGetDnodeEp(int32_t dnodeId, char *epstr, char *fqdn, uint16_t *port); void dnodeGetDnodeEp(int32_t dnodeId, char *epstr, char *fqdn, uint16_t *port);
#ifdef __cplusplus #ifdef __cplusplus
......
...@@ -19,36 +19,12 @@ ...@@ -19,36 +19,12 @@
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "taoserror.h"
#include "taosmsg.h" #include "taosmsg.h"
#include "tglobal.h"
#include "tlog.h" #include "tlog.h"
#include "trpc.h" #include "trpc.h"
#include "tstep.h"
#include "dnode.h" #include "dnode.h"
struct SDnCfg; extern int32_t dDebugFlag;
struct SDnCheck;
struct SDnEps;
struct SDnMnEps;
struct SDnStatus;
struct SDnTelem;
struct SDnTrans;
struct SDnMain;
typedef struct SDnode {
struct SSteps* steps;
struct SDnCfg* cfg;
struct SDnCheck* check;
struct SDnEps* eps;
struct SDnMnEps* meps;
struct SDnStatus* status;
struct SDnTelem* telem;
struct SDnTrans* trans;
struct SDnMain* main;
} SDnode;
SDnode* dnodeInst();
#define dFatal(...) { if (dDebugFlag & DEBUG_FATAL) { taosPrintLog("DND FATAL ", 255, __VA_ARGS__); }} #define dFatal(...) { if (dDebugFlag & DEBUG_FATAL) { taosPrintLog("DND FATAL ", 255, __VA_ARGS__); }}
#define dError(...) { if (dDebugFlag & DEBUG_ERROR) { taosPrintLog("DND ERROR ", 255, __VA_ARGS__); }} #define dError(...) { if (dDebugFlag & DEBUG_ERROR) { taosPrintLog("DND ERROR ", 255, __VA_ARGS__); }}
......
...@@ -27,14 +27,8 @@ typedef enum { ...@@ -27,14 +27,8 @@ typedef enum {
TD_RUN_STAT_STOPPED TD_RUN_STAT_STOPPED
} RunStat; } RunStat;
typedef struct SDnMain { int32_t dnodeInitMain();
RunStat runStatus; void dnodeCleanupMain();
void * dnodeTimer;
SStartupStep startup;
} SDnMain;
int32_t dnodeInitMain(SDnMain **main);
void dnodeCleanupMain(SDnMain **main);
int32_t dnodeInitStorage(); int32_t dnodeInitStorage();
void dnodeCleanupStorage(); void dnodeCleanupStorage();
void dnodeReportStartup(char *name, char *desc); void dnodeReportStartup(char *name, char *desc);
...@@ -42,6 +36,9 @@ void dnodeReportStartupFinished(char *name, char *desc); ...@@ -42,6 +36,9 @@ void dnodeReportStartupFinished(char *name, char *desc);
void dnodeProcessStartupReq(SRpcMsg *pMsg); void dnodeProcessStartupReq(SRpcMsg *pMsg);
void dnodeProcessCreateMnodeReq(SRpcMsg *pMsg); void dnodeProcessCreateMnodeReq(SRpcMsg *pMsg);
void dnodeProcessConfigDnodeReq(SRpcMsg *pMsg); void dnodeProcessConfigDnodeReq(SRpcMsg *pMsg);
RunStat dnodeGetRunStat();
void dnodeSetRunStat();
void* dnodeGetTimer();
#ifdef __cplusplus #ifdef __cplusplus
} }
......
...@@ -21,19 +21,12 @@ extern "C" { ...@@ -21,19 +21,12 @@ extern "C" {
#endif #endif
#include "dnodeInt.h" #include "dnodeInt.h"
typedef struct SDnMnEps { int32_t dnodeInitMnodeEps();
SRpcEpSet mnodeEpSet; void dnodeCleanupMnodeEps();
SMInfos mnodeInfos; void dnodeUpdateMnodeFromStatus(SMInfos *pMinfos);
char file[PATH_MAX + 20]; void dnodeUpdateMnodeFromPeer(SRpcEpSet *pEpSet);
pthread_mutex_t mutex; void dnodeGetEpSetForPeer(SRpcEpSet *epSet);
} SDnMnEps; void dnodeGetEpSetForShell(SRpcEpSet *epSet);
int32_t dnodeInitMnodeEps(SDnMnEps **meps);
void dnodeCleanupMnodeEps(SDnMnEps **meps);
void dnodeUpdateMnodeFromStatus(SDnMnEps *meps, SMInfos *pMinfos);
void dnodeUpdateMnodeFromPeer(SDnMnEps *meps, SRpcEpSet *pEpSet);
void dnodeGetEpSetForPeer(SDnMnEps *meps, SRpcEpSet *epSet);
void dnodeGetEpSetForShell(SDnMnEps *meps, SRpcEpSet *epSet);
void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell); void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell);
#ifdef __cplusplus #ifdef __cplusplus
......
...@@ -21,14 +21,8 @@ extern "C" { ...@@ -21,14 +21,8 @@ extern "C" {
#endif #endif
#include "dnodeInt.h" #include "dnodeInt.h"
typedef struct SDnStatus { int32_t dnodeInitStatus();
void * dnodeTimer; void dnodeCleanupStatus();
void * statusTimer;
uint32_t rebootTime;
} SDnStatus;
int32_t dnodeInitStatus(SDnStatus **status);
void dnodeCleanupStatus(SDnStatus **status);
void dnodeProcessStatusRsp(SRpcMsg *pMsg); void dnodeProcessStatusRsp(SRpcMsg *pMsg);
#ifdef __cplusplus #ifdef __cplusplus
......
...@@ -21,21 +21,8 @@ extern "C" { ...@@ -21,21 +21,8 @@ extern "C" {
#endif #endif
#include "dnodeInt.h" #include "dnodeInt.h"
/* int32_t dnodeInitTelem();
* sem_timedwait is NOT implemented on MacOSX void dnodeCleanupTelem();
* thus we use pthread_mutex_t/pthread_cond_t to simulate
*/
typedef struct SDnTelem {
bool enable;
pthread_mutex_t lock;
pthread_cond_t cond;
volatile int32_t exit;
pthread_t thread;
char email[TSDB_FQDN_LEN];
} SDnTelem;
int32_t dnodeInitTelem(SDnTelem **telem);
void dnodeCleanupTelem(SDnTelem **telem);
#ifdef __cplusplus #ifdef __cplusplus
} }
......
...@@ -21,20 +21,8 @@ extern "C" { ...@@ -21,20 +21,8 @@ extern "C" {
#endif #endif
#include "dnodeInt.h" #include "dnodeInt.h"
typedef void (*RpcMsgFp)( SRpcMsg *pMsg); int32_t dnodeInitTrans();
void dnodeCleanupTrans();
typedef struct SDnTrans {
void * serverRpc;
void * clientRpc;
void * shellRpc;
int32_t queryReqNum;
int32_t submitReqNum;
RpcMsgFp peerMsgFp[TSDB_MSG_TYPE_MAX];
RpcMsgFp shellMsgFp[TSDB_MSG_TYPE_MAX];
} SDnTrans;
int32_t dnodeInitTrans(SDnTrans **rans);
void dnodeCleanupTrans(SDnTrans **trans);
void dnodeSendMsgToMnode(SRpcMsg *rpcMsg); void dnodeSendMsgToMnode(SRpcMsg *rpcMsg);
void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg); void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg);
void dnodeSendMsgToDnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp, SRpcEpSet *epSet); void dnodeSendMsgToDnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp, SRpcEpSet *epSet);
......
...@@ -16,56 +16,65 @@ ...@@ -16,56 +16,65 @@
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "os.h"
#include "cJSON.h" #include "cJSON.h"
#include "tglobal.h"
#include "dnodeCfg.h" #include "dnodeCfg.h"
static int32_t dnodeReadCfg(SDnCfg *cfg) { static struct DnCfg {
int32_t dnodeId;
int32_t dropped;
char clusterId[TSDB_CLUSTER_ID_LEN];
char file[PATH_MAX + 20];
pthread_mutex_t mutex;
} tsDcfg;
static int32_t dnodeReadCfg() {
int32_t len = 0; int32_t len = 0;
int32_t maxLen = 200; int32_t maxLen = 200;
char * content = calloc(1, maxLen + 1); char * content = calloc(1, maxLen + 1);
cJSON * root = NULL; cJSON * root = NULL;
FILE * fp = NULL; FILE * fp = NULL;
fp = fopen(cfg->file, "r"); fp = fopen(tsDcfg.file, "r");
if (!fp) { if (!fp) {
dDebug("file %s not exist", cfg->file); dDebug("file %s not exist", tsDcfg.file);
goto PARSE_CFG_OVER; goto PARSE_CFG_OVER;
} }
len = (int32_t)fread(content, 1, maxLen, fp); len = (int32_t)fread(content, 1, maxLen, fp);
if (len <= 0) { if (len <= 0) {
dError("failed to read %s since content is null", cfg->file); dError("failed to read %s since content is null", tsDcfg.file);
goto PARSE_CFG_OVER; goto PARSE_CFG_OVER;
} }
content[len] = 0; content[len] = 0;
root = cJSON_Parse(content); root = cJSON_Parse(content);
if (root == NULL) { if (root == NULL) {
dError("failed to read %s since invalid json format", cfg->file); dError("failed to read %s since invalid json format", tsDcfg.file);
goto PARSE_CFG_OVER; goto PARSE_CFG_OVER;
} }
cJSON *dnodeId = cJSON_GetObjectItem(root, "dnodeId"); cJSON *dnodeId = cJSON_GetObjectItem(root, "dnodeId");
if (!dnodeId || dnodeId->type != cJSON_Number) { if (!dnodeId || dnodeId->type != cJSON_Number) {
dError("failed to read %s since dnodeId not found", cfg->file); dError("failed to read %s since dnodeId not found", tsDcfg.file);
goto PARSE_CFG_OVER; goto PARSE_CFG_OVER;
} }
cfg->dnodeId = (int32_t)dnodeId->valueint; tsDcfg.dnodeId = (int32_t)dnodeId->valueint;
cJSON *dropped = cJSON_GetObjectItem(root, "dropped"); cJSON *dropped = cJSON_GetObjectItem(root, "dropped");
if (!dropped || dropped->type != cJSON_Number) { if (!dropped || dropped->type != cJSON_Number) {
dError("failed to read %s since dropped not found", cfg->file); dError("failed to read %s since dropped not found", tsDcfg.file);
goto PARSE_CFG_OVER; goto PARSE_CFG_OVER;
} }
cfg->dropped = (int32_t)dropped->valueint; tsDcfg.dropped = (int32_t)dropped->valueint;
cJSON *clusterId = cJSON_GetObjectItem(root, "clusterId"); cJSON *clusterId = cJSON_GetObjectItem(root, "clusterId");
if (!clusterId || clusterId->type != cJSON_String) { if (!clusterId || clusterId->type != cJSON_String) {
dError("failed to read %s since clusterId not found", cfg->file); dError("failed to read %s since clusterId not found", tsDcfg.file);
goto PARSE_CFG_OVER; goto PARSE_CFG_OVER;
} }
tstrncpy(cfg->clusterId, clusterId->valuestring, TSDB_CLUSTER_ID_LEN); tstrncpy(tsDcfg.clusterId, clusterId->valuestring, TSDB_CLUSTER_ID_LEN);
dInfo("successed to read %s", cfg->file); dInfo("successed to read %s", tsDcfg.file);
PARSE_CFG_OVER: PARSE_CFG_OVER:
if (content != NULL) free(content); if (content != NULL) free(content);
...@@ -76,10 +85,10 @@ PARSE_CFG_OVER: ...@@ -76,10 +85,10 @@ PARSE_CFG_OVER:
return 0; return 0;
} }
static int32_t dnodeWriteCfg(SDnCfg *cfg) { static int32_t dnodeWriteCfg() {
FILE *fp = fopen(cfg->file, "w"); FILE *fp = fopen(tsDcfg.file, "w");
if (!fp) { if (!fp) {
dError("failed to write %s since %s", cfg->file, strerror(errno)); dError("failed to write %s since %s", tsDcfg.file, strerror(errno));
return -1; return -1;
} }
...@@ -88,9 +97,9 @@ static int32_t dnodeWriteCfg(SDnCfg *cfg) { ...@@ -88,9 +97,9 @@ static int32_t dnodeWriteCfg(SDnCfg *cfg) {
char * content = calloc(1, maxLen + 1); char * content = calloc(1, maxLen + 1);
len += snprintf(content + len, maxLen - len, "{\n"); len += snprintf(content + len, maxLen - len, "{\n");
len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", cfg->dnodeId); len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", tsDcfg.dnodeId);
len += snprintf(content + len, maxLen - len, " \"dropped\": %d,\n", cfg->dropped); len += snprintf(content + len, maxLen - len, " \"dropped\": %d,\n", tsDcfg.dropped);
len += snprintf(content + len, maxLen - len, " \"clusterId\": \"%s\"\n", cfg->clusterId); len += snprintf(content + len, maxLen - len, " \"clusterId\": \"%s\"\n", tsDcfg.clusterId);
len += snprintf(content + len, maxLen - len, "}\n"); len += snprintf(content + len, maxLen - len, "}\n");
fwrite(content, 1, len, fp); fwrite(content, 1, len, fp);
...@@ -99,27 +108,23 @@ static int32_t dnodeWriteCfg(SDnCfg *cfg) { ...@@ -99,27 +108,23 @@ static int32_t dnodeWriteCfg(SDnCfg *cfg) {
free(content); free(content);
terrno = 0; terrno = 0;
dInfo("successed to write %s", cfg->file); dInfo("successed to write %s", tsDcfg.file);
return 0; return 0;
} }
int32_t dnodeInitCfg(SDnCfg **out) { int32_t dnodeInitCfg() {
SDnCfg* cfg = calloc(1, sizeof(SDnCfg)); tsDcfg.dnodeId = 0;
if (cfg == NULL) return -1; tsDcfg.dropped = 0;
tsDcfg.clusterId[0] = 0;
cfg->dnodeId = 0; snprintf(tsDcfg.file, sizeof(tsDcfg.file), "%s/dnodeCfg.json", tsDnodeDir);
cfg->dropped = 0; pthread_mutex_init(&tsDcfg.mutex, NULL);
cfg->clusterId[0] = 0;
snprintf(cfg->file, sizeof(cfg->file), "%s/dnodeCfg.json", tsDnodeDir); int32_t ret = dnodeReadCfg();
pthread_mutex_init(&cfg->mutex, NULL);
*out = cfg;
int32_t ret = dnodeReadCfg(cfg);
if (ret == 0) { if (ret == 0) {
dInfo("dnode cfg is initialized"); dInfo("dnode cfg is initialized");
} }
if (cfg->dropped) { if (tsDcfg.dropped) {
dInfo("dnode is dropped and start to exit"); dInfo("dnode is dropped and start to exit");
return -1; return -1;
} }
...@@ -127,51 +132,47 @@ int32_t dnodeInitCfg(SDnCfg **out) { ...@@ -127,51 +132,47 @@ int32_t dnodeInitCfg(SDnCfg **out) {
return ret; return ret;
} }
void dnodeCleanupCfg(SDnCfg **out) { void dnodeCleanupCfg() {
SDnCfg* cfg = *out; pthread_mutex_destroy(&tsDcfg.mutex);
*out = NULL;
pthread_mutex_destroy(&cfg->mutex);
free(cfg);
} }
void dnodeUpdateCfg(SDnCfg *cfg, SDnodeCfg *data) { void dnodeUpdateCfg(SDnodeCfg *data) {
if (cfg == NULL || cfg->dnodeId == 0) return; if (tsDcfg.dnodeId != 0) return;
pthread_mutex_lock(&cfg->mutex); pthread_mutex_lock(&tsDcfg.mutex);
cfg->dnodeId = data->dnodeId; tsDcfg.dnodeId = data->dnodeId;
tstrncpy(cfg->clusterId, data->clusterId, TSDB_CLUSTER_ID_LEN); tstrncpy(tsDcfg.clusterId, data->clusterId, TSDB_CLUSTER_ID_LEN);
dInfo("dnodeId is set to %d, clusterId is set to %s", cfg->dnodeId, cfg->clusterId); dInfo("dnodeId is set to %d, clusterId is set to %s", data->dnodeId, data->clusterId);
dnodeWriteCfg(cfg); dnodeWriteCfg();
pthread_mutex_unlock(&cfg->mutex); pthread_mutex_unlock(&tsDcfg.mutex);
} }
void dnodeSetDropped(SDnCfg *cfg) { void dnodeSetDropped() {
pthread_mutex_lock(&cfg->mutex); pthread_mutex_lock(&tsDcfg.mutex);
cfg->dropped = 1; tsDcfg.dropped = 1;
dnodeWriteCfg(cfg); dnodeWriteCfg();
pthread_mutex_unlock(&cfg->mutex); pthread_mutex_unlock(&tsDcfg.mutex);
} }
int32_t dnodeGetDnodeId(SDnCfg *cfg) { int32_t dnodeGetDnodeId() {
int32_t dnodeId = 0; int32_t dnodeId = 0;
pthread_mutex_lock(&cfg->mutex); pthread_mutex_lock(&tsDcfg.mutex);
dnodeId = cfg->dnodeId; dnodeId = tsDcfg.dnodeId;
pthread_mutex_unlock(&cfg->mutex); pthread_mutex_unlock(&tsDcfg.mutex);
return dnodeId; return dnodeId;
} }
void dnodeGetClusterId(SDnCfg *cfg, char *clusterId) { void dnodeGetClusterId(char *clusterId) {
pthread_mutex_lock(&cfg->mutex); pthread_mutex_lock(&tsDcfg.mutex);
tstrncpy(clusterId, cfg->clusterId, TSDB_CLUSTER_ID_LEN); tstrncpy(clusterId, tsDcfg.clusterId, TSDB_CLUSTER_ID_LEN);
pthread_mutex_unlock(&cfg->mutex); pthread_mutex_unlock(&tsDcfg.mutex);
} }
void dnodeGetCfg(SDnCfg *cfg, int32_t *dnodeId, char *clusterId) { void dnodeGetCfg(int32_t *dnodeId, char *clusterId) {
pthread_mutex_lock(&cfg->mutex); pthread_mutex_lock(&tsDcfg.mutex);
*dnodeId = cfg->dnodeId; *dnodeId = tsDcfg.dnodeId;
tstrncpy(clusterId, cfg->clusterId, TSDB_CLUSTER_ID_LEN); tstrncpy(clusterId, tsDcfg.clusterId, TSDB_CLUSTER_ID_LEN);
pthread_mutex_unlock(&cfg->mutex); pthread_mutex_unlock(&tsDcfg.mutex);
} }
...@@ -118,7 +118,7 @@ static int32_t dnodeCheckMem() { ...@@ -118,7 +118,7 @@ static int32_t dnodeCheckMem() {
} }
static int32_t dnodeCheckDisk() { static int32_t dnodeCheckDisk() {
#if 0 #if 0
taosGetDisk(); taosGetDisk();
if (tsAvailDataDirGB < tsMinimalDataDirGB) { if (tsAvailDataDirGB < tsMinimalDataDirGB) {
...@@ -145,12 +145,7 @@ static int32_t dnodeCheckAccess() { return 0; } ...@@ -145,12 +145,7 @@ static int32_t dnodeCheckAccess() { return 0; }
static int32_t dnodeCheckVersion() { return 0; } static int32_t dnodeCheckVersion() { return 0; }
static int32_t dnodeCheckDatafile() { return 0; } static int32_t dnodeCheckDatafile() { return 0; }
int32_t dnodeInitCheck(SDnCheck **out) { int32_t dnodeInitCheck() {
SDnCheck *check = calloc(1, sizeof(SDnCheck));
if (check == NULL) return -1;
*out = check;
if (dnodeCheckNetwork() != 0) { if (dnodeCheckNetwork() != 0) {
dError("failed to check network"); dError("failed to check network");
return -1; return -1;
...@@ -195,9 +190,4 @@ int32_t dnodeInitCheck(SDnCheck **out) { ...@@ -195,9 +190,4 @@ int32_t dnodeInitCheck(SDnCheck **out) {
return 0; return 0;
} }
void dnodeCleanupCheck(SDnCheck **out) { void dnodeCleanupCheck() {}
SDnCheck *check = *out; \ No newline at end of file
*out = NULL;
free(check);
}
\ No newline at end of file
...@@ -16,86 +16,96 @@ ...@@ -16,86 +16,96 @@
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "os.h"
#include "cJSON.h" #include "cJSON.h"
#include "thash.h"
#include "tglobal.h" #include "tglobal.h"
#include "dnodeEps.h" #include "dnodeEps.h"
#include "dnodeCfg.h" #include "dnodeCfg.h"
static void dnodePrintEps(SDnEps *eps) { static struct {
dDebug("print dnodeEp, dnodeNum:%d", eps->dnodeNum); int32_t dnodeId;
for (int32_t i = 0; i < eps->dnodeNum; i++) { int32_t dnodeNum;
SDnodeEp *ep = &eps->dnodeList[i]; SDnodeEp * dnodeList;
SHashObj * dnodeHash;
char file[PATH_MAX + 20];
pthread_mutex_t mutex;
} tsDeps;
static void dnodePrintEps() {
dDebug("print dnodeEp, dnodeNum:%d", tsDeps.dnodeNum);
for (int32_t i = 0; i < tsDeps.dnodeNum; i++) {
SDnodeEp *ep = &tsDeps.dnodeList[i];
dDebug("dnode:%d, dnodeFqdn:%s dnodePort:%u", ep->dnodeId, ep->dnodeFqdn, ep->dnodePort); dDebug("dnode:%d, dnodeFqdn:%s dnodePort:%u", ep->dnodeId, ep->dnodeFqdn, ep->dnodePort);
} }
} }
static void dnodeResetEps(SDnEps *eps, SDnodeEps *data) { static void dnodeResetEps(SDnodeEps *data) {
assert(data != NULL); assert(data != NULL);
if (data->dnodeNum > eps->dnodeNum) { if (data->dnodeNum > tsDeps.dnodeNum) {
SDnodeEp *tmp = calloc(data->dnodeNum, sizeof(SDnodeEp)); SDnodeEp *tmp = calloc(data->dnodeNum, sizeof(SDnodeEp));
if (tmp == NULL) return; if (tmp == NULL) return;
tfree(eps->dnodeList); tfree(tsDeps.dnodeList);
eps->dnodeList = tmp; tsDeps.dnodeList = tmp;
eps->dnodeNum = data->dnodeNum; tsDeps.dnodeNum = data->dnodeNum;
memcpy(eps->dnodeList, data->dnodeEps, eps->dnodeNum * sizeof(SDnodeEp)); memcpy(tsDeps.dnodeList, data->dnodeEps, tsDeps.dnodeNum * sizeof(SDnodeEp));
dnodePrintEps(eps); dnodePrintEps();
for (int32_t i = 0; i < eps->dnodeNum; ++i) { for (int32_t i = 0; i < tsDeps.dnodeNum; ++i) {
SDnodeEp *ep = &eps->dnodeList[i]; SDnodeEp *ep = &tsDeps.dnodeList[i];
taosHashPut(eps->dnodeHash, &ep->dnodeId, sizeof(int32_t), ep, sizeof(SDnodeEp)); taosHashPut(tsDeps.dnodeHash, &ep->dnodeId, sizeof(int32_t), ep, sizeof(SDnodeEp));
} }
} }
} }
static int32_t dnodeReadEps(SDnEps *eps) { static int32_t dnodeReadEps() {
int32_t len = 0; int32_t len = 0;
int32_t maxLen = 30000; int32_t maxLen = 30000;
char * content = calloc(1, maxLen + 1); char * content = calloc(1, maxLen + 1);
cJSON * root = NULL; cJSON * root = NULL;
FILE * fp = NULL; FILE * fp = NULL;
fp = fopen(eps->file, "r"); fp = fopen(tsDeps.file, "r");
if (!fp) { if (!fp) {
dDebug("file %s not exist", eps->file); dDebug("file %s not exist", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
len = (int32_t)fread(content, 1, maxLen, fp); len = (int32_t)fread(content, 1, maxLen, fp);
if (len <= 0) { if (len <= 0) {
dError("failed to read %s since content is null", eps->file); dError("failed to read %s since content is null", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
content[len] = 0; content[len] = 0;
root = cJSON_Parse(content); root = cJSON_Parse(content);
if (root == NULL) { if (root == NULL) {
dError("failed to read %s since invalid json format", eps->file); dError("failed to read %s since invalid json format", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
cJSON *dnodeNum = cJSON_GetObjectItem(root, "dnodeNum"); cJSON *dnodeNum = cJSON_GetObjectItem(root, "dnodeNum");
if (!dnodeNum || dnodeNum->type != cJSON_Number) { if (!dnodeNum || dnodeNum->type != cJSON_Number) {
dError("failed to read %s since dnodeNum not found", eps->file); dError("failed to read %s since dnodeNum not found", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
cJSON *dnodeInfos = cJSON_GetObjectItem(root, "dnodeInfos"); cJSON *dnodeInfos = cJSON_GetObjectItem(root, "dnodeInfos");
if (!dnodeInfos || dnodeInfos->type != cJSON_Array) { if (!dnodeInfos || dnodeInfos->type != cJSON_Array) {
dError("failed to read %s since dnodeInfos not found", eps->file); dError("failed to read %s since dnodeInfos not found", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
int32_t dnodeInfosSize = cJSON_GetArraySize(dnodeInfos); int32_t dnodeInfosSize = cJSON_GetArraySize(dnodeInfos);
if (dnodeInfosSize != dnodeNum->valueint) { if (dnodeInfosSize != dnodeNum->valueint) {
dError("failed to read %s since dnodeInfos size:%d not matched dnodeNum:%d", eps->file, dnodeInfosSize, dError("failed to read %s since dnodeInfos size:%d not matched dnodeNum:%d", tsDeps.file, dnodeInfosSize,
(int32_t)dnodeNum->valueint); (int32_t)dnodeNum->valueint);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
eps->dnodeNum = dnodeInfosSize; tsDeps.dnodeNum = dnodeInfosSize;
eps->dnodeList = calloc(dnodeInfosSize, sizeof(SDnodeEp)); tsDeps.dnodeList = calloc(dnodeInfosSize, sizeof(SDnodeEp));
if (eps->dnodeList == NULL) { if (tsDeps.dnodeList == NULL) {
dError("failed to calloc dnodeEpList since %s", strerror(errno)); dError("failed to calloc dnodeEpList since %s", strerror(errno));
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
...@@ -104,40 +114,40 @@ static int32_t dnodeReadEps(SDnEps *eps) { ...@@ -104,40 +114,40 @@ static int32_t dnodeReadEps(SDnEps *eps) {
cJSON *dnodeInfo = cJSON_GetArrayItem(dnodeInfos, i); cJSON *dnodeInfo = cJSON_GetArrayItem(dnodeInfos, i);
if (dnodeInfo == NULL) break; if (dnodeInfo == NULL) break;
SDnodeEp *ep = &eps->dnodeList[i]; SDnodeEp *ep = &tsDeps.dnodeList[i];
cJSON *dnodeId = cJSON_GetObjectItem(dnodeInfo, "dnodeId"); cJSON *dnodeId = cJSON_GetObjectItem(dnodeInfo, "dnodeId");
if (!dnodeId || dnodeId->type != cJSON_Number) { if (!dnodeId || dnodeId->type != cJSON_Number) {
dError("failed to read %s, dnodeId not found", eps->file); dError("failed to read %s, dnodeId not found", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
ep->dnodeId = (int32_t)dnodeId->valueint; ep->dnodeId = (int32_t)dnodeId->valueint;
cJSON *dnodeFqdn = cJSON_GetObjectItem(dnodeInfo, "dnodeFqdn"); cJSON *dnodeFqdn = cJSON_GetObjectItem(dnodeInfo, "dnodeFqdn");
if (!dnodeFqdn || dnodeFqdn->type != cJSON_String || dnodeFqdn->valuestring == NULL) { if (!dnodeFqdn || dnodeFqdn->type != cJSON_String || dnodeFqdn->valuestring == NULL) {
dError("failed to read %s, dnodeFqdn not found", eps->file); dError("failed to read %s, dnodeFqdn not found", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
tstrncpy(ep->dnodeFqdn, dnodeFqdn->valuestring, TSDB_FQDN_LEN); tstrncpy(ep->dnodeFqdn, dnodeFqdn->valuestring, TSDB_FQDN_LEN);
cJSON *dnodePort = cJSON_GetObjectItem(dnodeInfo, "dnodePort"); cJSON *dnodePort = cJSON_GetObjectItem(dnodeInfo, "dnodePort");
if (!dnodePort || dnodePort->type != cJSON_Number) { if (!dnodePort || dnodePort->type != cJSON_Number) {
dError("failed to read %s, dnodePort not found", eps->file); dError("failed to read %s, dnodePort not found", tsDeps.file);
goto PRASE_EPS_OVER; goto PRASE_EPS_OVER;
} }
ep->dnodePort = (uint16_t)dnodePort->valueint; ep->dnodePort = (uint16_t)dnodePort->valueint;
} }
dInfo("succcessed to read file %s", eps->file); dInfo("succcessed to read file %s", tsDeps.file);
dnodePrintEps(eps); dnodePrintEps();
PRASE_EPS_OVER: PRASE_EPS_OVER:
if (content != NULL) free(content); if (content != NULL) free(content);
if (root != NULL) cJSON_Delete(root); if (root != NULL) cJSON_Delete(root);
if (fp != NULL) fclose(fp); if (fp != NULL) fclose(fp);
if (dnodeIsDnodeEpChanged(eps, eps->dnodeId, tsLocalEp)) { if (dnodeIsDnodeEpChanged(tsDeps.dnodeId, tsLocalEp)) {
dError("dnode:%d, localEp different from %s dnodeEps.json and need reconfigured", eps->dnodeId, tsLocalEp); dError("dnode:%d, localEp different from %s dnodeEps.json and need reconfigured", tsDeps.dnodeId, tsLocalEp);
return -1; return -1;
} }
...@@ -145,10 +155,10 @@ PRASE_EPS_OVER: ...@@ -145,10 +155,10 @@ PRASE_EPS_OVER:
return 0; return 0;
} }
static int32_t dnodeWriteEps(SDnEps *eps) { static int32_t dnodeWriteEps() {
FILE *fp = fopen(eps->file, "w"); FILE *fp = fopen(tsDeps.file, "w");
if (!fp) { if (!fp) {
dError("failed to write %s since %s", eps->file, strerror(errno)); dError("failed to write %s since %s", tsDeps.file, strerror(errno));
return -1; return -1;
} }
...@@ -157,14 +167,14 @@ static int32_t dnodeWriteEps(SDnEps *eps) { ...@@ -157,14 +167,14 @@ static int32_t dnodeWriteEps(SDnEps *eps) {
char * content = calloc(1, maxLen + 1); char * content = calloc(1, maxLen + 1);
len += snprintf(content + len, maxLen - len, "{\n"); len += snprintf(content + len, maxLen - len, "{\n");
len += snprintf(content + len, maxLen - len, " \"dnodeNum\": %d,\n", eps->dnodeNum); len += snprintf(content + len, maxLen - len, " \"dnodeNum\": %d,\n", tsDeps.dnodeNum);
len += snprintf(content + len, maxLen - len, " \"dnodeInfos\": [{\n"); len += snprintf(content + len, maxLen - len, " \"dnodeInfos\": [{\n");
for (int32_t i = 0; i < eps->dnodeNum; ++i) { for (int32_t i = 0; i < tsDeps.dnodeNum; ++i) {
SDnodeEp *ep = &eps->dnodeList[i]; SDnodeEp *ep = &tsDeps.dnodeList[i];
len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", ep->dnodeId); len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", ep->dnodeId);
len += snprintf(content + len, maxLen - len, " \"dnodeFqdn\": \"%s\",\n", ep->dnodeFqdn); len += snprintf(content + len, maxLen - len, " \"dnodeFqdn\": \"%s\",\n", ep->dnodeFqdn);
len += snprintf(content + len, maxLen - len, " \"dnodePort\": %u\n", ep->dnodePort); len += snprintf(content + len, maxLen - len, " \"dnodePort\": %u\n", ep->dnodePort);
if (i < eps->dnodeNum - 1) { if (i < tsDeps.dnodeNum - 1) {
len += snprintf(content + len, maxLen - len, " },{\n"); len += snprintf(content + len, maxLen - len, " },{\n");
} else { } else {
len += snprintf(content + len, maxLen - len, " }]\n"); len += snprintf(content + len, maxLen - len, " }]\n");
...@@ -178,24 +188,20 @@ static int32_t dnodeWriteEps(SDnEps *eps) { ...@@ -178,24 +188,20 @@ static int32_t dnodeWriteEps(SDnEps *eps) {
free(content); free(content);
terrno = 0; terrno = 0;
dInfo("successed to write %s", eps->file); dInfo("successed to write %s", tsDeps.file);
return 0; return 0;
} }
int32_t dnodeInitEps(SDnEps **out) { int32_t dnodeInitEps() {
SDnEps *eps = calloc(1, sizeof(SDnEps)); tsDeps.dnodeHash = taosHashInit(4, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK);
if (eps == NULL) return -1; if (tsDeps.dnodeHash == NULL) return -1;
eps->dnodeHash = taosHashInit(4, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK); tsDeps.dnodeId = dnodeGetDnodeId();
if (eps->dnodeHash == NULL) return -1; tsDeps.dnodeNum = 0;
snprintf(tsDeps.file, sizeof(tsDeps.file), "%s/dnodeEps.json", tsDnodeDir);
pthread_mutex_init(&tsDeps.mutex, NULL);
eps->dnodeId = dnodeInst()->cfg->dnodeId; int32_t ret = dnodeReadEps();
eps->dnodeNum = 0;
snprintf(eps->file, sizeof(eps->file), "%s/dnodeEps.json", tsDnodeDir);
pthread_mutex_init(&eps->mutex, NULL);
*out = eps;
int32_t ret = dnodeReadEps(eps);
if (ret == 0) { if (ret == 0) {
dInfo("dnode eps is initialized"); dInfo("dnode eps is initialized");
} }
...@@ -203,29 +209,25 @@ int32_t dnodeInitEps(SDnEps **out) { ...@@ -203,29 +209,25 @@ int32_t dnodeInitEps(SDnEps **out) {
return ret; return ret;
} }
void dnodeCleanupEps(SDnEps **out) { void dnodeCleanupEps() {
SDnEps *eps = *out; pthread_mutex_lock(&tsDeps.mutex);
*out = NULL;
pthread_mutex_lock(&eps->mutex);
if (eps->dnodeList != NULL) { if (tsDeps.dnodeList != NULL) {
free(eps->dnodeList); free(tsDeps.dnodeList);
eps->dnodeList = NULL; tsDeps.dnodeList = NULL;
} }
if (eps->dnodeHash) { if (tsDeps.dnodeHash) {
taosHashCleanup(eps->dnodeHash); taosHashCleanup(tsDeps.dnodeHash);
eps->dnodeHash = NULL; tsDeps.dnodeHash = NULL;
} }
eps->dnodeNum = 0; tsDeps.dnodeNum = 0;
pthread_mutex_unlock(&eps->mutex); pthread_mutex_unlock(&tsDeps.mutex);
pthread_mutex_destroy(&eps->mutex); pthread_mutex_destroy(&tsDeps.mutex);
free(eps);
} }
void dnodeUpdateEps(SDnEps *eps, SDnodeEps *data) { void dnodeUpdateEps(SDnodeEps *data) {
if (data == NULL || data->dnodeNum <= 0) return; if (data == NULL || data->dnodeNum <= 0) return;
data->dnodeNum = htonl(data->dnodeNum); data->dnodeNum = htonl(data->dnodeNum);
...@@ -234,28 +236,28 @@ void dnodeUpdateEps(SDnEps *eps, SDnodeEps *data) { ...@@ -234,28 +236,28 @@ void dnodeUpdateEps(SDnEps *eps, SDnodeEps *data) {
data->dnodeEps[i].dnodePort = htons(data->dnodeEps[i].dnodePort); data->dnodeEps[i].dnodePort = htons(data->dnodeEps[i].dnodePort);
} }
pthread_mutex_lock(&eps->mutex); pthread_mutex_lock(&tsDeps.mutex);
if (data->dnodeNum != eps->dnodeNum) { if (data->dnodeNum != tsDeps.dnodeNum) {
dnodeResetEps(eps, data); dnodeResetEps(data);
dnodeWriteEps(eps); dnodeWriteEps();
} else { } else {
int32_t size = data->dnodeNum * sizeof(SDnodeEp); int32_t size = data->dnodeNum * sizeof(SDnodeEp);
if (memcmp(eps->dnodeList, data->dnodeEps, size) != 0) { if (memcmp(tsDeps.dnodeList, data->dnodeEps, size) != 0) {
dnodeResetEps(eps, data); dnodeResetEps(data);
dnodeWriteEps(eps); dnodeWriteEps();
} }
} }
pthread_mutex_unlock(&eps->mutex); pthread_mutex_unlock(&tsDeps.mutex);
} }
bool dnodeIsDnodeEpChanged(SDnEps *eps, int32_t dnodeId, char *epstr) { bool dnodeIsDnodeEpChanged(int32_t dnodeId, char *epstr) {
bool changed = false; bool changed = false;
pthread_mutex_lock(&eps->mutex); pthread_mutex_lock(&tsDeps.mutex);
SDnodeEp *ep = taosHashGet(eps->dnodeHash, &dnodeId, sizeof(int32_t)); SDnodeEp *ep = taosHashGet(tsDeps.dnodeHash, &dnodeId, sizeof(int32_t));
if (ep != NULL) { if (ep != NULL) {
char epSaved[TSDB_EP_LEN + 1]; char epSaved[TSDB_EP_LEN + 1];
snprintf(epSaved, TSDB_EP_LEN, "%s:%u", ep->dnodeFqdn, ep->dnodePort); snprintf(epSaved, TSDB_EP_LEN, "%s:%u", ep->dnodeFqdn, ep->dnodePort);
...@@ -263,21 +265,20 @@ bool dnodeIsDnodeEpChanged(SDnEps *eps, int32_t dnodeId, char *epstr) { ...@@ -263,21 +265,20 @@ bool dnodeIsDnodeEpChanged(SDnEps *eps, int32_t dnodeId, char *epstr) {
tstrncpy(epstr, epSaved, TSDB_EP_LEN); tstrncpy(epstr, epSaved, TSDB_EP_LEN);
} }
pthread_mutex_unlock(&eps->mutex); pthread_mutex_unlock(&tsDeps.mutex);
return changed; return changed;
} }
void dnodeGetDnodeEp(int32_t dnodeId, char *epstr, char *fqdn, uint16_t *port) { void dnodeGetDnodeEp(int32_t dnodeId, char *epstr, char *fqdn, uint16_t *port) {
SDnEps *eps = dnodeInst()->eps; pthread_mutex_lock(&tsDeps.mutex);
pthread_mutex_lock(&eps->mutex);
SDnodeEp *ep = taosHashGet(eps->dnodeHash, &dnodeId, sizeof(int32_t)); SDnodeEp *ep = taosHashGet(tsDeps.dnodeHash, &dnodeId, sizeof(int32_t));
if (ep != NULL) { if (ep != NULL) {
if (port) *port = ep->dnodePort; if (port) *port = ep->dnodePort;
if (fqdn) tstrncpy(fqdn, ep->dnodeFqdn, TSDB_FQDN_LEN); if (fqdn) tstrncpy(fqdn, ep->dnodeFqdn, TSDB_FQDN_LEN);
if (epstr) snprintf(epstr, TSDB_EP_LEN, "%s:%u", ep->dnodeFqdn, ep->dnodePort); if (epstr) snprintf(epstr, TSDB_EP_LEN, "%s:%u", ep->dnodeFqdn, ep->dnodePort);
} }
pthread_mutex_unlock(&eps->mutex); pthread_mutex_unlock(&tsDeps.mutex);
} }
...@@ -33,10 +33,7 @@ ...@@ -33,10 +33,7 @@
#include "mnode.h" #include "mnode.h"
#include "vnode.h" #include "vnode.h"
SDnode *dnodeInst() { static struct SSteps *tsSteps;
static SDnode inst = {0};
return &inst;
}
static int32_t dnodeInitVnodeModule(void **unused) { static int32_t dnodeInitVnodeModule(void **unused) {
SVnodePara para; SVnodePara para;
...@@ -48,58 +45,50 @@ static int32_t dnodeInitVnodeModule(void **unused) { ...@@ -48,58 +45,50 @@ static int32_t dnodeInitVnodeModule(void **unused) {
} }
static int32_t dnodeInitMnodeModule(void **unused) { static int32_t dnodeInitMnodeModule(void **unused) {
SDnode *dnode = dnodeInst();
SMnodePara para; SMnodePara para;
para.fp.GetDnodeEp = dnodeGetDnodeEp; para.fp.GetDnodeEp = dnodeGetDnodeEp;
para.fp.SendMsgToDnode = dnodeSendMsgToDnode; para.fp.SendMsgToDnode = dnodeSendMsgToDnode;
para.fp.SendMsgToMnode = dnodeSendMsgToMnode; para.fp.SendMsgToMnode = dnodeSendMsgToMnode;
para.fp.SendRedirectMsg = dnodeSendRedirectMsg; para.fp.SendRedirectMsg = dnodeSendRedirectMsg;
para.dnodeId = dnode->cfg->dnodeId; dnodeGetCfg(&para.dnodeId, para.clusterId);
strncpy(para.clusterId, dnode->cfg->clusterId, sizeof(para.clusterId));
return mnodeInit(para); return mnodeInit(para);
} }
int32_t dnodeInit() { int32_t dnodeInit() {
struct SSteps *steps = taosStepInit(24, dnodeReportStartup); tsSteps = taosStepInit(24, dnodeReportStartup);
if (steps == NULL) return -1; if (tsSteps == NULL) return -1;
SDnode *dnode = dnodeInst(); taosStepAdd(tsSteps, "dnode-main", dnodeInitMain, dnodeCleanupMain);
taosStepAdd(tsSteps, "dnode-storage", dnodeInitStorage, dnodeCleanupStorage);
//taosStepAdd(tsSteps, "dnode-tfs", tfInit, tfCleanup);
taosStepAdd(tsSteps, "dnode-rpc", rpcInit, rpcCleanup);
taosStepAdd(tsSteps, "dnode-check", dnodeInitCheck, dnodeCleanupCheck);
taosStepAdd(tsSteps, "dnode-cfg", dnodeInitCfg, dnodeCleanupCfg);
taosStepAdd(tsSteps, "dnode-deps", dnodeInitEps, dnodeCleanupEps);
taosStepAdd(tsSteps, "dnode-meps", dnodeInitMnodeEps, dnodeCleanupMnodeEps);
//taosStepAdd(tsSteps, "dnode-wal", walInit, walCleanUp);
//taosStepAdd(tsSteps, "dnode-sync", syncInit, syncCleanUp);
taosStepAdd(tsSteps, "dnode-vnode", dnodeInitVnodeModule, vnodeCleanup);
taosStepAdd(tsSteps, "dnode-mnode", dnodeInitMnodeModule, mnodeCleanup);
taosStepAdd(tsSteps, "dnode-trans", dnodeInitTrans, dnodeCleanupTrans);
taosStepAdd(tsSteps, "dnode-status", dnodeInitStatus, dnodeCleanupStatus);
taosStepAdd(tsSteps, "dnode-telem", dnodeInitTelem, dnodeCleanupTelem);
//taosStepAdd(tsSteps, "dnode-script",scriptEnvPoolInit, scriptEnvPoolCleanup);
taosStepAdd(steps, "dnode-main", (void **)&dnode->main, (InitFp)dnodeInitMain, (CleanupFp)dnodeCleanupMain); taosStepExec(tsSteps);
taosStepAdd(steps, "dnode-storage", NULL, (InitFp)dnodeInitStorage, (CleanupFp)dnodeCleanupStorage);
//taosStepAdd(steps, "dnode-tfs", NULL, (InitFp)tfInit, (CleanupFp)tfCleanup);
taosStepAdd(steps, "dnode-rpc", NULL, (InitFp)rpcInit, (CleanupFp)rpcCleanup);
taosStepAdd(steps, "dnode-check", (void **)&dnode->check, (InitFp)dnodeInitCheck, (CleanupFp)dnodeCleanupCheck);
taosStepAdd(steps, "dnode-cfg", (void **)&dnode->cfg, (InitFp)dnodeInitCfg, (CleanupFp)dnodeCleanupCfg);
taosStepAdd(steps, "dnode-deps", (void **)&dnode->eps, (InitFp)dnodeInitEps, (CleanupFp)dnodeCleanupEps);
taosStepAdd(steps, "dnode-meps", (void **)&dnode->meps, (InitFp)dnodeInitMnodeEps, (CleanupFp)dnodeCleanupMnodeEps);
//taosStepAdd(steps, "dnode-wal", NULL, (InitFp)walInit, (CleanupFp)walCleanUp);
//taosStepAdd(steps, "dnode-sync", NULL, (InitFp)syncInit, (CleanupFp)syncCleanUp);
taosStepAdd(steps, "dnode-vnode", NULL, (InitFp)dnodeInitVnodeModule, (CleanupFp)vnodeCleanup);
taosStepAdd(steps, "dnode-mnode", NULL, (InitFp)dnodeInitMnodeModule, (CleanupFp)mnodeCleanup);
taosStepAdd(steps, "dnode-trans", (void **)&dnode->trans, (InitFp)dnodeInitTrans, (CleanupFp)dnodeCleanupTrans);
taosStepAdd(steps, "dnode-status", (void **)&dnode->status, (InitFp)dnodeInitStatus, (CleanupFp)dnodeCleanupStatus);
taosStepAdd(steps, "dnode-telem", (void **)&dnode->telem, (InitFp)dnodeInitTelem, (CleanupFp)dnodeCleanupTelem);
//taosStepAdd(steps, "dnode-script", NULL, (InitFp)scriptEnvPoolInit, (CleanupFp)scriptEnvPoolCleanup);
dnode->steps = steps; dnodeSetRunStat(TD_RUN_STAT_RUNNING);
taosStepExec(dnode->steps); dnodeReportStartupFinished("TDengine", "initialized successfully");
dInfo("TDengine is initialized successfully");
if (dnode->main) {
dnode->main->runStatus = TD_RUN_STAT_RUNNING;
dnodeReportStartupFinished("TDengine", "initialized successfully");
dInfo("TDengine is initialized successfully");
}
return 0; return 0;
} }
void dnodeCleanup() { void dnodeCleanup() {
SDnode *dnode = dnodeInst(); if (dnodeGetRunStat() != TD_RUN_STAT_STOPPED) {
if (dnode->main->runStatus != TD_RUN_STAT_STOPPED) { dnodeSetRunStat(TD_RUN_STAT_STOPPED);
dnode->main->runStatus = TD_RUN_STAT_STOPPED; taosStepCleanup(tsSteps);
taosStepCleanup(dnode->steps); tsSteps = NULL;
} }
} }
...@@ -17,15 +17,22 @@ ...@@ -17,15 +17,22 @@
#include "os.h" #include "os.h"
#include "tcache.h" #include "tcache.h"
#include "tconfig.h" #include "tconfig.h"
#include "tglobal.h"
#if 0 #if 0
#include "tfs.h" #include "tfs.h"
#endif #endif
#include "tnote.h"
#include "tcompression.h"
#include "ttimer.h"
#include "dnodeCfg.h" #include "dnodeCfg.h"
#include "dnodeMain.h" #include "dnodeMain.h"
#include "mnode.h" #include "mnode.h"
#include "tcompression.h"
#include "tnote.h" static struct {
#include "ttimer.h" RunStat runStatus;
void * dnodeTimer;
SStartupStep startup;
} tsDmain;
static void dnodeCheckDataDirOpenned(char *dir) { static void dnodeCheckDataDirOpenned(char *dir) {
#if 0 #if 0
...@@ -47,27 +54,14 @@ static void dnodeCheckDataDirOpenned(char *dir) { ...@@ -47,27 +54,14 @@ static void dnodeCheckDataDirOpenned(char *dir) {
#endif #endif
} }
void dnodePrintDiskInfo() { int32_t dnodeInitMain() {
dInfo("=================================="); tsDmain.runStatus = TD_RUN_STAT_STOPPED;
dInfo(" os totalDisk: %f(GB)", tsTotalDataDirGB); tsDmain.dnodeTimer = taosTmrInit(100, 200, 60000, "DND-TMR");
dInfo(" os usedDisk: %f(GB)", tsUsedDataDirGB); if (tsDmain.dnodeTimer == NULL) {
dInfo(" os availDisk: %f(GB)", tsAvailDataDirGB);
dInfo("==================================");
}
int32_t dnodeInitMain(SDnMain **out) {
SDnMain* main = calloc(1, sizeof(SDnMain));
if (main == NULL) return -1;
main->runStatus = TD_RUN_STAT_STOPPED;
main->dnodeTimer = taosTmrInit(100, 200, 60000, "DND-TMR");
if (main->dnodeTimer == NULL) {
dError("failed to init dnode timer"); dError("failed to init dnode timer");
return -1; return -1;
} }
*out = main;
tscEmbedded = 1; tscEmbedded = 1;
taosIgnSIGPIPE(); taosIgnSIGPIPE();
taosBlockSIGPIPE(); taosBlockSIGPIPE();
...@@ -76,7 +70,6 @@ int32_t dnodeInitMain(SDnMain **out) { ...@@ -76,7 +70,6 @@ int32_t dnodeInitMain(SDnMain **out) {
taosReadGlobalLogCfg(); taosReadGlobalLogCfg();
taosSetCoreDump(tsEnableCoreFile); taosSetCoreDump(tsEnableCoreFile);
if (!taosMkDir(tsLogDir)) { if (!taosMkDir(tsLogDir)) {
printf("failed to create dir: %s, reason: %s\n", tsLogDir, strerror(errno)); printf("failed to create dir: %s, reason: %s\n", tsLogDir, strerror(errno));
return -1; return -1;
...@@ -101,13 +94,10 @@ int32_t dnodeInitMain(SDnMain **out) { ...@@ -101,13 +94,10 @@ int32_t dnodeInitMain(SDnMain **out) {
return taosCheckGlobalCfg(); return taosCheckGlobalCfg();
} }
void dnodeCleanupMain(SDnMain **out) { void dnodeCleanupMain() {
SDnMain *main = *out; if (tsDmain.dnodeTimer != NULL) {
*out = NULL; taosTmrCleanUp(tsDmain.dnodeTimer);
tsDmain.dnodeTimer = NULL;
if (main->dnodeTimer != NULL) {
taosTmrCleanUp(main->dnodeTimer);
main->dnodeTimer = NULL;
} }
#if 0 #if 0
...@@ -115,8 +105,6 @@ void dnodeCleanupMain(SDnMain **out) { ...@@ -115,8 +105,6 @@ void dnodeCleanupMain(SDnMain **out) {
#endif #endif
taosCloseLog(); taosCloseLog();
taosStopCacheRefreshWorker(); taosStopCacheRefreshWorker();
free(main);
} }
int32_t dnodeInitStorage() { int32_t dnodeInitStorage() {
...@@ -138,7 +126,7 @@ int32_t dnodeInitStorage() { ...@@ -138,7 +126,7 @@ int32_t dnodeInitStorage() {
} }
strncpy(tsDataDir, TFS_PRIMARY_PATH(), TSDB_FILENAME_LEN); strncpy(tsDataDir, TFS_PRIMARY_PATH(), TSDB_FILENAME_LEN);
#endif #endif
sprintf(tsMnodeDir, "%s/mnode", tsDataDir); sprintf(tsMnodeDir, "%s/mnode", tsDataDir);
sprintf(tsVnodeDir, "%s/vnode", tsDataDir); sprintf(tsVnodeDir, "%s/vnode", tsDataDir);
sprintf(tsDnodeDir, "%s/dnode", tsDataDir); sprintf(tsDnodeDir, "%s/dnode", tsDataDir);
...@@ -164,7 +152,6 @@ int32_t dnodeInitStorage() { ...@@ -164,7 +152,6 @@ int32_t dnodeInitStorage() {
return -1; return -1;
} }
TDIR *tdir = tfsOpendir("vnode_bak/.staging"); TDIR *tdir = tfsOpendir("vnode_bak/.staging");
bool stagingNotEmpty = tfsReaddir(tdir) != NULL; bool stagingNotEmpty = tfsReaddir(tdir) != NULL;
tfsClosedir(tdir); tfsClosedir(tdir);
...@@ -190,7 +177,7 @@ int32_t dnodeInitStorage() { ...@@ -190,7 +177,7 @@ int32_t dnodeInitStorage() {
} }
void dnodeCleanupStorage() { void dnodeCleanupStorage() {
#if 0 #if 0
// storage destroy // storage destroy
tfsDestroy(); tfsDestroy();
...@@ -202,18 +189,14 @@ void dnodeCleanupStorage() { ...@@ -202,18 +189,14 @@ void dnodeCleanupStorage() {
} }
void dnodeReportStartup(char *name, char *desc) { void dnodeReportStartup(char *name, char *desc) {
SDnode *dnode = dnodeInst(); SStartupStep *startup = &tsDmain.startup;
if (dnode->main != NULL) { tstrncpy(startup->name, name, strlen(startup->name));
SStartupStep *startup = &dnode->main->startup; tstrncpy(startup->desc, desc, strlen(startup->desc));
tstrncpy(startup->name, name, strlen(startup->name)); startup->finished = 0;
tstrncpy(startup->desc, desc, strlen(startup->desc));
startup->finished = 0;
}
} }
void dnodeReportStartupFinished(char *name, char *desc) { void dnodeReportStartupFinished(char *name, char *desc) {
SDnode *dnode = dnodeInst(); SStartupStep *startup = &tsDmain.startup;
SStartupStep *startup = &dnode->main->startup;
tstrncpy(startup->name, name, strlen(startup->name)); tstrncpy(startup->name, name, strlen(startup->name));
tstrncpy(startup->desc, desc, strlen(startup->desc)); tstrncpy(startup->desc, desc, strlen(startup->desc));
startup->finished = 1; startup->finished = 1;
...@@ -222,9 +205,8 @@ void dnodeReportStartupFinished(char *name, char *desc) { ...@@ -222,9 +205,8 @@ void dnodeReportStartupFinished(char *name, char *desc) {
void dnodeProcessStartupReq(SRpcMsg *pMsg) { void dnodeProcessStartupReq(SRpcMsg *pMsg) {
dInfo("startup msg is received, cont:%s", (char *)pMsg->pCont); dInfo("startup msg is received, cont:%s", (char *)pMsg->pCont);
SDnode *dnode = dnodeInst();
SStartupStep *pStep = rpcMallocCont(sizeof(SStartupStep)); SStartupStep *pStep = rpcMallocCont(sizeof(SStartupStep));
memcpy(pStep, &dnode->main->startup, sizeof(SStartupStep)); memcpy(pStep, &tsDmain.startup, sizeof(SStartupStep));
dDebug("startup msg is sent, step:%s desc:%s finished:%d", pStep->name, pStep->desc, pStep->finished); dDebug("startup msg is sent, step:%s desc:%s finished:%d", pStep->name, pStep->desc, pStep->finished);
...@@ -234,12 +216,11 @@ void dnodeProcessStartupReq(SRpcMsg *pMsg) { ...@@ -234,12 +216,11 @@ void dnodeProcessStartupReq(SRpcMsg *pMsg) {
} }
static int32_t dnodeStartMnode(SRpcMsg *pMsg) { static int32_t dnodeStartMnode(SRpcMsg *pMsg) {
SDnode *dnode = dnodeInst();
SCreateMnodeMsg *pCfg = pMsg->pCont; SCreateMnodeMsg *pCfg = pMsg->pCont;
pCfg->dnodeId = htonl(pCfg->dnodeId); pCfg->dnodeId = htonl(pCfg->dnodeId);
if (pCfg->dnodeId != dnode->cfg->dnodeId) { if (pCfg->dnodeId != dnodeGetDnodeId()) {
dDebug("dnode:%d, in create meps msg is not equal with saved dnodeId:%d", pCfg->dnodeId, dDebug("dnode:%d, in create meps msg is not equal with saved dnodeId:%d", pCfg->dnodeId,
dnodeGetDnodeId(dnode->cfg)); dnodeGetDnodeId());
return TSDB_CODE_MND_DNODE_ID_NOT_CONFIGURED; return TSDB_CODE_MND_DNODE_ID_NOT_CONFIGURED;
} }
...@@ -277,4 +258,10 @@ void dnodeProcessConfigDnodeReq(SRpcMsg *pMsg) { ...@@ -277,4 +258,10 @@ void dnodeProcessConfigDnodeReq(SRpcMsg *pMsg) {
rpcSendResponse(&rspMsg); rpcSendResponse(&rspMsg);
rpcFreeCont(pMsg->pCont); rpcFreeCont(pMsg->pCont);
} }
\ No newline at end of file
RunStat dnodeGetRunStat() { return tsDmain.runStatus; }
void dnodeSetRunStat(RunStat stat) { tsDmain.runStatus = stat; }
void* dnodeGetTimer() { return tsDmain.dnodeTimer; }
\ No newline at end of file
...@@ -22,43 +22,51 @@ ...@@ -22,43 +22,51 @@
#include "dnodeMnodeEps.h" #include "dnodeMnodeEps.h"
#include "mnode.h" #include "mnode.h"
static void dnodePrintMnodeEps(SDnMnEps *meps) { static struct {
SRpcEpSet *epset = &meps->mnodeEpSet; SRpcEpSet mnodeEpSet;
SMInfos mnodeInfos;
char file[PATH_MAX + 20];
pthread_mutex_t mutex;
} tsDmeps;
static void dnodePrintMnodeEps() {
SRpcEpSet *epset = &tsDmeps.mnodeEpSet;
dInfo("print mnode eps, num:%d inuse:%d", epset->numOfEps, epset->inUse); dInfo("print mnode eps, num:%d inuse:%d", epset->numOfEps, epset->inUse);
for (int32_t i = 0; i < epset->numOfEps; i++) { for (int32_t i = 0; i < epset->numOfEps; i++) {
dInfo("ep index:%d, %s:%u", i, epset->fqdn[i], epset->port[i]); dInfo("ep index:%d, %s:%u", i, epset->fqdn[i], epset->port[i]);
} }
} }
static void dnodeResetMnodeEps(SDnMnEps *meps, SMInfos *mInfos) { static void dnodeResetMnodeEps(SMInfos *mInfos) {
if (mInfos == NULL || mInfos->mnodeNum == 0) { if (mInfos == NULL || mInfos->mnodeNum == 0) {
meps->mnodeEpSet.numOfEps = 1; tsDmeps.mnodeEpSet.numOfEps = 1;
taosGetFqdnPortFromEp(tsFirst, meps->mnodeEpSet.fqdn[0], &meps->mnodeEpSet.port[0]); taosGetFqdnPortFromEp(tsFirst, tsDmeps.mnodeEpSet.fqdn[0], &tsDmeps.mnodeEpSet.port[0]);
if (strcmp(tsSecond, tsFirst) != 0) { if (strcmp(tsSecond, tsFirst) != 0) {
meps->mnodeEpSet.numOfEps = 2; tsDmeps.mnodeEpSet.numOfEps = 2;
taosGetFqdnPortFromEp(tsSecond, meps->mnodeEpSet.fqdn[1], &meps->mnodeEpSet.port[1]); taosGetFqdnPortFromEp(tsSecond, tsDmeps.mnodeEpSet.fqdn[1], &tsDmeps.mnodeEpSet.port[1]);
} }
dnodePrintMnodeEps(meps); dnodePrintMnodeEps();
return; return;
} }
int32_t size = sizeof(SMInfos); int32_t size = sizeof(SMInfos);
memcpy(&meps->mnodeInfos, mInfos, size); memcpy(&tsDmeps.mnodeInfos, mInfos, size);
meps->mnodeEpSet.inUse = meps->mnodeInfos.inUse; tsDmeps.mnodeEpSet.inUse = tsDmeps.mnodeInfos.inUse;
meps->mnodeEpSet.numOfEps = meps->mnodeInfos.mnodeNum; tsDmeps.mnodeEpSet.numOfEps = tsDmeps.mnodeInfos.mnodeNum;
for (int32_t i = 0; i < meps->mnodeInfos.mnodeNum; i++) { for (int32_t i = 0; i < tsDmeps.mnodeInfos.mnodeNum; i++) {
taosGetFqdnPortFromEp(meps->mnodeInfos.mnodeInfos[i].mnodeEp, meps->mnodeEpSet.fqdn[i], &meps->mnodeEpSet.port[i]); taosGetFqdnPortFromEp(tsDmeps.mnodeInfos.mnodeInfos[i].mnodeEp, tsDmeps.mnodeEpSet.fqdn[i], &tsDmeps.mnodeEpSet.port[i]);
} }
dnodePrintMnodeEps(meps); dnodePrintMnodeEps();
} }
static int32_t dnodeWriteMnodeEps(SDnMnEps *meps) { static int32_t dnodeWriteMnodeEps() {
FILE *fp = fopen(meps->file, "w"); FILE *fp = fopen(tsDmeps.file, "w");
if (!fp) { if (!fp) {
dError("failed to write %s since %s", meps->file, strerror(errno)); dError("failed to write %s since %s", tsDmeps.file, strerror(errno));
return -1; return -1;
} }
...@@ -67,13 +75,13 @@ static int32_t dnodeWriteMnodeEps(SDnMnEps *meps) { ...@@ -67,13 +75,13 @@ static int32_t dnodeWriteMnodeEps(SDnMnEps *meps) {
char * content = calloc(1, maxLen + 1); char * content = calloc(1, maxLen + 1);
len += snprintf(content + len, maxLen - len, "{\n"); len += snprintf(content + len, maxLen - len, "{\n");
len += snprintf(content + len, maxLen - len, " \"inUse\": %d,\n", meps->mnodeInfos.inUse); len += snprintf(content + len, maxLen - len, " \"inUse\": %d,\n", tsDmeps.mnodeInfos.inUse);
len += snprintf(content + len, maxLen - len, " \"nodeNum\": %d,\n", meps->mnodeInfos.mnodeNum); len += snprintf(content + len, maxLen - len, " \"nodeNum\": %d,\n", tsDmeps.mnodeInfos.mnodeNum);
len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n"); len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n");
for (int32_t i = 0; i < meps->mnodeInfos.mnodeNum; i++) { for (int32_t i = 0; i < tsDmeps.mnodeInfos.mnodeNum; i++) {
len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", meps->mnodeInfos.mnodeInfos[i].mnodeId); len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", tsDmeps.mnodeInfos.mnodeInfos[i].mnodeId);
len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", meps->mnodeInfos.mnodeInfos[i].mnodeEp); len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", tsDmeps.mnodeInfos.mnodeInfos[i].mnodeEp);
if (i < meps->mnodeInfos.mnodeNum - 1) { if (i < tsDmeps.mnodeInfos.mnodeNum - 1) {
len += snprintf(content + len, maxLen - len, " },{\n"); len += snprintf(content + len, maxLen - len, " },{\n");
} else { } else {
len += snprintf(content + len, maxLen - len, " }]\n"); len += snprintf(content + len, maxLen - len, " }]\n");
...@@ -87,11 +95,11 @@ static int32_t dnodeWriteMnodeEps(SDnMnEps *meps) { ...@@ -87,11 +95,11 @@ static int32_t dnodeWriteMnodeEps(SDnMnEps *meps) {
free(content); free(content);
terrno = 0; terrno = 0;
dInfo("successed to write %s", meps->file); dInfo("successed to write %s", tsDmeps.file);
return 0; return 0;
} }
static int32_t dnodeReadMnodeEps(SDnMnEps *meps, SDnEps *deps) { static int32_t dnodeReadMnodeEps() {
int32_t len = 0; int32_t len = 0;
int32_t maxLen = 2000; int32_t maxLen = 2000;
char * content = calloc(1, maxLen + 1); char * content = calloc(1, maxLen + 1);
...@@ -100,22 +108,22 @@ static int32_t dnodeReadMnodeEps(SDnMnEps *meps, SDnEps *deps) { ...@@ -100,22 +108,22 @@ static int32_t dnodeReadMnodeEps(SDnMnEps *meps, SDnEps *deps) {
SMInfos mInfos = {0}; SMInfos mInfos = {0};
bool nodeChanged = false; bool nodeChanged = false;
fp = fopen(meps->file, "r"); fp = fopen(tsDmeps.file, "r");
if (!fp) { if (!fp) {
dDebug("file %s not exist", meps->file); dDebug("file %s not exist", tsDmeps.file);
goto PARSE_MINFOS_OVER; goto PARSE_MINFOS_OVER;
} }
len = (int32_t)fread(content, 1, maxLen, fp); len = (int32_t)fread(content, 1, maxLen, fp);
if (len <= 0) { if (len <= 0) {
dError("failed to read %s since content is null", meps->file); dError("failed to read %s since content is null", tsDmeps.file);
goto PARSE_MINFOS_OVER; goto PARSE_MINFOS_OVER;
} }
content[len] = 0; content[len] = 0;
root = cJSON_Parse(content); root = cJSON_Parse(content);
if (root == NULL) { if (root == NULL) {
dError("failed to read %s since invalid json format", meps->file); dError("failed to read %s since invalid json format", tsDmeps.file);
goto PARSE_MINFOS_OVER; goto PARSE_MINFOS_OVER;
} }
...@@ -124,7 +132,7 @@ static int32_t dnodeReadMnodeEps(SDnMnEps *meps, SDnEps *deps) { ...@@ -124,7 +132,7 @@ static int32_t dnodeReadMnodeEps(SDnMnEps *meps, SDnEps *deps) {
dError("failed to read mnodeEpSet.json since inUse not found"); dError("failed to read mnodeEpSet.json since inUse not found");
goto PARSE_MINFOS_OVER; goto PARSE_MINFOS_OVER;
} }
meps->mnodeInfos.inUse = (int8_t)inUse->valueint; tsDmeps.mnodeInfos.inUse = (int8_t)inUse->valueint;
cJSON *nodeNum = cJSON_GetObjectItem(root, "nodeNum"); cJSON *nodeNum = cJSON_GetObjectItem(root, "nodeNum");
if (!nodeNum || nodeNum->type != cJSON_Number) { if (!nodeNum || nodeNum->type != cJSON_Number) {
...@@ -165,11 +173,11 @@ static int32_t dnodeReadMnodeEps(SDnMnEps *meps, SDnEps *deps) { ...@@ -165,11 +173,11 @@ static int32_t dnodeReadMnodeEps(SDnMnEps *meps, SDnEps *deps) {
mInfo->mnodeId = (int32_t)nodeId->valueint; mInfo->mnodeId = (int32_t)nodeId->valueint;
tstrncpy(mInfo->mnodeEp, nodeEp->valuestring, TSDB_EP_LEN); tstrncpy(mInfo->mnodeEp, nodeEp->valuestring, TSDB_EP_LEN);
bool changed = dnodeIsDnodeEpChanged(deps, mInfo->mnodeId, mInfo->mnodeEp); bool changed = dnodeIsDnodeEpChanged(mInfo->mnodeId, mInfo->mnodeEp);
if (changed) nodeChanged = changed; if (changed) nodeChanged = changed;
} }
dInfo("successed to read file %s", meps->file); dInfo("successed to read file %s", tsDmeps.file);
PARSE_MINFOS_OVER: PARSE_MINFOS_OVER:
if (content != NULL) free(content); if (content != NULL) free(content);
...@@ -182,25 +190,24 @@ PARSE_MINFOS_OVER: ...@@ -182,25 +190,24 @@ PARSE_MINFOS_OVER:
dnodeGetDnodeEp(mInfo->mnodeId, mInfo->mnodeEp, NULL, NULL); dnodeGetDnodeEp(mInfo->mnodeId, mInfo->mnodeEp, NULL, NULL);
} }
dnodeResetMnodeEps(meps, &mInfos); dnodeResetMnodeEps(&mInfos);
if (nodeChanged) { if (nodeChanged) {
dnodeWriteMnodeEps(meps); dnodeWriteMnodeEps();
} }
return 0; return 0;
} }
void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) { void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) {
SDnMnEps *meps = dnodeInst()->meps;
SRpcConnInfo connInfo = {0}; SRpcConnInfo connInfo = {0};
rpcGetConnInfo(rpcMsg->handle, &connInfo); rpcGetConnInfo(rpcMsg->handle, &connInfo);
SRpcEpSet epSet = {0}; SRpcEpSet epSet = {0};
if (forShell) { if (forShell) {
dnodeGetEpSetForShell(meps, &epSet); dnodeGetEpSetForShell(&epSet);
} else { } else {
dnodeGetEpSetForPeer(meps, &epSet); dnodeGetEpSetForPeer(&epSet);
} }
dDebug("msg:%s will be redirected, dnodeIp:%s user:%s, numOfEps:%d inUse:%d", taosMsg[rpcMsg->msgType], dDebug("msg:%s will be redirected, dnodeIp:%s user:%s, numOfEps:%d inUse:%d", taosMsg[rpcMsg->msgType],
...@@ -222,16 +229,12 @@ void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) { ...@@ -222,16 +229,12 @@ void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) {
rpcSendRedirectRsp(rpcMsg->handle, &epSet); rpcSendRedirectRsp(rpcMsg->handle, &epSet);
} }
int32_t dnodeInitMnodeEps(SDnMnEps **out) { int32_t dnodeInitMnodeEps() {
SDnMnEps *meps = calloc(1, sizeof(SDnMnEps)); snprintf(tsDmeps.file, sizeof(tsDmeps.file), "%s/mnodeEpSet.json", tsDnodeDir);
if (meps == NULL) return -1; pthread_mutex_init(&tsDmeps.mutex, NULL);
snprintf(meps->file, sizeof(meps->file), "%s/mnodeEpSet.json", tsDnodeDir);
pthread_mutex_init(&meps->mutex, NULL);
*out = meps;
dnodeResetMnodeEps(meps, NULL); dnodeResetMnodeEps(NULL);
int32_t ret = dnodeReadMnodeEps(meps, dnodeInst()->eps); int32_t ret = dnodeReadMnodeEps();
if (ret == 0) { if (ret == 0) {
dInfo("dnode mInfos is initialized"); dInfo("dnode mInfos is initialized");
} }
...@@ -239,17 +242,11 @@ int32_t dnodeInitMnodeEps(SDnMnEps **out) { ...@@ -239,17 +242,11 @@ int32_t dnodeInitMnodeEps(SDnMnEps **out) {
return ret; return ret;
} }
void dnodeCleanupMnodeEps(SDnMnEps **out) { void dnodeCleanupMnodeEps() {
SDnMnEps *meps = *out; pthread_mutex_destroy(&tsDmeps.mutex);
*out = NULL;
if (meps != NULL) {
pthread_mutex_destroy(&meps->mutex);
free(meps);
}
} }
void dnodeUpdateMnodeFromStatus(SDnMnEps *meps, SMInfos *mInfos) { void dnodeUpdateMnodeFromStatus(SMInfos *mInfos) {
if (mInfos->mnodeNum <= 0 || mInfos->mnodeNum > TSDB_MAX_REPLICA) { if (mInfos->mnodeNum <= 0 || mInfos->mnodeNum > TSDB_MAX_REPLICA) {
dError("invalid mInfos since num:%d invalid", mInfos->mnodeNum); dError("invalid mInfos since num:%d invalid", mInfos->mnodeNum);
return; return;
...@@ -264,53 +261,51 @@ void dnodeUpdateMnodeFromStatus(SDnMnEps *meps, SMInfos *mInfos) { ...@@ -264,53 +261,51 @@ void dnodeUpdateMnodeFromStatus(SDnMnEps *meps, SMInfos *mInfos) {
} }
} }
pthread_mutex_lock(&meps->mutex); pthread_mutex_lock(&tsDmeps.mutex);
if (mInfos->mnodeNum != meps->mnodeInfos.mnodeNum) { if (mInfos->mnodeNum != tsDmeps.mnodeInfos.mnodeNum) {
dnodeResetMnodeEps(meps, mInfos); dnodeResetMnodeEps(mInfos);
dnodeWriteMnodeEps(meps); dnodeWriteMnodeEps();
} else { } else {
int32_t size = sizeof(SMInfos); int32_t size = sizeof(SMInfos);
if (memcmp(mInfos, &meps->mnodeInfos, size) != 0) { if (memcmp(mInfos, &tsDmeps.mnodeInfos, size) != 0) {
dnodeResetMnodeEps(meps, mInfos); dnodeResetMnodeEps(mInfos);
dnodeWriteMnodeEps(meps); dnodeWriteMnodeEps();
} }
} }
pthread_mutex_unlock(&meps->mutex); pthread_mutex_unlock(&tsDmeps.mutex);
} }
void dnodeUpdateMnodeFromPeer(SDnMnEps *meps, SRpcEpSet *ep) { void dnodeUpdateMnodeFromPeer(SRpcEpSet *ep) {
if (ep->numOfEps <= 0) { if (ep->numOfEps <= 0) {
dError("mInfos is changed, but content is invalid, discard it"); dError("mInfos is changed, but content is invalid, discard it");
return; return;
} }
pthread_mutex_lock(&meps->mutex); pthread_mutex_lock(&tsDmeps.mutex);
dInfo("mInfos is changed, numOfEps:%d inUse:%d", ep->numOfEps, ep->inUse); dInfo("mInfos is changed, numOfEps:%d inUse:%d", ep->numOfEps, ep->inUse);
for (int32_t i = 0; i < ep->numOfEps; ++i) { for (int32_t i = 0; i < ep->numOfEps; ++i) {
ep->port[i] -= TSDB_PORT_DNODEDNODE; ep->port[i] -= TSDB_PORT_DNODEDNODE;
dInfo("minfo:%d %s:%u", i, ep->fqdn[i], ep->port[i]); dInfo("minfo:%d %s:%u", i, ep->fqdn[i], ep->port[i]);
} }
meps->mnodeEpSet = *ep; tsDmeps.mnodeEpSet = *ep;
pthread_mutex_unlock(&meps->mutex); pthread_mutex_unlock(&tsDmeps.mutex);
} }
void dnodeGetEpSetForPeer(SDnMnEps *meps, SRpcEpSet *epSet) { void dnodeGetEpSetForPeer(SRpcEpSet *epSet) {
pthread_mutex_lock(&meps->mutex); pthread_mutex_lock(&tsDmeps.mutex);
*epSet = meps->mnodeEpSet; *epSet = tsDmeps.mnodeEpSet;
for (int32_t i = 0; i < epSet->numOfEps; ++i) { for (int32_t i = 0; i < epSet->numOfEps; ++i) {
epSet->port[i] += TSDB_PORT_DNODEDNODE; epSet->port[i] += TSDB_PORT_DNODEDNODE;
} }
pthread_mutex_unlock(&meps->mutex); pthread_mutex_unlock(&tsDmeps.mutex);
} }
void dnodeGetEpSetForShell(SDnMnEps *meps, SRpcEpSet *epSet) { void dnodeGetEpSetForShell(SRpcEpSet *epSet) {
pthread_mutex_lock(&meps->mutex); pthread_mutex_lock(&tsDmeps.mutex);
*epSet = tsDmeps.mnodeEpSet;
*epSet = meps->mnodeEpSet; pthread_mutex_unlock(&tsDmeps.mutex);
pthread_mutex_unlock(&meps->mutex);
} }
...@@ -25,15 +25,15 @@ ...@@ -25,15 +25,15 @@
#include "dnodeMain.h" #include "dnodeMain.h"
#include "vnode.h" #include "vnode.h"
static void dnodeSendStatusMsg(void *handle, void *tmrId) { static struct {
SDnStatus *status = handle; void * dnodeTimer;
if (status->dnodeTimer == NULL) { void * statusTimer;
dError("dnode timer is already released"); uint32_t rebootTime;
return; } tsStatus;
}
if (status->statusTimer == NULL) { static void dnodeSendStatusMsg(void *handle, void *tmrId) {
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, status, status->dnodeTimer, &status->statusTimer); if (tsStatus.statusTimer == NULL) {
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsStatus.dnodeTimer, &tsStatus.statusTimer);
dError("failed to start status timer"); dError("failed to start status timer");
return; return;
} }
...@@ -41,16 +41,15 @@ static void dnodeSendStatusMsg(void *handle, void *tmrId) { ...@@ -41,16 +41,15 @@ static void dnodeSendStatusMsg(void *handle, void *tmrId) {
int32_t contLen = sizeof(SStatusMsg) + TSDB_MAX_VNODES * sizeof(SVnodeLoad); int32_t contLen = sizeof(SStatusMsg) + TSDB_MAX_VNODES * sizeof(SVnodeLoad);
SStatusMsg *pStatus = rpcMallocCont(contLen); SStatusMsg *pStatus = rpcMallocCont(contLen);
if (pStatus == NULL) { if (pStatus == NULL) {
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, status, status->dnodeTimer, &status->statusTimer); taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsStatus.dnodeTimer, &tsStatus.statusTimer);
dError("failed to malloc status message"); dError("failed to malloc status message");
return; return;
} }
SDnode *dnode = dnodeInst(); dnodeGetCfg(&pStatus->dnodeId, pStatus->clusterId);
dnodeGetCfg(dnode->cfg, &pStatus->dnodeId, pStatus->clusterId); pStatus->dnodeId = htonl(dnodeGetDnodeId());
pStatus->dnodeId = htonl(dnodeGetDnodeId(dnode->cfg));
pStatus->version = htonl(tsVersion); pStatus->version = htonl(tsVersion);
pStatus->lastReboot = htonl(status->rebootTime); pStatus->lastReboot = htonl(tsStatus.rebootTime);
pStatus->numOfCores = htons((uint16_t)tsNumOfCores); pStatus->numOfCores = htons((uint16_t)tsNumOfCores);
pStatus->diskAvailable = tsAvailDataDirGB; pStatus->diskAvailable = tsAvailDataDirGB;
pStatus->alternativeRole = tsAlternativeRole; pStatus->alternativeRole = tsAlternativeRole;
...@@ -80,69 +79,58 @@ static void dnodeSendStatusMsg(void *handle, void *tmrId) { ...@@ -80,69 +79,58 @@ static void dnodeSendStatusMsg(void *handle, void *tmrId) {
contLen = sizeof(SStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad); contLen = sizeof(SStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad);
pStatus->openVnodes = htons(pStatus->openVnodes); pStatus->openVnodes = htons(pStatus->openVnodes);
SRpcMsg rpcMsg = {.ahandle = status, .pCont = pStatus, .contLen = contLen, .msgType = TSDB_MSG_TYPE_DM_STATUS}; SRpcMsg rpcMsg = {.ahandle = NULL, .pCont = pStatus, .contLen = contLen, .msgType = TSDB_MSG_TYPE_DM_STATUS};
dnodeSendMsgToMnode(&rpcMsg); dnodeSendMsgToMnode(&rpcMsg);
} }
void dnodeProcessStatusRsp(SRpcMsg *pMsg) { void dnodeProcessStatusRsp(SRpcMsg *pMsg) {
SDnode *dnode = dnodeInst();
SDnStatus *status = pMsg->ahandle;
if (pMsg->code != TSDB_CODE_SUCCESS) { if (pMsg->code != TSDB_CODE_SUCCESS) {
dError("status rsp is received, error:%s", tstrerror(pMsg->code)); dError("status rsp is received, error:%s", tstrerror(pMsg->code));
if (pMsg->code == TSDB_CODE_MND_DNODE_NOT_EXIST) { if (pMsg->code == TSDB_CODE_MND_DNODE_NOT_EXIST) {
char clusterId[TSDB_CLUSTER_ID_LEN]; char clusterId[TSDB_CLUSTER_ID_LEN];
dnodeGetClusterId(dnode->cfg, clusterId); dnodeGetClusterId(clusterId);
if (clusterId[0] != '\0') { if (clusterId[0] != '\0') {
dnodeSetDropped(dnode->cfg); dnodeSetDropped();
dError("exit zombie dropped dnode"); dError("exit zombie dropped dnode");
exit(EXIT_FAILURE); exit(EXIT_FAILURE);
} }
} }
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, status, status->dnodeTimer, &status->statusTimer); taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsStatus.dnodeTimer, &tsStatus.statusTimer);
return; return;
} }
SStatusRsp *pStatusRsp = pMsg->pCont; SStatusRsp *pStatusRsp = pMsg->pCont;
SMInfos * minfos = &pStatusRsp->mnodes; SMInfos * minfos = &pStatusRsp->mnodes;
dnodeUpdateMnodeFromStatus(dnode->meps, minfos); dnodeUpdateMnodeFromStatus(minfos);
SDnodeCfg *pCfg = &pStatusRsp->dnodeCfg; SDnodeCfg *pCfg = &pStatusRsp->dnodeCfg;
pCfg->numOfVnodes = htonl(pCfg->numOfVnodes); pCfg->numOfVnodes = htonl(pCfg->numOfVnodes);
pCfg->moduleStatus = htonl(pCfg->moduleStatus); pCfg->moduleStatus = htonl(pCfg->moduleStatus);
pCfg->dnodeId = htonl(pCfg->dnodeId); pCfg->dnodeId = htonl(pCfg->dnodeId);
dnodeUpdateCfg(dnode->cfg, pCfg); dnodeUpdateCfg(pCfg);
vnodeSetAccess(pStatusRsp->vgAccess, pCfg->numOfVnodes); vnodeSetAccess(pStatusRsp->vgAccess, pCfg->numOfVnodes);
SDnodeEps *pEps = (SDnodeEps *)((char *)pStatusRsp->vgAccess + pCfg->numOfVnodes * sizeof(SVgroupAccess)); SDnodeEps *pEps = (SDnodeEps *)((char *)pStatusRsp->vgAccess + pCfg->numOfVnodes * sizeof(SVgroupAccess));
dnodeUpdateEps(dnode->eps, pEps); dnodeUpdateEps(pEps);
taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, status, status->dnodeTimer, &status->statusTimer); taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsStatus.dnodeTimer, &tsStatus.statusTimer);
} }
int32_t dnodeInitStatus(SDnStatus **out) { int32_t dnodeInitStatus() {
SDnStatus *status = calloc(1, sizeof(SDnStatus)); tsStatus.statusTimer = NULL;
if (status == NULL) return -1; tsStatus.dnodeTimer = dnodeGetTimer();
status->statusTimer = NULL; tsStatus.rebootTime = taosGetTimestampSec();
status->dnodeTimer = dnodeInst()->main->dnodeTimer; taosTmrReset(dnodeSendStatusMsg, 500, NULL, tsStatus.dnodeTimer, &tsStatus.statusTimer);
status->rebootTime = taosGetTimestampSec();
taosTmrReset(dnodeSendStatusMsg, 500, status, status->dnodeTimer, &status->statusTimer);
*out = status;
dInfo("dnode status timer is initialized"); dInfo("dnode status timer is initialized");
return TSDB_CODE_SUCCESS; return TSDB_CODE_SUCCESS;
} }
void dnodeCleanupStatus(SDnStatus **out) { void dnodeCleanupStatus() {
SDnStatus *status = *out; if (tsStatus.statusTimer != NULL) {
*out = NULL; taosTmrStopA(&tsStatus.statusTimer);
tsStatus.statusTimer = NULL;
if (status->statusTimer != NULL) {
taosTmrStopA(&status->statusTimer);
status->statusTimer = NULL;
} }
free(status);
} }
...@@ -25,6 +25,19 @@ ...@@ -25,6 +25,19 @@
#define TELEMETRY_PORT 80 #define TELEMETRY_PORT 80
#define REPORT_INTERVAL 86400 #define REPORT_INTERVAL 86400
/*
* sem_timedwait is NOT implemented on MacOSX
* thus we use pthread_mutex_t/pthread_cond_t to simulate
*/
static struct {
bool enable;
pthread_mutex_t lock;
pthread_cond_t cond;
volatile int32_t exit;
pthread_t thread;
char email[TSDB_FQDN_LEN];
} tsTelem;
static void dnodeBeginObject(SBufferWriter* bw) { tbufWriteChar(bw, '{'); } static void dnodeBeginObject(SBufferWriter* bw) { tbufWriteChar(bw, '{'); }
static void dnodeCloseObject(SBufferWriter* bw) { static void dnodeCloseObject(SBufferWriter* bw) {
...@@ -154,14 +167,14 @@ static void dnodeAddMemoryInfo(SBufferWriter* bw) { ...@@ -154,14 +167,14 @@ static void dnodeAddMemoryInfo(SBufferWriter* bw) {
fclose(fp); fclose(fp);
} }
static void dnodeAddVersionInfo(SDnTelem* telem, SBufferWriter* bw) { static void dnodeAddVersionInfo(SBufferWriter* bw) {
dnodeAddStringField(bw, "version", version); dnodeAddStringField(bw, "version", version);
dnodeAddStringField(bw, "buildInfo", buildinfo); dnodeAddStringField(bw, "buildInfo", buildinfo);
dnodeAddStringField(bw, "gitInfo", gitinfo); dnodeAddStringField(bw, "gitInfo", gitinfo);
dnodeAddStringField(bw, "email", telem->email); dnodeAddStringField(bw, "email", tsTelem.email);
} }
static void dnodeAddRuntimeInfo(SDnTelem* telem, SBufferWriter* bw) { static void dnodeAddRuntimeInfo(SBufferWriter* bw) {
SMnodeStat stat = {0}; SMnodeStat stat = {0};
if (mnodeGetStatistics(&stat) != 0) { if (mnodeGetStatistics(&stat) != 0) {
return; return;
...@@ -179,7 +192,7 @@ static void dnodeAddRuntimeInfo(SDnTelem* telem, SBufferWriter* bw) { ...@@ -179,7 +192,7 @@ static void dnodeAddRuntimeInfo(SDnTelem* telem, SBufferWriter* bw) {
dnodeAddIntField(bw, "compStorage", stat.compStorage); dnodeAddIntField(bw, "compStorage", stat.compStorage);
} }
static void dnodeSendTelemetryReport(SDnTelem* telem) { static void dnodeSendTelemetryReport() {
char buf[128] = {0}; char buf[128] = {0};
uint32_t ip = taosGetIpv4FromFqdn(TELEMETRY_SERVER); uint32_t ip = taosGetIpv4FromFqdn(TELEMETRY_SERVER);
if (ip == 0xffffffff) { if (ip == 0xffffffff) {
...@@ -192,16 +205,18 @@ static void dnodeSendTelemetryReport(SDnTelem* telem) { ...@@ -192,16 +205,18 @@ static void dnodeSendTelemetryReport(SDnTelem* telem) {
return; return;
} }
SDnode *dnode = dnodeInst(); char clusterId[TSDB_CLUSTER_ID_LEN] = {0};
dnodeGetClusterId(clusterId);
SBufferWriter bw = tbufInitWriter(NULL, false); SBufferWriter bw = tbufInitWriter(NULL, false);
dnodeBeginObject(&bw); dnodeBeginObject(&bw);
dnodeAddStringField(&bw, "instanceId", dnode->cfg->clusterId); dnodeAddStringField(&bw, "instanceId", clusterId);
dnodeAddIntField(&bw, "reportVersion", 1); dnodeAddIntField(&bw, "reportVersion", 1);
dnodeAddOsInfo(&bw); dnodeAddOsInfo(&bw);
dnodeAddCpuInfo(&bw); dnodeAddCpuInfo(&bw);
dnodeAddMemoryInfo(&bw); dnodeAddMemoryInfo(&bw);
dnodeAddVersionInfo(telem, &bw); dnodeAddVersionInfo(&bw);
dnodeAddRuntimeInfo(telem, &bw); dnodeAddRuntimeInfo(&bw);
dnodeCloseObject(&bw); dnodeCloseObject(&bw);
const char* header = const char* header =
...@@ -227,25 +242,23 @@ static void dnodeSendTelemetryReport(SDnTelem* telem) { ...@@ -227,25 +242,23 @@ static void dnodeSendTelemetryReport(SDnTelem* telem) {
} }
static void* dnodeTelemThreadFp(void* param) { static void* dnodeTelemThreadFp(void* param) {
SDnTelem* telem = param;
struct timespec end = {0}; struct timespec end = {0};
clock_gettime(CLOCK_REALTIME, &end); clock_gettime(CLOCK_REALTIME, &end);
end.tv_sec += 300; // wait 5 minutes before send first report end.tv_sec += 300; // wait 5 minutes before send first report
setThreadName("dnode-telem"); setThreadName("dnode-telem");
while (!telem->exit) { while (!tsTelem.exit) {
int32_t r = 0; int32_t r = 0;
struct timespec ts = end; struct timespec ts = end;
pthread_mutex_lock(&telem->lock); pthread_mutex_lock(&tsTelem.lock);
r = pthread_cond_timedwait(&telem->cond, &telem->lock, &ts); r = pthread_cond_timedwait(&tsTelem.cond, &tsTelem.lock, &ts);
pthread_mutex_unlock(&telem->lock); pthread_mutex_unlock(&tsTelem.lock);
if (r == 0) break; if (r == 0) break;
if (r != ETIMEDOUT) continue; if (r != ETIMEDOUT) continue;
if (mnodeIsServing()) { if (mnodeIsServing()) {
dnodeSendTelemetryReport(telem); dnodeSendTelemetryReport();
} }
end.tv_sec += REPORT_INTERVAL; end.tv_sec += REPORT_INTERVAL;
} }
...@@ -253,40 +266,35 @@ static void* dnodeTelemThreadFp(void* param) { ...@@ -253,40 +266,35 @@ static void* dnodeTelemThreadFp(void* param) {
return NULL; return NULL;
} }
static void dnodeGetEmail(SDnTelem* telem, char* filepath) { static void dnodeGetEmail(char* filepath) {
int32_t fd = taosOpenFileRead(filepath); int32_t fd = taosOpenFileRead(filepath);
if (fd < 0) { if (fd < 0) {
return; return;
} }
if (taosReadFile(fd, (void*)telem->email, TSDB_FQDN_LEN) < 0) { if (taosReadFile(fd, (void*)tsTelem.email, TSDB_FQDN_LEN) < 0) {
dError("failed to read %d bytes from file %s since %s", TSDB_FQDN_LEN, filepath, strerror(errno)); dError("failed to read %d bytes from file %s since %s", TSDB_FQDN_LEN, filepath, strerror(errno));
} }
taosCloseFile(fd); taosCloseFile(fd);
} }
int32_t dnodeInitTelem(SDnTelem** out) { int32_t dnodeInitTelem() {
SDnTelem* telem = calloc(1, sizeof(SDnTelem)); tsTelem.enable = tsEnableTelemetryReporting;
if (telem == NULL) return -1; if (!tsTelem.enable) return 0;
telem->enable = tsEnableTelemetryReporting;
*out = telem;
if (!telem->enable) return 0; tsTelem.exit = 0;
pthread_mutex_init(&tsTelem.lock, NULL);
pthread_cond_init(&tsTelem.cond, NULL);
tsTelem.email[0] = 0;
telem->exit = 0; dnodeGetEmail("/usr/local/taos/email");
pthread_mutex_init(&telem->lock, NULL);
pthread_cond_init(&telem->cond, NULL);
telem->email[0] = 0;
dnodeGetEmail(telem, "/usr/local/taos/email");
pthread_attr_t attr; pthread_attr_t attr;
pthread_attr_init(&attr); pthread_attr_init(&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE); pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE);
int32_t code = pthread_create(&telem->thread, &attr, dnodeTelemThreadFp, telem); int32_t code = pthread_create(&tsTelem.thread, &attr, dnodeTelemThreadFp, NULL);
pthread_attr_destroy(&attr); pthread_attr_destroy(&attr);
if (code != 0) { if (code != 0) {
dTrace("failed to create telemetry thread since :%s", strerror(code)); dTrace("failed to create telemetry thread since :%s", strerror(code));
...@@ -296,26 +304,18 @@ int32_t dnodeInitTelem(SDnTelem** out) { ...@@ -296,26 +304,18 @@ int32_t dnodeInitTelem(SDnTelem** out) {
return 0; return 0;
} }
void dnodeCleanupTelem(SDnTelem** out) { void dnodeCleanupTelem() {
SDnTelem* telem = *out; if (!tsTelem.enable) return;
*out = NULL;
if (!telem->enable) { if (taosCheckPthreadValid(tsTelem.thread)) {
free(telem); pthread_mutex_lock(&tsTelem.lock);
return; tsTelem.exit = 1;
} pthread_cond_signal(&tsTelem.cond);
pthread_mutex_unlock(&tsTelem.lock);
if (taosCheckPthreadValid(telem->thread)) { pthread_join(tsTelem.thread, NULL);
pthread_mutex_lock(&telem->lock);
telem->exit = 1;
pthread_cond_signal(&telem->cond);
pthread_mutex_unlock(&telem->lock);
pthread_join(telem->thread, NULL);
} }
pthread_mutex_destroy(&telem->lock); pthread_mutex_destroy(&tsTelem.lock);
pthread_cond_destroy(&telem->cond); pthread_cond_destroy(&tsTelem.cond);
free(telem);
} }
...@@ -14,7 +14,7 @@ ...@@ -14,7 +14,7 @@
*/ */
/* this file is mainly responsible for the communication between DNODEs. Each /* this file is mainly responsible for the communication between DNODEs. Each
* dnode works as both server and client. SDnode may send status, grant, config * dnode works as both server and client. Dnode may send status, grant, config
* messages to mnode, mnode may send create/alter/drop table/vnode messages * messages to mnode, mnode may send create/alter/drop table/vnode messages
* to dnode. All theses messages are handled from here * to dnode. All theses messages are handled from here
*/ */
...@@ -29,8 +29,19 @@ ...@@ -29,8 +29,19 @@
#include "vnode.h" #include "vnode.h"
#include "mnode.h" #include "mnode.h"
typedef void (*RpcMsgFp)( SRpcMsg *pMsg);
static struct {
void * serverRpc;
void * clientRpc;
void * shellRpc;
int32_t queryReqNum;
int32_t submitReqNum;
RpcMsgFp peerMsgFp[TSDB_MSG_TYPE_MAX];
RpcMsgFp shellMsgFp[TSDB_MSG_TYPE_MAX];
} tsTrans;
static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
SDnode * dnode = dnodeInst();
SRpcMsg rspMsg = {.handle = pMsg->handle, .pCont = NULL, .contLen = 0}; SRpcMsg rspMsg = {.handle = pMsg->handle, .pCont = NULL, .contLen = 0};
if (pMsg->pCont == NULL) return; if (pMsg->pCont == NULL) return;
...@@ -39,7 +50,7 @@ static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { ...@@ -39,7 +50,7 @@ static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
return; return;
} }
if (dnode->main->runStatus != TD_RUN_STAT_RUNNING) { if (dnodeGetRunStat() != TD_RUN_STAT_RUNNING) {
rspMsg.code = TSDB_CODE_APP_NOT_READY; rspMsg.code = TSDB_CODE_APP_NOT_READY;
rpcSendResponse(&rspMsg); rpcSendResponse(&rspMsg);
rpcFreeCont(pMsg->pCont); rpcFreeCont(pMsg->pCont);
...@@ -53,7 +64,7 @@ static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { ...@@ -53,7 +64,7 @@ static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
return; return;
} }
RpcMsgFp fp = dnode->trans->peerMsgFp[pMsg->msgType]; RpcMsgFp fp = tsTrans.peerMsgFp[pMsg->msgType];
if (fp != NULL) { if (fp != NULL) {
(*fp)(pMsg); (*fp)(pMsg);
} else { } else {
...@@ -64,27 +75,27 @@ static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { ...@@ -64,27 +75,27 @@ static void dnodeProcessPeerReq(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
} }
} }
int32_t dnodeInitServer(SDnTrans *trans) { int32_t dnodeInitServer() {
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeProcessConfigDnodeReq; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeProcessConfigDnodeReq;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeProcessCreateMnodeReq; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeProcessCreateMnodeReq;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_TABLE] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_TABLE] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_VNODE] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_VNODE] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_AUTH] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_AUTH] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_GRANT] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_GRANT] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_STATUS] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_STATUS] = mnodeProcessMsg;
SRpcInit rpcInit; SRpcInit rpcInit;
memset(&rpcInit, 0, sizeof(rpcInit)); memset(&rpcInit, 0, sizeof(rpcInit));
...@@ -96,8 +107,8 @@ int32_t dnodeInitServer(SDnTrans *trans) { ...@@ -96,8 +107,8 @@ int32_t dnodeInitServer(SDnTrans *trans) {
rpcInit.connType = TAOS_CONN_SERVER; rpcInit.connType = TAOS_CONN_SERVER;
rpcInit.idleTime = tsShellActivityTimer * 1000; rpcInit.idleTime = tsShellActivityTimer * 1000;
trans->serverRpc = rpcOpen(&rpcInit); tsTrans.serverRpc = rpcOpen(&rpcInit);
if (trans->serverRpc == NULL) { if (tsTrans.serverRpc == NULL) {
dError("failed to init peer rpc server"); dError("failed to init peer rpc server");
return -1; return -1;
} }
...@@ -106,17 +117,16 @@ int32_t dnodeInitServer(SDnTrans *trans) { ...@@ -106,17 +117,16 @@ int32_t dnodeInitServer(SDnTrans *trans) {
return 0; return 0;
} }
void dnodeCleanupServer(SDnTrans *trans) { void dnodeCleanupServer() {
if (trans->serverRpc) { if (tsTrans.serverRpc) {
rpcClose(trans->serverRpc); rpcClose(tsTrans.serverRpc);
trans->serverRpc = NULL; tsTrans.serverRpc = NULL;
dInfo("dnode peer server is closed"); dInfo("dnode peer server is closed");
} }
} }
static void dnodeProcessRspFromPeer(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { static void dnodeProcessRspFromPeer(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
SDnode *dnode = dnodeInst(); if (dnodeGetRunStat() == TD_RUN_STAT_STOPPED) {
if (dnode->main->runStatus == TD_RUN_STAT_STOPPED) {
if (pMsg == NULL || pMsg->pCont == NULL) return; if (pMsg == NULL || pMsg->pCont == NULL) return;
dTrace("msg:%p is ignored since dnode is stopping", pMsg); dTrace("msg:%p is ignored since dnode is stopping", pMsg);
rpcFreeCont(pMsg->pCont); rpcFreeCont(pMsg->pCont);
...@@ -124,10 +134,10 @@ static void dnodeProcessRspFromPeer(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { ...@@ -124,10 +134,10 @@ static void dnodeProcessRspFromPeer(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
} }
if (pMsg->msgType == TSDB_MSG_TYPE_DM_STATUS_RSP && pEpSet) { if (pMsg->msgType == TSDB_MSG_TYPE_DM_STATUS_RSP && pEpSet) {
dnodeUpdateMnodeFromPeer(dnode->meps, pEpSet); dnodeUpdateMnodeFromPeer(pEpSet);
} }
RpcMsgFp fp = dnode->trans->peerMsgFp[pMsg->msgType]; RpcMsgFp fp = tsTrans.peerMsgFp[pMsg->msgType];
if (fp != NULL) { if (fp != NULL) {
(*fp)(pMsg); (*fp)(pMsg);
} else { } else {
...@@ -141,27 +151,27 @@ static void dnodeProcessRspFromPeer(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { ...@@ -141,27 +151,27 @@ static void dnodeProcessRspFromPeer(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
rpcFreeCont(pMsg->pCont); rpcFreeCont(pMsg->pCont);
} }
int32_t dnodeInitClient(SDnTrans *trans) { int32_t dnodeInitClient() {
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_TABLE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_TABLE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_VNODE_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_CONFIG_VNODE_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_AUTH_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_AUTH_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_GRANT_RSP] = mnodeProcessMsg; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_GRANT_RSP] = mnodeProcessMsg;
trans->peerMsgFp[TSDB_MSG_TYPE_DM_STATUS_RSP] = dnodeProcessStatusRsp; tsTrans.peerMsgFp[TSDB_MSG_TYPE_DM_STATUS_RSP] = dnodeProcessStatusRsp;
char secret[TSDB_KEY_LEN] = "secret"; char secret[TSDB_KEY_LEN] = "secret";
SRpcInit rpcInit; SRpcInit rpcInit;
...@@ -176,8 +186,8 @@ int32_t dnodeInitClient(SDnTrans *trans) { ...@@ -176,8 +186,8 @@ int32_t dnodeInitClient(SDnTrans *trans) {
rpcInit.ckey = "key"; rpcInit.ckey = "key";
rpcInit.secret = secret; rpcInit.secret = secret;
trans->clientRpc = rpcOpen(&rpcInit); tsTrans.clientRpc = rpcOpen(&rpcInit);
if (trans->clientRpc == NULL) { if (tsTrans.clientRpc == NULL) {
dError("failed to init peer rpc client"); dError("failed to init peer rpc client");
return -1; return -1;
} }
...@@ -186,26 +196,25 @@ int32_t dnodeInitClient(SDnTrans *trans) { ...@@ -186,26 +196,25 @@ int32_t dnodeInitClient(SDnTrans *trans) {
return 0; return 0;
} }
void dnodeCleanupClient(SDnTrans *trans) { void dnodeCleanupClient() {
if (trans->clientRpc) { if (tsTrans.clientRpc) {
rpcClose(trans->clientRpc); rpcClose(tsTrans.clientRpc);
trans->clientRpc = NULL; tsTrans.clientRpc = NULL;
dInfo("dnode peer rpc client is closed"); dInfo("dnode peer rpc client is closed");
} }
} }
static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
SDnode * dnode = dnodeInst();
SRpcMsg rpcMsg = {.handle = pMsg->handle, .pCont = NULL, .contLen = 0}; SRpcMsg rpcMsg = {.handle = pMsg->handle, .pCont = NULL, .contLen = 0};
if (pMsg->pCont == NULL) return; if (pMsg->pCont == NULL) return;
if (dnode->main->runStatus == TD_RUN_STAT_STOPPED) { if (dnodeGetRunStat() == TD_RUN_STAT_STOPPED) {
dError("RPC %p, shell msg:%s is ignored since dnode exiting", pMsg->handle, taosMsg[pMsg->msgType]); dError("RPC %p, shell msg:%s is ignored since dnode exiting", pMsg->handle, taosMsg[pMsg->msgType]);
rpcMsg.code = TSDB_CODE_DND_EXITING; rpcMsg.code = TSDB_CODE_DND_EXITING;
rpcSendResponse(&rpcMsg); rpcSendResponse(&rpcMsg);
rpcFreeCont(pMsg->pCont); rpcFreeCont(pMsg->pCont);
return; return;
} else if (dnode->main->runStatus != TD_RUN_STAT_RUNNING) { } else if (dnodeGetRunStat() != TD_RUN_STAT_RUNNING) {
dError("RPC %p, shell msg:%s is ignored since dnode not running", pMsg->handle, taosMsg[pMsg->msgType]); dError("RPC %p, shell msg:%s is ignored since dnode not running", pMsg->handle, taosMsg[pMsg->msgType]);
rpcMsg.code = TSDB_CODE_APP_NOT_READY; rpcMsg.code = TSDB_CODE_APP_NOT_READY;
rpcSendResponse(&rpcMsg); rpcSendResponse(&rpcMsg);
...@@ -213,14 +222,13 @@ static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { ...@@ -213,14 +222,13 @@ static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *pEpSet) {
return; return;
} }
SDnTrans *trans = dnode->trans;
if (pMsg->msgType == TSDB_MSG_TYPE_QUERY) { if (pMsg->msgType == TSDB_MSG_TYPE_QUERY) {
atomic_fetch_add_32(&trans->queryReqNum, 1); atomic_fetch_add_32(&tsTrans.queryReqNum, 1);
} else if (pMsg->msgType == TSDB_MSG_TYPE_SUBMIT) { } else if (pMsg->msgType == TSDB_MSG_TYPE_SUBMIT) {
atomic_fetch_add_32(&trans->submitReqNum, 1); atomic_fetch_add_32(&tsTrans.submitReqNum, 1);
} else {} } else {}
RpcMsgFp fp = trans->shellMsgFp[pMsg->msgType]; RpcMsgFp fp = tsTrans.shellMsgFp[pMsg->msgType];
if (fp != NULL) { if (fp != NULL) {
(*fp)(pMsg); (*fp)(pMsg);
} else { } else {
...@@ -247,27 +255,23 @@ static int32_t dnodeAuthNetTest(char *user, char *spi, char *encrypt, char *secr ...@@ -247,27 +255,23 @@ static int32_t dnodeAuthNetTest(char *user, char *spi, char *encrypt, char *secr
} }
void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg) { void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg) {
SDnode *dnode = dnodeInst(); rpcSendRequest(tsTrans.clientRpc, epSet, rpcMsg, NULL);
rpcSendRequest(dnode->trans->clientRpc, epSet, rpcMsg, NULL);
} }
void dnodeSendMsgToMnode(SRpcMsg *rpcMsg) { void dnodeSendMsgToMnode(SRpcMsg *rpcMsg) {
SDnode * dnode = dnodeInst();
SRpcEpSet epSet = {0}; SRpcEpSet epSet = {0};
dnodeGetEpSetForPeer(dnode->meps, &epSet); dnodeGetEpSetForPeer(&epSet);
dnodeSendMsgToDnode(&epSet, rpcMsg); dnodeSendMsgToDnode(&epSet, rpcMsg);
} }
void dnodeSendMsgToMnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp) { void dnodeSendMsgToMnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp) {
SDnode * dnode = dnodeInst();
SRpcEpSet epSet = {0}; SRpcEpSet epSet = {0};
dnodeGetEpSetForPeer(dnode->meps, &epSet); dnodeGetEpSetForPeer(&epSet);
rpcSendRecv(dnode->trans->clientRpc, &epSet, rpcMsg, rpcRsp); rpcSendRecv(tsTrans.clientRpc, &epSet, rpcMsg, rpcRsp);
} }
void dnodeSendMsgToDnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp, SRpcEpSet *epSet) { void dnodeSendMsgToDnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp, SRpcEpSet *epSet) {
SDnode *dnode = dnodeInst(); rpcSendRecv(tsTrans.clientRpc, epSet, rpcMsg, rpcRsp);
rpcSendRecv(dnode->trans->clientRpc, epSet, rpcMsg, rpcRsp);
} }
static int32_t dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char *secret, char *ckey) { static int32_t dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char *secret, char *ckey) {
...@@ -303,52 +307,52 @@ static int32_t dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, c ...@@ -303,52 +307,52 @@ static int32_t dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, c
return rpcRsp.code; return rpcRsp.code;
} }
int32_t dnodeInitShell(SDnTrans *trans) { int32_t dnodeInitShell() {
trans->shellMsgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessMsg;
// the following message shall be treated as mnode write // the following message shall be treated as mnode write
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_ACCT] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_ACCT] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_ACCT] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_ACCT] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_DROP_ACCT] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_DROP_ACCT] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_USER] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_USER] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_USER] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_USER] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_DROP_USER] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_DROP_USER] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_DNODE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_DNODE] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_DROP_DNODE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_DROP_DNODE] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_DB] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_DB] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_TP] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_TP] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_FUNCTION] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_FUNCTION] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_DROP_DB] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_DROP_DB] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_SYNC_DB] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_SYNC_DB] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_DROP_TP] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_DROP_TP] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_DROP_FUNCTION] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_DROP_FUNCTION] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_DB] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_DB] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_TP] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_TP] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_TABLE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CREATE_TABLE] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_DROP_TABLE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_DROP_TABLE] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_TABLE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_TABLE] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_STREAM] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_ALTER_STREAM] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_KILL_QUERY] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_KILL_QUERY] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_KILL_STREAM] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_KILL_STREAM] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_KILL_CONN] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_KILL_CONN] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CONFIG_DNODE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CONFIG_DNODE] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_COMPACT_VNODE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_COMPACT_VNODE] = mnodeProcessMsg;
// the following message shall be treated as mnode query // the following message shall be treated as mnode query
trans->shellMsgFp[TSDB_MSG_TYPE_CM_HEARTBEAT] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_HEARTBEAT] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_CONNECT] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_CONNECT] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_USE_DB] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_USE_DB] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_TABLE_META] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_TABLE_META] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_STABLE_VGROUP] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_STABLE_VGROUP] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_TABLES_META] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_TABLES_META] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_SHOW] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_SHOW] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_RETRIEVE] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_RETRIEVE] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_CM_RETRIEVE_FUNC] = mnodeProcessMsg; tsTrans.shellMsgFp[TSDB_MSG_TYPE_CM_RETRIEVE_FUNC] = mnodeProcessMsg;
trans->shellMsgFp[TSDB_MSG_TYPE_NETWORK_TEST] = dnodeProcessStartupReq; tsTrans.shellMsgFp[TSDB_MSG_TYPE_NETWORK_TEST] = dnodeProcessStartupReq;
int32_t numOfThreads = (int32_t)((tsNumOfCores * tsNumOfThreadsPerCore) / 2.0); int32_t numOfThreads = (int32_t)((tsNumOfCores * tsNumOfThreadsPerCore) / 2.0);
if (numOfThreads < 1) { if (numOfThreads < 1) {
...@@ -366,8 +370,8 @@ int32_t dnodeInitShell(SDnTrans *trans) { ...@@ -366,8 +370,8 @@ int32_t dnodeInitShell(SDnTrans *trans) {
rpcInit.idleTime = tsShellActivityTimer * 1000; rpcInit.idleTime = tsShellActivityTimer * 1000;
rpcInit.afp = dnodeRetrieveUserAuthInfo; rpcInit.afp = dnodeRetrieveUserAuthInfo;
trans->shellRpc = rpcOpen(&rpcInit); tsTrans.shellRpc = rpcOpen(&rpcInit);
if (trans->shellRpc == NULL) { if (tsTrans.shellRpc == NULL) {
dError("failed to init shell rpc server"); dError("failed to init shell rpc server");
return -1; return -1;
} }
...@@ -376,41 +380,31 @@ int32_t dnodeInitShell(SDnTrans *trans) { ...@@ -376,41 +380,31 @@ int32_t dnodeInitShell(SDnTrans *trans) {
return 0; return 0;
} }
void dnodeCleanupShell(SDnTrans *trans) { void dnodeCleanupShell() {
if (trans->shellRpc) { if (tsTrans.shellRpc) {
rpcClose(trans->shellRpc); rpcClose(tsTrans.shellRpc);
trans->shellRpc = NULL; tsTrans.shellRpc = NULL;
} }
} }
int32_t dnodeInitTrans(SDnTrans **out) { int32_t dnodeInitTrans() {
SDnTrans *trans = calloc(1, sizeof(SDnTrans)); if (dnodeInitClient() != 0) {
if (trans == NULL) return -1;
*out = trans;
if (dnodeInitClient(trans) != 0) {
return -1; return -1;
} }
if (dnodeInitServer(trans) != 0) { if (dnodeInitServer() != 0) {
return -1; return -1;
} }
if (dnodeInitShell(trans) != 0) { if (dnodeInitShell() != 0) {
return -1; return -1;
} }
return 0; return 0;
} }
void dnodeCleanupTrans(SDnTrans **out) { void dnodeCleanupTrans() {
SDnTrans* trans = *out; dnodeCleanupShell();
*out = NULL; dnodeCleanupServer();
dnodeCleanupClient();
dnodeCleanupShell(trans);
dnodeCleanupServer(trans);
dnodeCleanupClient(trans);
free(trans);
} }
...@@ -15,9 +15,11 @@ target_link_libraries( ...@@ -15,9 +15,11 @@ target_link_libraries(
PUBLIC meta PUBLIC meta
PUBLIC tq PUBLIC tq
PUBLIC tsdb PUBLIC tsdb
PUBLIC wal
PUBLIC cjson
) )
# test # test
if(${BUILD_TEST}) if(${BUILD_TEST})
add_subdirectory(test) add_subdirectory(test)
endif(${BUILD_TEST}) endif(${BUILD_TEST})
\ No newline at end of file
...@@ -13,19 +13,19 @@ ...@@ -13,19 +13,19 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_CFG_H #ifndef _TD_VNODE_CFG_H_
#define TDENGINE_VNODE_CFG_H #define _TD_VNODE_CFG_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeReadCfg(SVnodeObj *pVnode); int32_t vnodeReadCfg(SVnode *pVnode);
int32_t vnodeWriteCfg(SCreateVnodeMsg *pVnodeCfg); int32_t vnodeWriteCfg(SCreateVnodeMsg *pVnodeCfg);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_CFG_H_*/
...@@ -16,7 +16,6 @@ ...@@ -16,7 +16,6 @@
#ifndef _TD_VNODE_INT_H_ #ifndef _TD_VNODE_INT_H_
#define _TD_VNODE_INT_H_ #define _TD_VNODE_INT_H_
#include "os.h" #include "os.h"
#include "amalloc.h" #include "amalloc.h"
#include "meta.h" #include "meta.h"
...@@ -25,20 +24,83 @@ ...@@ -25,20 +24,83 @@
#include "trpc.h" #include "trpc.h"
#include "tsdb.h" #include "tsdb.h"
#include "vnode.h" #include "vnode.h"
#include "tlog.h"
#include "tqueue.h"
#include "wal.h"
#include "tworker.h"
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
typedef struct SVnode { extern int32_t vDebugFlag;
#define vFatal(...) { if (vDebugFlag & DEBUG_FATAL) { taosPrintLog("VND FATAL ", 255, __VA_ARGS__); }}
#define vError(...) { if (vDebugFlag & DEBUG_ERROR) { taosPrintLog("VND ERROR ", 255, __VA_ARGS__); }}
#define vWarn(...) { if (vDebugFlag & DEBUG_WARN) { taosPrintLog("VND WARN ", 255, __VA_ARGS__); }}
#define vInfo(...) { if (vDebugFlag & DEBUG_INFO) { taosPrintLog("VND ", 255, __VA_ARGS__); }}
#define vDebug(...) { if (vDebugFlag & DEBUG_DEBUG) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
#define vTrace(...) { if (vDebugFlag & DEBUG_TRACE) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
typedef struct {
SMeta * pMeta; SMeta * pMeta;
STsdb * pTsdb; STsdb * pTsdb;
STQ * pTQ; STQ * pTQ;
SMemAllocator *allocator; SMemAllocator *allocator;
int32_t vgId; // global vnode group ID
int32_t refCount; // reference count
int64_t queuedWMsgSize;
int32_t queuedWMsg;
int32_t queuedRMsg;
int32_t numOfExistQHandle; // current initialized and existed query handle in current dnode
int32_t flowctrlLevel;
int8_t preClose; // drop and close switch
int8_t reserved[3];
int64_t sequence; // for topic
int8_t status;
int8_t role;
int8_t accessState;
int8_t isFull;
int8_t isCommiting;
int8_t dbReplica;
int8_t dropped;
int8_t dbType;
uint64_t version; // current version
uint64_t cversion; // version while commit start
uint64_t fversion; // version on saved data file
void * wqueue; // write queue
void * qqueue; // read query queue
void * fqueue; // read fetch/cancel queue
void * wal;
void * tsdb;
int64_t sync;
void * events;
void * cq; // continuous query
int32_t dbCfgVersion;
int32_t vgCfgVersion;
STsdbCfg tsdbCfg;
#if 0
SSyncCfg syncCfg;
#endif
SWalCfg walCfg;
void * qMgmt;
char * rootDir;
tsem_t sem;
char db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN];
pthread_mutex_t statusMutex;
} SVnode; } SVnode;
typedef struct {
int32_t len;
void * rsp;
void * qhandle; // used by query and retrieve msg
} SVnRsp;
void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif /*_TD_VNODE_INT_H_*/ #endif /*_TD_VNODE_INT_H_*/
\ No newline at end of file
...@@ -13,25 +13,35 @@ ...@@ -13,25 +13,35 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_MAIN_H #ifndef _TD_VNODE_MAIN_H_
#define TDENGINE_VNODE_MAIN_H #define _TD_VNODE_MAIN_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeInitMain();
void vnodeCleanupMain();
int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg); int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg);
int32_t vnodeDrop(int32_t vgId); int32_t vnodeDrop(int32_t vgId);
int32_t vnodeOpen(int32_t vgId); int32_t vnodeOpen(int32_t vgId);
int32_t vnodeAlter(void *pVnode, SCreateVnodeMsg *pVnodeCfg); int32_t vnodeAlter(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg);
int32_t vnodeSync(int32_t vgId); int32_t vnodeSync(int32_t vgId);
int32_t vnodeClose(int32_t vgId); int32_t vnodeClose(int32_t vgId);
void vnodeCleanUp(SVnodeObj *pVnode); void vnodeCleanUp(SVnode *pVnode);
void vnodeDestroy(SVnodeObj *pVnode); void vnodeDestroy(SVnode *pVnode);
int32_t vnodeCompact(int32_t vgId);
void vnodeBackup(int32_t vgId);
void vnodeGetStatus(struct SStatusMsg *status);
SVnode *vnodeAcquire(int32_t vgId);
SVnode *vnodeAcquireNotClose(int32_t vgId);
void vnodeRelease(SVnode *pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_MAIN_H_*/
...@@ -13,20 +13,20 @@ ...@@ -13,20 +13,20 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_BACKUP_H #ifndef _TD_VNODE_MGMT_H_
#define TDENGINE_VNODE_BACKUP_H #define _TD_VNODE_MGMT_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeInitBackup(); int32_t vnodeInitMgmt();
void vnodeCleanupBackup(); void vnodeCleanupMgmt();
int32_t vnodeBackup(int32_t vgId); void vnodeProcessMgmtMsg(SRpcMsg *pMsg);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_MGMT_H_*/
...@@ -13,24 +13,23 @@ ...@@ -13,24 +13,23 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_READ_H #ifndef _TD_VNODE_MGMT_MSG_H_
#define TDENGINE_VNODE_READ_H #define _TD_VNODE_MGMT_MSG_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeInitRead(void); int32_t vnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg);
void vnodeCleanupRead(void); int32_t vnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg);
int32_t vnodeProcessSyncVnodeMsg(SRpcMsg *rpcMsg);
int32_t vnodeWriteToRQueue(void *pVnode, void *pCont, int32_t contLen, int8_t qtype, void *rparam); int32_t vnodeProcessCompactVnodeMsg(SRpcMsg *rpcMsg);
void vnodeFreeFromRQueue(void *pVnode, SVReadMsg *pRead); int32_t vnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg);
int32_t vnodeProcessRead(void *pVnode, SVReadMsg *pRead); int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg);
void vnodeWaitReadCompleted(SVnodeObj *pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_MGMT_H_*/
...@@ -13,30 +13,30 @@ ...@@ -13,30 +13,30 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_MGMT_H #ifndef _TD_VNODE_READ_H_
#define TDENGINE_VNODE_MGMT_H #define _TD_VNODE_READ_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeInitMgmt(); int32_t vnodeInitRead();
void vnodeCleanupMgmt(); void vnodeCleanupRead();
taos_queue vnodeAllocQueryQueue(SVnode *pVnode);
taos_queue vnodeAllocFetchQueue(SVnode *pVnode);
void vnodeFreeQueryQueue(taos_queue pQueue);
void vnodeFreeFetchQueue(taos_queue pQueue);
void* vnodeAcquire(int32_t vgId); void vnodeProcessReadMsg(SRpcMsg *pRpcMsg);
void vnodeRelease(void *pVnode); int32_t vnodeReputPutToRQueue(SVnode *pVnode, void **qhandle, void *ahandle);
void* vnodeGetWal(void *pVnode);
int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes); void vnodeStartRead(SVnode *pVnode);
void vnodeBuildStatusMsg(void *pStatus); void vnodeStopRead(SVnode *pVnode);
void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes); void vnodeWaitReadCompleted(SVnode *pVnode);
void vnodeAddIntoHash(SVnodeObj* pVnode);
void vnodeRemoveFromHash(SVnodeObj * pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_READ_H_*/
...@@ -13,24 +13,32 @@ ...@@ -13,24 +13,32 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_WRITE_H #ifndef _TD_VNODE_READ_MSG_H_
#define TDENGINE_VNODE_WRITE_H #define _TD_VNODE_READ_MSG_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeInitWrite(void); typedef struct SReadMsg {
void vnodeCleanupWrite(void); int32_t code;
int32_t contLen;
int8_t qtype;
int8_t msgType;
SVnode *pVnode;
SVnRsp rspRet;
void * rpcHandle;
void * rpcAhandle;
void * qhandle;
char pCont[];
} SReadMsg;
int32_t vnodeWriteToWQueue(void *pVnode, void *pHead, int32_t qtype, void *pRpcMsg); int32_t vnodeProcessQueryMsg(SVnode *pVnode, SReadMsg *pRead);
void vnodeFreeFromWQueue(void *pVnode, SVWriteMsg *pWrite); int32_t vnodeProcessFetchMsg(SVnode *pVnode, SReadMsg *pRead);
int32_t vnodeProcessWrite(void *pVnode, void *pHead, int32_t qtype, void *pRspRet);
void vnodeWaitWriteCompleted(SVnodeObj *pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_READ_MSG_H_*/
\ No newline at end of file
...@@ -13,8 +13,8 @@ ...@@ -13,8 +13,8 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_STATUS_H #ifndef _TD_VNODE_STATUS_H_
#define TDENGINE_VNODE_STATUS_H #define _TD_VNODE_STATUS_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
...@@ -25,24 +25,23 @@ typedef enum _VN_STATUS { ...@@ -25,24 +25,23 @@ typedef enum _VN_STATUS {
TAOS_VN_STATUS_INIT = 0, TAOS_VN_STATUS_INIT = 0,
TAOS_VN_STATUS_READY = 1, TAOS_VN_STATUS_READY = 1,
TAOS_VN_STATUS_CLOSING = 2, TAOS_VN_STATUS_CLOSING = 2,
TAOS_VN_STATUS_UPDATING = 3, TAOS_VN_STATUS_UPDATING = 3
TAOS_VN_STATUS_RESET = 4,
} EVnodeStatus; } EVnodeStatus;
bool vnodeSetInitStatus(SVnodeObj* pVnode); // vnodeStatus
bool vnodeSetReadyStatus(SVnodeObj* pVnode); extern char* vnodeStatus[];
bool vnodeSetClosingStatus(SVnodeObj* pVnode);
bool vnodeSetUpdatingStatus(SVnodeObj* pVnode);
bool vnodeSetResetStatus(SVnodeObj* pVnode);
bool vnodeInInitStatus(SVnodeObj* pVnode); bool vnodeSetInitStatus(SVnode* pVnode);
bool vnodeInReadyStatus(SVnodeObj* pVnode); bool vnodeSetReadyStatus(SVnode* pVnode);
bool vnodeInReadyOrUpdatingStatus(SVnodeObj* pVnode); bool vnodeSetClosingStatus(SVnode* pVnode);
bool vnodeInClosingStatus(SVnodeObj* pVnode); bool vnodeSetUpdatingStatus(SVnode* pVnode);
bool vnodeInResetStatus(SVnodeObj* pVnode);
bool vnodeInInitStatus(SVnode* pVnode);
bool vnodeInReadyStatus(SVnode* pVnode);
bool vnodeInClosingStatus(SVnode* pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_STATUS_H_*/
\ No newline at end of file \ No newline at end of file
...@@ -13,19 +13,19 @@ ...@@ -13,19 +13,19 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_VERSION_H #ifndef _TD_VNODE_VERSION_H_
#define TDENGINE_VNODE_VERSION_H #define _TD_VNODE_VERSION_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeReadVersion(SVnodeObj *pVnode); int32_t vnodeReadVersion(SVnode *pVnode);
int32_t vnodeSaveVersion(SVnodeObj *pVnode); int32_t vnodeSaveVersion(SVnode *pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_VERSION_H_*/
...@@ -13,21 +13,22 @@ ...@@ -13,21 +13,22 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef TDENGINE_VNODE_WORKER_H #ifndef _TD_VNODE_WORKER_H_
#define TDENGINE_VNODE_WORKER_H #define _TD_VNODE_WORKER_H_
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif
#include "vnodeInt.h" #include "vnodeInt.h"
int32_t vnodeInitMWorker(); int32_t vnodeInitWorker();
void vnodeCleanupMWorker(); void vnodeCleanupWorker();
int32_t vnodeCleanupInMWorker(SVnodeObj *pVnode); void vnodeProcessCleanupTask(SVnode *pVnode);
int32_t vnodeDestroyInMWorker(SVnodeObj *pVnode); void vnodeProcessDestroyTask(SVnode *pVnode);
void vnodeProcessBackupTask(SVnode *pVnode);
#ifdef __cplusplus #ifdef __cplusplus
} }
#endif #endif
#endif #endif /*_TD_VNODE_WORKER_H_*/
\ No newline at end of file \ No newline at end of file
...@@ -16,6 +16,25 @@ ...@@ -16,6 +16,25 @@
#ifndef _TD_VNODE_WRITE_H_ #ifndef _TD_VNODE_WRITE_H_
#define _TD_VNODE_WRITE_H_ #define _TD_VNODE_WRITE_H_
int vnodeProcessSubmitReq(SVnode *pVnode, SSubmitReq *pReq, SSubmitRsp *pRsp); #ifdef __cplusplus
extern "C" {
#endif
#include "vnodeInt.h"
int32_t vnodeInitWrite();
void vnodeCleanupWrite();
taos_queue vnodeAllocWriteQueue(SVnode *pVnode);
void vnodeFreeWriteQueue(taos_queue pQueue);
void vnodeProcessWriteMsg(SRpcMsg *pRpcMsg);
int32_t vnodeProcessWalMsg(SVnode *pVnode, SWalHead *pHead);
void vnodeStartWrite(SVnode *pVnode);
void vnodeStopWrite(SVnode *pVnode);
void vnodeWaitWriteCompleted(SVnode *pVnode);
#ifdef __cplusplus
}
#endif
#endif /*_TD_VNODE_WRITE_H_*/ #endif /*_TD_VNODE_WRITE_H_*/
\ No newline at end of file
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef _TD_VNODE_WRITE_MSG_H_
#define _TD_VNODE_WRITE_MSG_H_
#ifdef __cplusplus
extern "C" {
#endif
#include "vnodeInt.h"
int32_t vnodeProcessSubmitReq(SVnode *pVnode, SSubmitReq *pReq, SSubmitRsp *pRsp);
int32_t vnodeProcessCreateTableReq(SVnode *pVnode, SCreateTableReq *pReq, SCreateTableRsp *pRsp);
int32_t vnodeProcessDropTableReq(SVnode *pVnode, SDropTableReq *pReq, SDropTableRsp *pRsp);
int32_t vnodeProcessAlterTableReq(SVnode *pVnode, SAlterTableReq *pReq, SAlterTableRsp *pRsp);
int32_t vnodeProcessDropStableReq(SVnode *pVnode, SDropStableReq *pReq, SDropStableRsp *pRsp);
int32_t vnodeProcessUpdateTagValReq(SVnode *pVnode, SUpdateTagValReq *pReq, SUpdateTagValRsp *pRsp);
#ifdef __cplusplus
}
#endif
#endif /*_TD_VNODE_WRITE_MSG_H_*/
\ No newline at end of file
...@@ -17,10 +17,10 @@ ...@@ -17,10 +17,10 @@
#include "os.h" #include "os.h"
#include "cJSON.h" #include "cJSON.h"
#include "tglobal.h" #include "tglobal.h"
#include "dnode.h"
#include "vnodeCfg.h" #include "vnodeCfg.h"
static void vnodeLoadCfg(SVnodeObj *pVnode, SCreateVnodeMsg* vnodeMsg) { static void vnodeLoadCfg(SVnode *pVnode, SCreateVnodeMsg *vnodeMsg) {
#if 0
tstrncpy(pVnode->db, vnodeMsg->db, sizeof(pVnode->db)); tstrncpy(pVnode->db, vnodeMsg->db, sizeof(pVnode->db));
pVnode->dbCfgVersion = vnodeMsg->cfg.dbCfgVersion; pVnode->dbCfgVersion = vnodeMsg->cfg.dbCfgVersion;
pVnode->vgCfgVersion = vnodeMsg->cfg.vgCfgVersion; pVnode->vgCfgVersion = vnodeMsg->cfg.vgCfgVersion;
...@@ -56,9 +56,11 @@ static void vnodeLoadCfg(SVnodeObj *pVnode, SCreateVnodeMsg* vnodeMsg) { ...@@ -56,9 +56,11 @@ static void vnodeLoadCfg(SVnodeObj *pVnode, SCreateVnodeMsg* vnodeMsg) {
SNodeInfo *node = &pVnode->syncCfg.nodeInfo[i]; SNodeInfo *node = &pVnode->syncCfg.nodeInfo[i];
vInfo("vgId:%d, dnode:%d, %s:%u", pVnode->vgId, node->nodeId, node->nodeFqdn, node->nodePort); vInfo("vgId:%d, dnode:%d, %s:%u", pVnode->vgId, node->nodeId, node->nodeFqdn, node->nodePort);
} }
#endif
} }
int32_t vnodeReadCfg(SVnodeObj *pVnode) { int32_t vnodeReadCfg(SVnode *pVnode) {
#if 0
int32_t ret = TSDB_CODE_VND_APP_ERROR; int32_t ret = TSDB_CODE_VND_APP_ERROR;
int32_t len = 0; int32_t len = 0;
int maxLen = 1000; int maxLen = 1000;
...@@ -66,6 +68,7 @@ int32_t vnodeReadCfg(SVnodeObj *pVnode) { ...@@ -66,6 +68,7 @@ int32_t vnodeReadCfg(SVnodeObj *pVnode) {
cJSON * root = NULL; cJSON * root = NULL;
FILE * fp = NULL; FILE * fp = NULL;
bool nodeChanged = false; bool nodeChanged = false;
SCreateVnodeMsg vnodeMsg; SCreateVnodeMsg vnodeMsg;
char file[TSDB_FILENAME_LEN + 30] = {0}; char file[TSDB_FILENAME_LEN + 30] = {0};
...@@ -286,8 +289,13 @@ int32_t vnodeReadCfg(SVnodeObj *pVnode) { ...@@ -286,8 +289,13 @@ int32_t vnodeReadCfg(SVnodeObj *pVnode) {
} }
tstrncpy(node->nodeEp, nodeEp->valuestring, TSDB_EP_LEN); tstrncpy(node->nodeEp, nodeEp->valuestring, TSDB_EP_LEN);
bool changed = dnodeCheckEpChanged(node->nodeId, node->nodeEp); char nodeEpStr[TSDB_EP_LEN];
if (changed) nodeChanged = changed; vnodeGetDnodeEp(node->nodeId, nodeEpStr, NULL, NULL);
bool changed = (strcmp(node->nodeEp, nodeEpStr) != 0);
if (changed) {
tstrncpy(node->nodeEp, nodeEpStr, TSDB_EP_LEN);
nodeChanged = changed;
}
} }
ret = TSDB_CODE_SUCCESS; ret = TSDB_CODE_SUCCESS;
...@@ -350,7 +358,7 @@ int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) { ...@@ -350,7 +358,7 @@ int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) {
len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n"); len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n");
for (int32_t i = 0; i < pMsg->cfg.vgReplica; i++) { for (int32_t i = 0; i < pMsg->cfg.vgReplica; i++) {
SVnodeDesc *node = &pMsg->nodes[i]; SVnodeDesc *node = &pMsg->nodes[i];
dnodeUpdateEp(node->nodeId, node->nodeEp, NULL, NULL); vnodeGetDnodeEp(node->nodeId, node->nodeEp, NULL, NULL);
len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", node->nodeId); len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", node->nodeId);
len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", node->nodeEp); len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", node->nodeEp);
if (i < pMsg->cfg.vgReplica - 1) { if (i < pMsg->cfg.vgReplica - 1) {
...@@ -368,5 +376,6 @@ int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) { ...@@ -368,5 +376,6 @@ int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) {
terrno = 0; terrno = 0;
vInfo("vgId:%d, successed to write %s", pMsg->cfg.vgId, file); vInfo("vgId:%d, successed to write %s", pMsg->cfg.vgId, file);
#endif
return TSDB_CODE_SUCCESS; return TSDB_CODE_SUCCESS;
} }
...@@ -13,16 +13,39 @@ ...@@ -13,16 +13,39 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#include "vnodeInt.h" #define _DEFAULT_SOURCE
#include "os.h"
#include "tstep.h"
#include "vnodeMain.h"
#include "vnodeMgmt.h"
#include "vnodeRead.h"
#include "vnodeWorker.h"
#include "vnodeWrite.h"
int32_t vnodeInit(SVnodePara para) { return 0; } static struct {
struct SSteps *steps;
SVnodeFp fp;
} tsVint;
void vnodeCleanup() {} int32_t vnodeInit(SVnodePara para) {
tsVint.fp = para.fp;
int32_t vnodeGetStatistics(SVnodeStat *stat) { return 0; } struct SSteps *steps = taosStepInit(8, NULL);
if (steps == NULL) return -1;
void vnodeGetStatus(struct SStatusMsg *status) {} taosStepAdd(steps, "vnode-main", vnodeInitMain, vnodeCleanupMain);
taosStepAdd(steps, "vnode-worker",vnodeInitWorker, vnodeCleanupWorker);
taosStepAdd(steps, "vnode-read", vnodeInitRead, vnodeCleanupRead);
taosStepAdd(steps, "vnode-mgmt", vnodeInitMgmt, vnodeCleanupMgmt);
taosStepAdd(steps, "vnode-write", vnodeInitWrite, vnodeCleanupWrite);
// taosStepAdd(steps, "vnode-queue", tsdbInitCommitQueue, tsdbDestroyCommitQueue);
void vnodeSetAccess(struct SVgroupAccess *access, int32_t numOfVnodes) {} tsVint.steps = steps;
return taosStepExec(tsVint.steps);
}
void vnodeProcessMsg(SRpcMsg *msg) {} void vnodeCleanup() { taosStepCleanup(tsVint.steps); }
void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port) {
return (*tsVint.fp.GetDnodeEp)(dnodeId, ep, fqdn, port);
}
\ No newline at end of file
...@@ -18,30 +18,104 @@ ...@@ -18,30 +18,104 @@
#include "taoserror.h" #include "taoserror.h"
#include "taosmsg.h" #include "taosmsg.h"
#include "tglobal.h" #include "tglobal.h"
#include "tfs.h" #include "ttimer.h"
#include "query.h" #include "thash.h"
#include "dnode.h" // #include "query.h"
#include "vnodeCfg.h" #include "vnodeCfg.h"
#include "vnodeMain.h"
#include "vnodeMgmt.h"
#include "vnodeRead.h"
#include "vnodeStatus.h" #include "vnodeStatus.h"
#include "vnodeSync.h"
#include "vnodeVersion.h" #include "vnodeVersion.h"
#include "vnodeMgmt.h"
#include "vnodeWorker.h" #include "vnodeWorker.h"
#include "vnodeBackup.h" #include "vnodeWrite.h"
#include "vnodeMain.h"
typedef struct {
pthread_t thread;
int32_t threadIndex;
int32_t failed;
int32_t opened;
int32_t vnodeNum;
int32_t * vnodeList;
} SOpenVnodeThread;
static struct {
void * timer;
SHashObj *hash;
int32_t openVnodes;
int32_t totalVnodes;
void (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
} tsVmain;
static void vnodeIncRef(void *ptNode) {
assert(ptNode != NULL);
SVnode **ppVnode = (SVnode **)ptNode;
assert(ppVnode);
assert(*ppVnode);
SVnode *pVnode = *ppVnode;
atomic_add_fetch_32(&pVnode->refCount, 1);
vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
}
SVnode *vnodeAcquire(int32_t vgId) {
SVnode *pVnode = NULL;
#if 0
taosHashGetClone(tsVmain.hash, &vgId, sizeof(int32_t), vnodeIncRef, &pVnode);
#endif
if (pVnode == NULL) {
terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
vDebug("vgId:%d, not exist", vgId);
return NULL;
}
return pVnode;
}
SVnode *vnodeAcquireNotClose(int32_t vgId) {
SVnode *pVnode = vnodeAcquire(vgId);
if (pVnode != NULL && pVnode->preClose == 1) {
vnodeRelease(pVnode);
terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
vDebug("vgId:%d, not exist, pre closing", vgId);
return NULL;
}
return pVnode;
}
void vnodeRelease(SVnode *pVnode) {
if (pVnode == NULL) return;
int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1);
int32_t vgId = pVnode->vgId;
vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", vgId, refCount, pVnode);
assert(refCount >= 0);
if (refCount <= 0) {
vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", vgId, refCount, pVnode);
vnodeProcessDestroyTask(pVnode);
int32_t count = taosHashGetSize(tsVmain.hash);
vDebug("vgId:%d, vnode is destroyed, vnodes:%d", vgId, count);
}
}
static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno); static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno);
int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) { int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) {
int32_t code; int32_t code;
SVnodeObj *pVnode = vnodeAcquire(pVnodeCfg->cfg.vgId); SVnode *pVnode = vnodeAcquire(pVnodeCfg->cfg.vgId);
if (pVnode != NULL) { if (pVnode != NULL) {
vDebug("vgId:%d, vnode already exist, refCount:%d pVnode:%p", pVnodeCfg->cfg.vgId, pVnode->refCount, pVnode); vDebug("vgId:%d, vnode already exist, refCount:%d pVnode:%p", pVnodeCfg->cfg.vgId, pVnode->refCount, pVnode);
vnodeRelease(pVnode); vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS; return TSDB_CODE_SUCCESS;
} }
#if 0
if (tfsMkdir("vnode") < 0) { if (tfsMkdir("vnode") < 0) {
vError("vgId:%d, failed to create vnode dir, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno)); vError("vgId:%d, failed to create vnode dir, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno));
return terrno; return terrno;
...@@ -64,7 +138,7 @@ int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) { ...@@ -64,7 +138,7 @@ int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) {
vError("vgId:%d, failed to create tsdb in vnode, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno)); vError("vgId:%d, failed to create tsdb in vnode, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno));
return TSDB_CODE_VND_INIT_FAILED; return TSDB_CODE_VND_INIT_FAILED;
} }
#endif
vInfo("vgId:%d, vnode dir is created, walLevel:%d fsyncPeriod:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.walLevel, vInfo("vgId:%d, vnode dir is created, walLevel:%d fsyncPeriod:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.walLevel,
pVnodeCfg->cfg.fsyncPeriod); pVnodeCfg->cfg.fsyncPeriod);
code = vnodeOpen(pVnodeCfg->cfg.vgId); code = vnodeOpen(pVnodeCfg->cfg.vgId);
...@@ -73,7 +147,8 @@ int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) { ...@@ -73,7 +147,8 @@ int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) {
} }
int32_t vnodeSync(int32_t vgId) { int32_t vnodeSync(int32_t vgId) {
SVnodeObj *pVnode = vnodeAcquireNotClose(vgId); #if 0
SVnode *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) { if (pVnode == NULL) {
vDebug("vgId:%d, failed to sync, vnode not find", vgId); vDebug("vgId:%d, failed to sync, vnode not find", vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID; return TSDB_CODE_VND_INVALID_VGROUP_ID;
...@@ -90,13 +165,12 @@ int32_t vnodeSync(int32_t vgId) { ...@@ -90,13 +165,12 @@ int32_t vnodeSync(int32_t vgId) {
} }
vnodeRelease(pVnode); vnodeRelease(pVnode);
#endif
return TSDB_CODE_SUCCESS; return TSDB_CODE_SUCCESS;
} }
int32_t vnodeDrop(int32_t vgId) { int32_t vnodeDrop(int32_t vgId) {
SVnodeObj *pVnode = vnodeAcquireNotClose(vgId); SVnode *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) { if (pVnode == NULL) {
vDebug("vgId:%d, failed to drop, vnode not find", vgId); vDebug("vgId:%d, failed to drop, vnode not find", vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID; return TSDB_CODE_VND_INVALID_VGROUP_ID;
...@@ -110,25 +184,29 @@ int32_t vnodeDrop(int32_t vgId) { ...@@ -110,25 +184,29 @@ int32_t vnodeDrop(int32_t vgId) {
pVnode->dropped = 1; pVnode->dropped = 1;
vnodeRelease(pVnode); vnodeRelease(pVnode);
vnodeCleanupInMWorker(pVnode); vnodeProcessCleanupTask(pVnode);
return TSDB_CODE_SUCCESS; return TSDB_CODE_SUCCESS;
} }
int32_t vnodeCompact(int32_t vgId) { int32_t vnodeCompact(int32_t vgId) {
void *pVnode = vnodeAcquire(vgId); #if 0
SVnode *pVnode = vnodeAcquire(vgId);
if (pVnode != NULL) { if (pVnode != NULL) {
vDebug("vgId:%d, compact vnode msg is received", vgId); vDebug("vgId:%d, compact vnode msg is received", vgId);
//not care success or not // not care success or not
tsdbCompact(((SVnodeObj*)pVnode)->tsdb); tsdbCompact(((SVnode *)pVnode)->tsdb);
vnodeRelease(pVnode); vnodeRelease(pVnode);
} else { } else {
vInfo("vgId:%d, vnode not exist, can't compact it", vgId); vInfo("vgId:%d, vnode not exist, can't compact it", vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID; return TSDB_CODE_VND_INVALID_VGROUP_ID;
} }
return TSDB_CODE_SUCCESS; #endif
return TSDB_CODE_SUCCESS;
} }
static int32_t vnodeAlterImp(SVnodeObj *pVnode, SCreateVnodeMsg *pVnodeCfg) { static int32_t vnodeAlterImp(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg) {
#if 0
STsdbCfg tsdbCfg = pVnode->tsdbCfg; STsdbCfg tsdbCfg = pVnode->tsdbCfg;
SSyncCfg syncCfg = pVnode->syncCfg; SSyncCfg syncCfg = pVnode->syncCfg;
int32_t dbCfgVersion = pVnode->dbCfgVersion; int32_t dbCfgVersion = pVnode->dbCfgVersion;
...@@ -204,13 +282,11 @@ static int32_t vnodeAlterImp(SVnodeObj *pVnode, SCreateVnodeMsg *pVnodeCfg) { ...@@ -204,13 +282,11 @@ static int32_t vnodeAlterImp(SVnodeObj *pVnode, SCreateVnodeMsg *pVnodeCfg) {
vnodeSetReadyStatus(pVnode); vnodeSetReadyStatus(pVnode);
} }
#endif
return 0; return 0;
} }
int32_t vnodeAlter(void *vparam, SCreateVnodeMsg *pVnodeCfg) { int32_t vnodeAlter(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg) {
SVnodeObj *pVnode = vparam;
vDebug("vgId:%d, current dbCfgVersion:%d vgCfgVersion:%d, input dbCfgVersion:%d vgCfgVersion:%d", pVnode->vgId, vDebug("vgId:%d, current dbCfgVersion:%d vgCfgVersion:%d, input dbCfgVersion:%d vgCfgVersion:%d", pVnode->vgId,
pVnode->dbCfgVersion, pVnode->vgCfgVersion, pVnodeCfg->cfg.dbCfgVersion, pVnodeCfg->cfg.vgCfgVersion); pVnode->dbCfgVersion, pVnode->vgCfgVersion, pVnodeCfg->cfg.dbCfgVersion, pVnodeCfg->cfg.vgCfgVersion);
...@@ -231,6 +307,7 @@ int32_t vnodeAlter(void *vparam, SCreateVnodeMsg *pVnodeCfg) { ...@@ -231,6 +307,7 @@ int32_t vnodeAlter(void *vparam, SCreateVnodeMsg *pVnodeCfg) {
} }
static void vnodeFindWalRootDir(int32_t vgId, char *walRootDir) { static void vnodeFindWalRootDir(int32_t vgId, char *walRootDir) {
#if 0
char vnodeDir[TSDB_FILENAME_LEN] = "\0"; char vnodeDir[TSDB_FILENAME_LEN] = "\0";
snprintf(vnodeDir, TSDB_FILENAME_LEN, "/vnode/vnode%d/wal", vgId); snprintf(vnodeDir, TSDB_FILENAME_LEN, "/vnode/vnode%d/wal", vgId);
...@@ -246,15 +323,17 @@ static void vnodeFindWalRootDir(int32_t vgId, char *walRootDir) { ...@@ -246,15 +323,17 @@ static void vnodeFindWalRootDir(int32_t vgId, char *walRootDir) {
sprintf(walRootDir, "%s/vnode/vnode%d", TFS_DISK_PATH(tfile->level, tfile->id), vgId); sprintf(walRootDir, "%s/vnode/vnode%d", TFS_DISK_PATH(tfile->level, tfile->id), vgId);
tfsClosedir(tdir); tfsClosedir(tdir);
#endif
} }
int32_t vnodeOpen(int32_t vgId) { int32_t vnodeOpen(int32_t vgId) {
#if 0
char temp[TSDB_FILENAME_LEN * 3]; char temp[TSDB_FILENAME_LEN * 3];
char rootDir[TSDB_FILENAME_LEN * 2]; char rootDir[TSDB_FILENAME_LEN * 2];
char walRootDir[TSDB_FILENAME_LEN * 2] = {0}; char walRootDir[TSDB_FILENAME_LEN * 2] = {0};
snprintf(rootDir, TSDB_FILENAME_LEN * 2, "%s/vnode%d", tsVnodeDir, vgId); snprintf(rootDir, TSDB_FILENAME_LEN * 2, "%s/vnode%d", tsVnodeDir, vgId);
SVnodeObj *pVnode = calloc(sizeof(SVnodeObj), 1); SVnode *pVnode = calloc(sizeof(SVnode), 1);
if (pVnode == NULL) { if (pVnode == NULL) {
vError("vgId:%d, failed to open vnode since no enough memory", vgId); vError("vgId:%d, failed to open vnode since no enough memory", vgId);
return TAOS_SYSTEM_ERROR(errno); return TAOS_SYSTEM_ERROR(errno);
...@@ -262,9 +341,9 @@ int32_t vnodeOpen(int32_t vgId) { ...@@ -262,9 +341,9 @@ int32_t vnodeOpen(int32_t vgId) {
atomic_add_fetch_32(&pVnode->refCount, 1); atomic_add_fetch_32(&pVnode->refCount, 1);
pVnode->vgId = vgId; pVnode->vgId = vgId;
pVnode->fversion = 0; pVnode->fversion = 0;
pVnode->version = 0; pVnode->version = 0;
pVnode->tsdbCfg.tsdbId = pVnode->vgId; pVnode->tsdbCfg.tsdbId = pVnode->vgId;
pVnode->rootDir = strdup(rootDir); pVnode->rootDir = strdup(rootDir);
pVnode->accessState = TSDB_VN_ALL_ACCCESS; pVnode->accessState = TSDB_VN_ALL_ACCCESS;
...@@ -279,7 +358,7 @@ int32_t vnodeOpen(int32_t vgId) { ...@@ -279,7 +358,7 @@ int32_t vnodeOpen(int32_t vgId) {
vError("vgId:%d, failed to read config file, set cfgVersion to 0", pVnode->vgId); vError("vgId:%d, failed to read config file, set cfgVersion to 0", pVnode->vgId);
vnodeCleanUp(pVnode); vnodeCleanUp(pVnode);
return 0; return 0;
} }
code = vnodeReadVersion(pVnode); code = vnodeReadVersion(pVnode);
if (code != TSDB_CODE_SUCCESS) { if (code != TSDB_CODE_SUCCESS) {
...@@ -291,29 +370,15 @@ int32_t vnodeOpen(int32_t vgId) { ...@@ -291,29 +370,15 @@ int32_t vnodeOpen(int32_t vgId) {
} }
pVnode->fversion = pVnode->version; pVnode->fversion = pVnode->version;
pVnode->wqueue = dnodeAllocVWriteQueue(pVnode); pVnode->wqueue = vnodeAllocWriteQueue(pVnode);
pVnode->qqueue = dnodeAllocVQueryQueue(pVnode); pVnode->qqueue = vnodeAllocQueryQueue(pVnode);
pVnode->fqueue = dnodeAllocVFetchQueue(pVnode); pVnode->fqueue = vnodeAllocFetchQueue(pVnode);
if (pVnode->wqueue == NULL || pVnode->qqueue == NULL || pVnode->fqueue == NULL) { if (pVnode->wqueue == NULL || pVnode->qqueue == NULL || pVnode->fqueue == NULL) {
vnodeCleanUp(pVnode); vnodeCleanUp(pVnode);
return terrno; return terrno;
} }
if (tsEnableStream) {
SCqCfg cqCfg = {0};
sprintf(cqCfg.user, "_root");
strcpy(cqCfg.pass, tsInternalPass);
strcpy(cqCfg.db, pVnode->db);
cqCfg.vgId = vgId;
cqCfg.cqWrite = vnodeWriteToCache;
pVnode->cq = cqOpen(pVnode, &cqCfg);
if (pVnode->cq == NULL) {
vnodeCleanUp(pVnode);
return terrno;
}
}
STsdbAppH appH = {0}; STsdbAppH appH = {0};
appH.appH = (void *)pVnode; appH.appH = (void *)pVnode;
appH.notifyStatus = vnodeProcessTsdbStatus; appH.notifyStatus = vnodeProcessTsdbStatus;
...@@ -327,8 +392,8 @@ int32_t vnodeOpen(int32_t vgId) { ...@@ -327,8 +392,8 @@ int32_t vnodeOpen(int32_t vgId) {
vnodeCleanUp(pVnode); vnodeCleanUp(pVnode);
return terrno; return terrno;
} else if (tsdbGetState(pVnode->tsdb) != TSDB_STATE_OK) { } else if (tsdbGetState(pVnode->tsdb) != TSDB_STATE_OK) {
vError("vgId:%d, failed to open tsdb(state: %d), replica:%d reason:%s", pVnode->vgId, vError("vgId:%d, failed to open tsdb(state: %d), replica:%d reason:%s", pVnode->vgId, tsdbGetState(pVnode->tsdb),
tsdbGetState(pVnode->tsdb), pVnode->syncCfg.replica, tstrerror(terrno)); pVnode->syncCfg.replica, tstrerror(terrno));
if (pVnode->syncCfg.replica <= 1) { if (pVnode->syncCfg.replica <= 1) {
vnodeCleanUp(pVnode); vnodeCleanUp(pVnode);
return TSDB_CODE_VND_INVALID_TSDB_STATE; return TSDB_CODE_VND_INVALID_TSDB_STATE;
...@@ -355,12 +420,12 @@ int32_t vnodeOpen(int32_t vgId) { ...@@ -355,12 +420,12 @@ int32_t vnodeOpen(int32_t vgId) {
sprintf(temp, "%s/wal", walRootDir); sprintf(temp, "%s/wal", walRootDir);
pVnode->walCfg.vgId = pVnode->vgId; pVnode->walCfg.vgId = pVnode->vgId;
pVnode->wal = walOpen(temp, &pVnode->walCfg); pVnode->wal = walOpen(temp, &pVnode->walCfg);
if (pVnode->wal == NULL) { if (pVnode->wal == NULL) {
vnodeCleanUp(pVnode); vnodeCleanUp(pVnode);
return terrno; return terrno;
} }
walRestore(pVnode->wal, pVnode, vnodeProcessWrite); walRestore(pVnode->wal, pVnode, (FWalWrite)vnodeProcessWalMsg);
if (pVnode->version == 0) { if (pVnode->version == 0) {
pVnode->fversion = 0; pVnode->fversion = 0;
pVnode->version = walGetVersion(pVnode->wal); pVnode->version = walGetVersion(pVnode->wal);
...@@ -386,40 +451,16 @@ int32_t vnodeOpen(int32_t vgId) { ...@@ -386,40 +451,16 @@ int32_t vnodeOpen(int32_t vgId) {
vDebug("vgId:%d, vnode is opened in %s - %s, pVnode:%p", pVnode->vgId, rootDir, walRootDir, pVnode); vDebug("vgId:%d, vnode is opened in %s - %s, pVnode:%p", pVnode->vgId, rootDir, walRootDir, pVnode);
vnodeAddIntoHash(pVnode); taosHashPut(tsVmain.hash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnode *));
SSyncInfo syncInfo;
syncInfo.vgId = pVnode->vgId;
syncInfo.version = pVnode->version;
syncInfo.syncCfg = pVnode->syncCfg;
tstrncpy(syncInfo.path, walRootDir, TSDB_FILENAME_LEN);
syncInfo.getWalInfoFp = vnodeGetWalInfo;
syncInfo.writeToCacheFp = vnodeWriteToCache;
syncInfo.confirmForward = vnodeConfirmForard;
syncInfo.notifyRoleFp = vnodeNotifyRole;
syncInfo.notifyFlowCtrlFp = vnodeCtrlFlow;
syncInfo.startSyncFileFp = vnodeStartSyncFile;
syncInfo.stopSyncFileFp = vnodeStopSyncFile;
syncInfo.getVersionFp = vnodeGetVersion;
syncInfo.sendFileFp = tsdbSyncSend;
syncInfo.recvFileFp = tsdbSyncRecv;
syncInfo.pTsdb = pVnode->tsdb;
pVnode->sync = syncStart(&syncInfo);
if (pVnode->sync <= 0) {
vError("vgId:%d, failed to open sync, replica:%d reason:%s", pVnode->vgId, pVnode->syncCfg.replica,
tstrerror(terrno));
vnodeRemoveFromHash(pVnode);
vnodeCleanUp(pVnode);
return terrno;
}
vnodeSetReadyStatus(pVnode); vnodeSetReadyStatus(pVnode);
pVnode->role = TAOS_SYNC_ROLE_MASTER;
#endif
return TSDB_CODE_SUCCESS; return TSDB_CODE_SUCCESS;
} }
int32_t vnodeClose(int32_t vgId) { int32_t vnodeClose(int32_t vgId) {
SVnodeObj *pVnode = vnodeAcquireNotClose(vgId); SVnode *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) return 0; if (pVnode == NULL) return 0;
if (pVnode->dropped) { if (pVnode->dropped) {
vnodeRelease(pVnode); vnodeRelease(pVnode);
...@@ -435,10 +476,11 @@ int32_t vnodeClose(int32_t vgId) { ...@@ -435,10 +476,11 @@ int32_t vnodeClose(int32_t vgId) {
return 0; return 0;
} }
void vnodeDestroy(SVnodeObj *pVnode) { void vnodeDestroy(SVnode *pVnode) {
#if 0
int32_t code = 0; int32_t code = 0;
int32_t vgId = pVnode->vgId; int32_t vgId = pVnode->vgId;
if (pVnode->qMgmt) { if (pVnode->qMgmt) {
qCleanupQueryMgmt(pVnode->qMgmt); qCleanupQueryMgmt(pVnode->qMgmt);
pVnode->qMgmt = NULL; pVnode->qMgmt = NULL;
...@@ -475,47 +517,49 @@ void vnodeDestroy(SVnodeObj *pVnode) { ...@@ -475,47 +517,49 @@ void vnodeDestroy(SVnodeObj *pVnode) {
} }
if (pVnode->wqueue) { if (pVnode->wqueue) {
dnodeFreeVWriteQueue(pVnode->wqueue); vnodeFreeWriteQueue(pVnode->wqueue);
pVnode->wqueue = NULL; pVnode->wqueue = NULL;
} }
if (pVnode->qqueue) { if (pVnode->qqueue) {
dnodeFreeVQueryQueue(pVnode->qqueue); vnodeFreeQueryQueue(pVnode->qqueue);
pVnode->qqueue = NULL; pVnode->qqueue = NULL;
} }
if (pVnode->fqueue) { if (pVnode->fqueue) {
dnodeFreeVFetchQueue(pVnode->fqueue); vnodeFreeFetchQueue(pVnode->fqueue);
pVnode->fqueue = NULL; pVnode->fqueue = NULL;
} }
tfree(pVnode->rootDir); tfree(pVnode->rootDir);
if (pVnode->dropped) { if (pVnode->dropped) {
char rootDir[TSDB_FILENAME_LEN] = {0}; char rootDir[TSDB_FILENAME_LEN] = {0};
char stagingDir[TSDB_FILENAME_LEN] = {0}; char stagingDir[TSDB_FILENAME_LEN] = {0};
sprintf(rootDir, "%s/vnode%d", "vnode", vgId); sprintf(rootDir, "%s/vnode%d", "vnode", vgId);
sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId); sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId);
tfsRename(rootDir, stagingDir); tfsRename(rootDir, stagingDir);
vnodeBackup(vgId); vnodeProcessBackupTask(pVnode);
dnodeSendStatusMsgToMnode(); // dnodeSendStatusMsgToMnode();
} }
tsem_destroy(&pVnode->sem); tsem_destroy(&pVnode->sem);
pthread_mutex_destroy(&pVnode->statusMutex); pthread_mutex_destroy(&pVnode->statusMutex);
free(pVnode); free(pVnode);
tsdbDecCommitRef(vgId); tsdbDecCommitRef(vgId);
#endif
} }
void vnodeCleanUp(SVnodeObj *pVnode) { void vnodeCleanUp(SVnode *pVnode) {
#if 0
vDebug("vgId:%d, vnode will cleanup, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode); vDebug("vgId:%d, vnode will cleanup, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
vnodeSetClosingStatus(pVnode); vnodeSetClosingStatus(pVnode);
vnodeRemoveFromHash(pVnode); taosHashRemove(tsVmain.hash, &pVnode->vgId, sizeof(int32_t));
// stop replication module // stop replication module
if (pVnode->sync > 0) { if (pVnode->sync > 0) {
...@@ -526,10 +570,12 @@ void vnodeCleanUp(SVnodeObj *pVnode) { ...@@ -526,10 +570,12 @@ void vnodeCleanUp(SVnodeObj *pVnode) {
vDebug("vgId:%d, vnode is cleaned, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode); vDebug("vgId:%d, vnode is cleaned, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
vnodeRelease(pVnode); vnodeRelease(pVnode);
#endif
} }
#if 0
static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) { static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) {
SVnodeObj *pVnode = arg; SVnode *pVnode = arg;
if (eno != TSDB_CODE_SUCCESS) { if (eno != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to commit since %s, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, tstrerror(eno), vError("vgId:%d, failed to commit since %s, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, tstrerror(eno),
...@@ -561,9 +607,301 @@ static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) { ...@@ -561,9 +607,301 @@ static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) {
} }
// timer thread callback // timer thread callback
if(status == TSDB_STATUS_COMMIT_NOBLOCK) { if (status == TSDB_STATUS_COMMIT_NOBLOCK) {
qSolveCommitNoBlock(pVnode->tsdb, pVnode->qMgmt); qSolveCommitNoBlock(pVnode->tsdb, pVnode->qMgmt);
} }
return 0; return 0;
} }
#endif
static void *vnodeOpenVnode(void *param) {
SOpenVnodeThread *pThread = param;
vDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
setThreadName("vnodeOpenVnode");
for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
int32_t vgId = pThread->vnodeList[v];
char stepDesc[TSDB_STEP_DESC_LEN] = {0};
snprintf(stepDesc, TSDB_STEP_DESC_LEN, "vgId:%d, start to restore, %d of %d have been opened", vgId,
tsVmain.openVnodes, tsVmain.totalVnodes);
// (*vnodeInst()->fp.ReportStartup)("open-vnodes", stepDesc);
if (vnodeOpen(vgId) < 0) {
vError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex);
pThread->failed++;
} else {
vDebug("vgId:%d, is opened by thread:%d", vgId, pThread->threadIndex);
pThread->opened++;
}
atomic_add_fetch_32(&tsVmain.openVnodes, 1);
}
vDebug("thread:%d, total vnodes:%d, opened:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
pThread->failed);
return NULL;
}
static int32_t vnodeGetVnodeListFromDisk(int32_t vnodeList[], int32_t *numOfVnodes) {
#if 0
DIR *dir = opendir(tsVnodeDir);
if (dir == NULL) return TSDB_CODE_DND_NO_WRITE_ACCESS;
*numOfVnodes = 0;
struct dirent *de = NULL;
while ((de = readdir(dir)) != NULL) {
if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
if (de->d_type & DT_DIR) {
if (strncmp("vnode", de->d_name, 5) != 0) continue;
int32_t vnode = atoi(de->d_name + 5);
if (vnode == 0) continue;
(*numOfVnodes)++;
if (*numOfVnodes >= TSDB_MAX_VNODES) {
vError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES);
closedir(dir);
return TSDB_CODE_DND_TOO_MANY_VNODES;
} else {
vnodeList[*numOfVnodes - 1] = vnode;
}
}
}
closedir(dir);
#endif
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeOpenVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t status = vnodeGetVnodeListFromDisk(vnodeList, &numOfVnodes);
if (status != TSDB_CODE_SUCCESS) {
vInfo("failed to get vnode list from disk since code:%d", status);
return status;
}
tsVmain.totalVnodes = numOfVnodes;
int32_t threadNum = tsNumOfCores;
int32_t vnodesPerThread = numOfVnodes / threadNum + 1;
SOpenVnodeThread *threads = calloc(threadNum, sizeof(SOpenVnodeThread));
for (int32_t t = 0; t < threadNum; ++t) {
threads[t].threadIndex = t;
threads[t].vnodeList = calloc(vnodesPerThread, sizeof(int32_t));
}
for (int32_t v = 0; v < numOfVnodes; ++v) {
int32_t t = v % threadNum;
SOpenVnodeThread *pThread = &threads[t];
pThread->vnodeList[pThread->vnodeNum++] = vnodeList[v];
}
vInfo("start %d threads to open %d vnodes", threadNum, numOfVnodes);
for (int32_t t = 0; t < threadNum; ++t) {
SOpenVnodeThread *pThread = &threads[t];
if (pThread->vnodeNum == 0) continue;
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&pThread->thread, &thAttr, vnodeOpenVnode, pThread) != 0) {
vError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno));
}
pthread_attr_destroy(&thAttr);
}
int32_t openVnodes = 0;
int32_t failedVnodes = 0;
for (int32_t t = 0; t < threadNum; ++t) {
SOpenVnodeThread *pThread = &threads[t];
if (pThread->vnodeNum > 0 && taosCheckPthreadValid(pThread->thread)) {
pthread_join(pThread->thread, NULL);
}
openVnodes += pThread->opened;
failedVnodes += pThread->failed;
free(pThread->vnodeList);
}
free(threads);
vInfo("there are total vnodes:%d, opened:%d", numOfVnodes, openVnodes);
if (failedVnodes != 0) {
vError("there are total vnodes:%d, failed:%d", numOfVnodes, failedVnodes);
return -1;
}
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) {
void *pIter = taosHashIterate(tsVmain.hash, NULL);
while (pIter) {
SVnode **pVnode = pIter;
if (*pVnode) {
(*numOfVnodes)++;
if (*numOfVnodes >= TSDB_MAX_VNODES) {
vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES);
continue;
} else {
vnodeList[*numOfVnodes - 1] = (*pVnode)->vgId;
}
}
pIter = taosHashIterate(tsVmain.hash, pIter);
}
return TSDB_CODE_SUCCESS;
}
static void vnodeCleanupVnodes() {
int32_t vnodeList[TSDB_MAX_VNODES] = {0};
int32_t numOfVnodes = 0;
int32_t code = vnodeGetVnodeList(vnodeList, &numOfVnodes);
if (code != TSDB_CODE_SUCCESS) {
vInfo("failed to get dnode list since code %d", code);
return;
}
for (int32_t i = 0; i < numOfVnodes; ++i) {
vnodeClose(vnodeList[i]);
}
vInfo("total vnodes:%d are all closed", numOfVnodes);
}
static void vnodeInitMsgFp() {
tsVmain.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessMgmtMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessMgmtMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessMgmtMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessMgmtMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessMgmtMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessMgmtMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessWriteMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessWriteMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessWriteMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessWriteMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessWriteMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessWriteMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessReadMsg;
tsVmain.msgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessReadMsg;
}
void vnodeProcessMsg(SRpcMsg *pMsg) {
if (tsVmain.msgFp[pMsg->msgType]) {
(*tsVmain.msgFp[pMsg->msgType])(pMsg);
} else {
assert(0);
}
}
int32_t vnodeInitMain() {
vnodeInitMsgFp();
tsVmain.timer = taosTmrInit(100, 200, 60000, "VND-TIMER");
if (tsVmain.timer == NULL) {
vError("failed to init vnode timer");
return -1;
}
tsVmain.hash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK);
if (tsVmain.hash == NULL) {
taosTmrCleanUp(tsVmain.timer);
vError("failed to init vnode mgmt");
return -1;
}
vInfo("vnode main is initialized");
return vnodeOpenVnodes();
}
void vnodeCleanupMain() {
taosTmrCleanUp(tsVmain.timer);
tsVmain.timer = NULL;
vnodeCleanupVnodes();
taosHashCleanup(tsVmain.hash);
tsVmain.hash = NULL;
}
static void vnodeBuildVloadMsg(SVnode *pVnode, SStatusMsg *pStatus) {
#if 0
int64_t totalStorage = 0;
int64_t compStorage = 0;
int64_t pointsWritten = 0;
if (vnodeInClosingStatus(pVnode)) return;
if (pStatus->openVnodes >= TSDB_MAX_VNODES) return;
if (pVnode->tsdb) {
tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage);
}
SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++];
pLoad->vgId = htonl(pVnode->vgId);
pLoad->dbCfgVersion = htonl(pVnode->dbCfgVersion);
pLoad->vgCfgVersion = htonl(pVnode->vgCfgVersion);
pLoad->totalStorage = htobe64(totalStorage);
pLoad->compStorage = htobe64(compStorage);
pLoad->pointsWritten = htobe64(pointsWritten);
pLoad->vnodeVersion = htobe64(pVnode->version);
pLoad->status = pVnode->status;
pLoad->role = pVnode->role;
pLoad->replica = pVnode->syncCfg.replica;
pLoad->compact = (pVnode->tsdb != NULL) ? tsdbGetCompactState(pVnode->tsdb) : 0;
#endif
}
void vnodeGetStatus(struct SStatusMsg *pStatus) {
void *pIter = taosHashIterate(tsVmain.hash, NULL);
while (pIter) {
SVnode **pVnode = pIter;
if (*pVnode) {
vnodeBuildVloadMsg(*pVnode, pStatus);
}
pIter = taosHashIterate(tsVmain.hash, pIter);
}
}
void vnodeSetAccess(struct SVgroupAccess *pAccess, int32_t numOfVnodes) {
for (int32_t i = 0; i < numOfVnodes; ++i) {
pAccess[i].vgId = htonl(pAccess[i].vgId);
SVnode *pVnode = vnodeAcquireNotClose(pAccess[i].vgId);
if (pVnode != NULL) {
pVnode->accessState = pAccess[i].accessState;
if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) {
vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState);
}
vnodeRelease(pVnode);
}
}
}
void vnodeBackup(int32_t vgId) {
char newDir[TSDB_FILENAME_LEN] = {0};
char stagingDir[TSDB_FILENAME_LEN] = {0};
sprintf(newDir, "%s/vnode%d", "vnode_bak", vgId);
sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId);
#if 0
if (tsEnableVnodeBak) {
tfsRmdir(newDir);
tfsRename(stagingDir, newDir);
} else {
vInfo("vgId:%d, vnode backup not enabled", vgId);
tfsRmdir(stagingDir);
}
#endif
}
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "vnodeMain.h"
#include "vnodeMgmt.h"
#include "vnodeMgmtMsg.h"
typedef struct {
SRpcMsg rpcMsg;
char pCont[];
} SVnMgmtMsg;
static struct {
SWorkerPool pool;
taos_queue pQueue;
int32_t (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
} tsVmgmt = {0};
static int32_t vnodeProcessMgmtStart(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype) {
SRpcMsg *pMsg = &pMgmt->rpcMsg;
int32_t msgType = pMsg->msgType;
if (tsVmgmt.msgFp[msgType]) {
vTrace("msg:%p, ahandle:%p type:%s will be processed", pMgmt, pMsg->ahandle, taosMsg[msgType]);
return (*tsVmgmt.msgFp[msgType])(pMsg);
} else {
vError("msg:%p, ahandle:%p type:%s not processed since no handle", pMgmt, pMsg->ahandle, taosMsg[msgType]);
return TSDB_CODE_DND_MSG_NOT_PROCESSED;
}
}
static void vnodeSendMgmtEnd(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype, int32_t code) {
SRpcMsg *pMsg = &pMgmt->rpcMsg;
SRpcMsg rsp = {0};
rsp.code = code;
vTrace("msg:%p, is processed, code:0x%x", pMgmt, rsp.code);
if (rsp.code != TSDB_CODE_DND_ACTION_IN_PROGRESS) {
rsp.handle = pMsg->handle;
rsp.pCont = NULL;
rpcSendResponse(&rsp);
}
taosFreeQitem(pMsg);
}
static void vnodeInitMgmtReqFp() {
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessCreateVnodeMsg;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessAlterVnodeMsg;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessSyncVnodeMsg;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE]= vnodeProcessCompactVnodeMsg;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessDropVnodeMsg;
tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessAlterStreamReq;
}
static int32_t vnodeWriteToMgmtQueue(SRpcMsg *pMsg) {
int32_t size = sizeof(SVnMgmtMsg) + pMsg->contLen;
SVnMgmtMsg *pMgmt = taosAllocateQitem(size);
if (pMgmt == NULL) return TSDB_CODE_DND_OUT_OF_MEMORY;
pMgmt->rpcMsg = *pMsg;
pMgmt->rpcMsg.pCont = pMgmt->pCont;
memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen);
taosWriteQitem(tsVmgmt.pQueue, TAOS_QTYPE_RPC, pMgmt);
return TSDB_CODE_SUCCESS;
}
void vnodeProcessMgmtMsg(SRpcMsg *pMsg) {
int32_t code = vnodeWriteToMgmtQueue(pMsg);
if (code != TSDB_CODE_SUCCESS) {
SRpcMsg rsp = {.handle = pMsg->handle, .code = code};
rpcSendResponse(&rsp);
}
rpcFreeCont(pMsg->pCont);
}
int32_t vnodeInitMgmt() {
vnodeInitMgmtReqFp();
SWorkerPool *pPool = &tsVmgmt.pool;
pPool->name = "vmgmt";
pPool->startFp = (ProcessStartFp)vnodeProcessMgmtStart;
pPool->endFp = (ProcessEndFp)vnodeSendMgmtEnd;
pPool->min = 1;
pPool->max = 1;
if (tWorkerInit(pPool) != 0) {
return TSDB_CODE_VND_OUT_OF_MEMORY;
}
tsVmgmt.pQueue = tWorkerAllocQueue(pPool, NULL);
vInfo("vmgmt is initialized, max worker %d", pPool->max);
return TSDB_CODE_SUCCESS;
}
void vnodeCleanupMgmt() {
tWorkerFreeQueue(&tsVmgmt.pool, tsVmgmt.pQueue);
tWorkerCleanup(&tsVmgmt.pool);
tsVmgmt.pQueue = NULL;
vInfo("vmgmt is closed");
}
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "vnodeMain.h"
#include "vnodeMgmtMsg.h"
static SCreateVnodeMsg* vnodeParseVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = rpcMsg->pCont;
pCreate->cfg.vgId = htonl(pCreate->cfg.vgId);
pCreate->cfg.dbCfgVersion = htonl(pCreate->cfg.dbCfgVersion);
pCreate->cfg.vgCfgVersion = htonl(pCreate->cfg.vgCfgVersion);
pCreate->cfg.maxTables = htonl(pCreate->cfg.maxTables);
pCreate->cfg.cacheBlockSize = htonl(pCreate->cfg.cacheBlockSize);
pCreate->cfg.totalBlocks = htonl(pCreate->cfg.totalBlocks);
pCreate->cfg.daysPerFile = htonl(pCreate->cfg.daysPerFile);
pCreate->cfg.daysToKeep1 = htonl(pCreate->cfg.daysToKeep1);
pCreate->cfg.daysToKeep2 = htonl(pCreate->cfg.daysToKeep2);
pCreate->cfg.daysToKeep = htonl(pCreate->cfg.daysToKeep);
pCreate->cfg.minRowsPerFileBlock = htonl(pCreate->cfg.minRowsPerFileBlock);
pCreate->cfg.maxRowsPerFileBlock = htonl(pCreate->cfg.maxRowsPerFileBlock);
pCreate->cfg.fsyncPeriod = htonl(pCreate->cfg.fsyncPeriod);
pCreate->cfg.commitTime = htonl(pCreate->cfg.commitTime);
for (int32_t j = 0; j < pCreate->cfg.vgReplica; ++j) {
pCreate->nodes[j].nodeId = htonl(pCreate->nodes[j].nodeId);
}
return pCreate;
}
int32_t vnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) {
SCreateVnodeMsg *pCreate = vnodeParseVnodeMsg(rpcMsg);
SVnode *pVnode = vnodeAcquire(pCreate->cfg.vgId);
if (pVnode != NULL) {
vDebug("vgId:%d, already exist, return success", pCreate->cfg.vgId);
vnodeRelease(pVnode);
return TSDB_CODE_SUCCESS;
} else {
vDebug("vgId:%d, create vnode msg is received", pCreate->cfg.vgId);
return vnodeCreate(pCreate);
}
}
int32_t vnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) {
SAlterVnodeMsg *pAlter = vnodeParseVnodeMsg(rpcMsg);
void *pVnode = vnodeAcquireNotClose(pAlter->cfg.vgId);
if (pVnode != NULL) {
vDebug("vgId:%d, alter vnode msg is received", pAlter->cfg.vgId);
int32_t code = vnodeAlter(pVnode, pAlter);
vnodeRelease(pVnode);
return code;
} else {
vInfo("vgId:%d, vnode not exist, can't alter it", pAlter->cfg.vgId);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
}
int32_t vnodeProcessSyncVnodeMsg(SRpcMsg *rpcMsg) {
SSyncVnodeMsg *pSyncVnode = rpcMsg->pCont;
pSyncVnode->vgId = htonl(pSyncVnode->vgId);
return vnodeSync(pSyncVnode->vgId);
}
int32_t vnodeProcessCompactVnodeMsg(SRpcMsg *rpcMsg) {
SCompactVnodeMsg *pCompactVnode = rpcMsg->pCont;
pCompactVnode->vgId = htonl(pCompactVnode->vgId);
return vnodeCompact(pCompactVnode->vgId);
}
int32_t vnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) {
SDropVnodeMsg *pDrop = rpcMsg->pCont;
pDrop->vgId = htonl(pDrop->vgId);
return vnodeDrop(pDrop->vgId);
}
int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg) { return 0; }
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "taosmsg.h"
#include "tglobal.h"
// #include "query.h"
#include "vnodeMain.h"
#include "vnodeRead.h"
#include "vnodeReadMsg.h"
#include "vnodeStatus.h"
static struct {
SWorkerPool query;
SWorkerPool fetch;
int32_t (*msgFp[TSDB_MSG_TYPE_MAX])(SVnode *, struct SReadMsg *);
} tsVread = {0};
void vnodeStartRead(SVnode *pVnode) {}
void vnodeStopRead(SVnode *pVnode) {}
void vnodeWaitReadCompleted(SVnode *pVnode) {
while (pVnode->queuedRMsg > 0) {
vTrace("vgId:%d, queued rmsg num:%d", pVnode->vgId, pVnode->queuedRMsg);
taosMsleep(10);
}
}
static int32_t vnodeWriteToRQueue(SVnode *pVnode, void *pCont, int32_t contLen, int8_t qtype, SRpcMsg *pRpcMsg) {
if (pVnode->dropped) {
return TSDB_CODE_APP_NOT_READY;
}
#if 0
if (!((pVnode->role == TAOS_SYNC_ROLE_MASTER) || (tsEnableSlaveQuery && pVnode->role == TAOS_SYNC_ROLE_SLAVE))) {
return TSDB_CODE_APP_NOT_READY;
}
#endif
if (!vnodeInReadyStatus(pVnode)) {
vDebug("vgId:%d, failed to write into vread queue, vnode status is %s", pVnode->vgId, vnodeStatus[pVnode->status]);
return TSDB_CODE_APP_NOT_READY;
}
int32_t size = sizeof(SReadMsg) + contLen;
SReadMsg *pRead = taosAllocateQitem(size);
if (pRead == NULL) {
return TSDB_CODE_VND_OUT_OF_MEMORY;
}
if (pRpcMsg != NULL) {
pRead->rpcHandle = pRpcMsg->handle;
pRead->rpcAhandle = pRpcMsg->ahandle;
pRead->msgType = pRpcMsg->msgType;
pRead->code = pRpcMsg->code;
}
if (contLen != 0) {
pRead->contLen = contLen;
memcpy(pRead->pCont, pCont, contLen);
} else {
pRead->qhandle = pCont;
}
pRead->qtype = qtype;
atomic_add_fetch_32(&pVnode->refCount, 1);
atomic_add_fetch_32(&pVnode->queuedRMsg, 1);
if (pRead->code == TSDB_CODE_RPC_NETWORK_UNAVAIL || pRead->msgType == TSDB_MSG_TYPE_FETCH) {
return taosWriteQitem(pVnode->fqueue, qtype, pRead);
} else {
return taosWriteQitem(pVnode->qqueue, qtype, pRead);
}
}
static void vnodeFreeFromRQueue(SVnode *pVnode, SReadMsg *pRead) {
atomic_sub_fetch_32(&pVnode->queuedRMsg, 1);
taosFreeQitem(pRead);
vnodeRelease(pVnode);
}
int32_t vnodeReputPutToRQueue(SVnode *pVnode, void **qhandle, void *ahandle) {
SRpcMsg rpcMsg = {0};
rpcMsg.msgType = TSDB_MSG_TYPE_QUERY;
rpcMsg.ahandle = ahandle;
int32_t code = vnodeWriteToRQueue(pVnode, qhandle, 0, TAOS_QTYPE_QUERY, &rpcMsg);
if (code == TSDB_CODE_SUCCESS) {
vTrace("QInfo:%p add to vread queue for exec query", *qhandle);
}
return code;
}
void vnodeProcessReadMsg(SRpcMsg *pMsg) {
int32_t queuedMsgNum = 0;
int32_t leftLen = pMsg->contLen;
int32_t code = TSDB_CODE_VND_INVALID_VGROUP_ID;
char * pCont = pMsg->pCont;
while (leftLen > 0) {
SMsgHead *pHead = (SMsgHead *)pCont;
pHead->vgId = htonl(pHead->vgId);
pHead->contLen = htonl(pHead->contLen);
assert(pHead->contLen > 0);
SVnode *pVnode = vnodeAcquireNotClose(pHead->vgId);
if (pVnode != NULL) {
code = vnodeWriteToRQueue(pVnode, pCont, pHead->contLen, TAOS_QTYPE_RPC, pMsg);
if (code == TSDB_CODE_SUCCESS) queuedMsgNum++;
vnodeRelease(pVnode);
}
leftLen -= pHead->contLen;
pCont -= pHead->contLen;
}
if (queuedMsgNum == 0) {
SRpcMsg rpcRsp = {.handle = pMsg->handle, .code = code};
rpcSendResponse(&rpcRsp);
}
rpcFreeCont(pMsg->pCont);
}
static void vnodeInitReadMsgFp() {
tsVread.msgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessQueryMsg;
tsVread.msgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessFetchMsg;
}
static int32_t vnodeProcessReadStart(SVnode *pVnode, SReadMsg *pRead, int32_t qtype) {
int32_t msgType = pRead->msgType;
if (tsVread.msgFp[msgType] == NULL) {
vDebug("vgId:%d, msgType:%s not processed, no handle", pVnode->vgId, taosMsg[msgType]);
return TSDB_CODE_VND_MSG_NOT_PROCESSED;
} else {
vTrace("msg:%p, app:%p type:%s will be processed", pRead, pRead->rpcAhandle, taosMsg[msgType]);
}
return (*tsVread.msgFp[msgType])(pVnode, pRead);
}
static void vnodeSendReadRsp(SReadMsg *pRead, int32_t code) {
SRpcMsg rpcRsp = {
.handle = pRead->rpcHandle,
.pCont = pRead->rspRet.rsp,
.contLen = pRead->rspRet.len,
.code = code,
};
rpcSendResponse(&rpcRsp);
}
static void vnodeProcessReadEnd(SVnode *pVnode, SReadMsg *pRead, int32_t qtype, int32_t code) {
if (qtype == TAOS_QTYPE_RPC && code != TSDB_CODE_QRY_NOT_READY) {
vnodeSendReadRsp(pRead, code);
} else {
if (code == TSDB_CODE_QRY_HAS_RSP) {
vnodeSendReadRsp(pRead, pRead->code);
} else { // code == TSDB_CODE_QRY_NOT_READY, do not return msg to client
assert(pRead->rpcHandle == NULL || (pRead->rpcHandle != NULL && pRead->msgType == 5));
}
}
vnodeFreeFromRQueue(pVnode, pRead);
}
int32_t vnodeInitRead() {
vnodeInitReadMsgFp();
int32_t maxFetchThreads = 4;
float threadsForQuery = MAX(tsNumOfCores * tsRatioOfQueryCores, 1);
SWorkerPool *pPool = &tsVread.query;
pPool->name = "vquery";
pPool->startFp = (ProcessStartFp)vnodeProcessReadStart;
pPool->endFp = (ProcessEndFp)vnodeProcessReadEnd;
pPool->min = (int32_t)threadsForQuery;
pPool->max = pPool->min;
if (tWorkerInit(pPool) != 0) return -1;
pPool = &tsVread.fetch;
pPool->name = "vfetch";
pPool->startFp = (ProcessStartFp)vnodeProcessReadStart;
pPool->endFp = (ProcessEndFp)vnodeProcessReadEnd;
pPool->min = MIN(maxFetchThreads, tsNumOfCores);
pPool->max = pPool->min;
if (tWorkerInit(pPool) != 0) return -1;
vInfo("vread is initialized, max worker %d", pPool->max);
return 0;
}
void vnodeCleanupRead() {
tWorkerCleanup(&tsVread.fetch);
tWorkerCleanup(&tsVread.query);
vInfo("vread is closed");
}
taos_queue vnodeAllocQueryQueue(SVnode *pVnode) { return tWorkerAllocQueue(&tsVread.query, pVnode); }
taos_queue vnodeAllocFetchQueue(SVnode *pVnode) { return tWorkerAllocQueue(&tsVread.fetch, pVnode); }
void vnodeFreeQueryQueue(taos_queue pQueue) { tWorkerFreeQueue(&tsVread.query, pQueue); }
void vnodeFreeFetchQueue(taos_queue pQueue) { tWorkerFreeQueue(&tsVread.fetch, pQueue); }
...@@ -16,155 +16,26 @@ ...@@ -16,155 +16,26 @@
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "os.h"
#include "taosmsg.h" #include "taosmsg.h"
#include "tqueue.h"
#include "tglobal.h" #include "tglobal.h"
#include "query.h" // #include "query.h"
#include "vnodeStatus.h" #include "vnodeStatus.h"
#include "vnodeRead.h"
#include "vnodeReadMsg.h"
int32_t vNumOfExistedQHandle; // current initialized and existed query handle in current dnode #if 0
// notify connection(handle) that current qhandle is created, if current connection from
static int32_t (*vnodeProcessReadMsgFp[TSDB_MSG_TYPE_MAX])(SVnodeObj *pVnode, SVReadMsg *pRead); // client is broken, the query needs to be killed immediately.
static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead); static int32_t vnodeNotifyCurrentQhandle(void *handle, uint64_t qId, void *qhandle, int32_t vgId) {
static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead); SRetrieveTableMsg *pMsg = rpcMallocCont(sizeof(SRetrieveTableMsg));
pMsg->qId = htobe64(qId);
static int32_t vnodeNotifyCurrentQhandle(void* handle, uint64_t qId, void* qhandle, int32_t vgId); pMsg->header.vgId = htonl(vgId);
pMsg->header.contLen = htonl(sizeof(SRetrieveTableMsg));
int32_t vnodeInitRead(void) {
vnodeProcessReadMsgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessQueryMsg;
vnodeProcessReadMsgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessFetchMsg;
return 0;
}
void vnodeCleanupRead() {}
//
// After the fetch request enters the vnode queue, if the vnode cannot provide services, the process function are
// still required, or there will be a deadlock, so we don’t do any check here, but put the check codes before the
// request enters the queue
//
int32_t vnodeProcessRead(void *vparam, SVReadMsg *pRead) {
SVnodeObj *pVnode = vparam;
int32_t msgType = pRead->msgType;
if (vnodeProcessReadMsgFp[msgType] == NULL) {
vDebug("vgId:%d, msgType:%s not processed, no handle", pVnode->vgId, taosMsg[msgType]);
return TSDB_CODE_VND_MSG_NOT_PROCESSED;
}
return (*vnodeProcessReadMsgFp[msgType])(pVnode, pRead);
}
static int32_t vnodeCheckRead(SVnodeObj *pVnode) {
if (!vnodeInReadyStatus(pVnode)) {
vDebug("vgId:%d, vnode status is %s, refCount:%d pVnode:%p", pVnode->vgId, vnodeStatus[pVnode->status],
pVnode->refCount, pVnode);
return TSDB_CODE_APP_NOT_READY;
}
// tsdb may be in reset state
if (pVnode->tsdb == NULL) {
vDebug("vgId:%d, tsdb is null, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
return TSDB_CODE_APP_NOT_READY;
}
if (pVnode->role == TAOS_SYNC_ROLE_MASTER) {
return TSDB_CODE_SUCCESS;
}
if (tsEnableSlaveQuery && pVnode->role == TAOS_SYNC_ROLE_SLAVE) {
return TSDB_CODE_SUCCESS;
}
vDebug("vgId:%d, replica:%d role:%s, refCount:%d pVnode:%p, cant provide query service", pVnode->vgId, pVnode->syncCfg.replica,
syncRole[pVnode->role], pVnode->refCount, pVnode);
return TSDB_CODE_APP_NOT_READY;
}
void vnodeFreeFromRQueue(void *vparam, SVReadMsg *pRead) {
SVnodeObj *pVnode = vparam;
atomic_sub_fetch_32(&pVnode->queuedRMsg, 1);
vTrace("vgId:%d, free from vrqueue, refCount:%d queued:%d", pVnode->vgId, pVnode->refCount, pVnode->queuedRMsg);
taosFreeQitem(pRead);
vnodeRelease(pVnode);
}
static SVReadMsg *vnodeBuildVReadMsg(SVnodeObj *pVnode, void *pCont, int32_t contLen, int8_t qtype, SRpcMsg *pRpcMsg) {
int32_t size = sizeof(SVReadMsg) + contLen;
SVReadMsg *pRead = taosAllocateQitem(size);
if (pRead == NULL) {
terrno = TSDB_CODE_VND_OUT_OF_MEMORY;
return NULL;
}
if (pRpcMsg != NULL) {
pRead->rpcHandle = pRpcMsg->handle;
pRead->rpcAhandle = pRpcMsg->ahandle;
pRead->msgType = pRpcMsg->msgType;
pRead->code = pRpcMsg->code;
}
if (contLen != 0) {
pRead->contLen = contLen;
memcpy(pRead->pCont, pCont, contLen);
} else {
pRead->qhandle = pCont;
}
pRead->qtype = qtype;
atomic_add_fetch_32(&pVnode->refCount, 1);
return pRead;
}
int32_t vnodeWriteToRQueue(void *vparam, void *pCont, int32_t contLen, int8_t qtype, void *rparam) {
SVnodeObj *pVnode = vparam;
if (pVnode->dropped) {
return TSDB_CODE_APP_NOT_READY;
}
SVReadMsg *pRead = vnodeBuildVReadMsg(vparam, pCont, contLen, qtype, rparam);
if (pRead == NULL) {
assert(terrno != 0);
return terrno;
}
int32_t code = vnodeCheckRead(pVnode);
if (code != TSDB_CODE_SUCCESS) {
taosFreeQitem(pRead);
vnodeRelease(pVnode);
return code;
}
atomic_add_fetch_32(&pVnode->queuedRMsg, 1);
if (pRead->code == TSDB_CODE_RPC_NETWORK_UNAVAIL || pRead->msgType == TSDB_MSG_TYPE_FETCH) {
vTrace("vgId:%d, write into vfetch queue, refCount:%d queued:%d", pVnode->vgId, pVnode->refCount,
pVnode->queuedRMsg);
return taosWriteQitem(pVnode->fqueue, qtype, pRead);
} else {
vTrace("vgId:%d, write into vquery queue, refCount:%d queued:%d", pVnode->vgId, pVnode->refCount,
pVnode->queuedRMsg);
return taosWriteQitem(pVnode->qqueue, qtype, pRead);
}
}
static int32_t vnodePutItemIntoReadQueue(SVnodeObj *pVnode, void **qhandle, void *ahandle) {
SRpcMsg rpcMsg = {0};
rpcMsg.msgType = TSDB_MSG_TYPE_QUERY;
rpcMsg.ahandle = ahandle;
int32_t code = vnodeWriteToRQueue(pVnode, qhandle, 0, TAOS_QTYPE_QUERY, &rpcMsg);
if (code == TSDB_CODE_SUCCESS) {
vTrace("QInfo:%p add to vread queue for exec query", *qhandle);
}
return code; vTrace("QInfo:0x%" PRIx64 "-%p register qhandle to connect:%p", qId, qhandle, handle);
return rpcReportProgress(handle, (char *)pMsg, sizeof(SRetrieveTableMsg));
} }
/** /**
*
* @param pRet response message object * @param pRet response message object
* @param pVnode the vnode object * @param pVnode the vnode object
* @param handle qhandle for executing query * @param handle qhandle for executing query
...@@ -172,14 +43,16 @@ static int32_t vnodePutItemIntoReadQueue(SVnodeObj *pVnode, void **qhandle, void ...@@ -172,14 +43,16 @@ static int32_t vnodePutItemIntoReadQueue(SVnodeObj *pVnode, void **qhandle, void
* @param ahandle sqlObj address at client side * @param ahandle sqlObj address at client side
* @return * @return
*/ */
static int32_t vnodeDumpQueryResult(SRspRet *pRet, void *pVnode, uint64_t qId, void **handle, bool *freeHandle, void *ahandle) { static int32_t vnodeDumpQueryResult(SVnRsp *pRet, void *pVnode, uint64_t qId, void **handle, bool *freeHandle,
void *ahandle) {
bool continueExec = false; bool continueExec = false;
int32_t code = TSDB_CODE_SUCCESS; int32_t code = TSDB_CODE_SUCCESS;
if ((code = qDumpRetrieveResult(*handle, (SRetrieveTableRsp **)&pRet->rsp, &pRet->len, &continueExec)) == TSDB_CODE_SUCCESS) { if ((code = qDumpRetrieveResult(*handle, (SRetrieveTableRsp **)&pRet->rsp, &pRet->len, &continueExec)) ==
TSDB_CODE_SUCCESS) {
if (continueExec) { if (continueExec) {
*freeHandle = false; *freeHandle = false;
code = vnodePutItemIntoReadQueue(pVnode, handle, ahandle); code = vnodeReputPutToRQueue(pVnode, handle, ahandle);
if (code != TSDB_CODE_SUCCESS) { if (code != TSDB_CODE_SUCCESS) {
*freeHandle = true; *freeHandle = true;
return code; return code;
...@@ -188,7 +61,7 @@ static int32_t vnodeDumpQueryResult(SRspRet *pRet, void *pVnode, uint64_t qId, v ...@@ -188,7 +61,7 @@ static int32_t vnodeDumpQueryResult(SRspRet *pRet, void *pVnode, uint64_t qId, v
} }
} else { } else {
*freeHandle = true; *freeHandle = true;
vTrace("QInfo:0x%"PRIx64"-%p exec completed, free handle:%d", qId, *handle, *freeHandle); vTrace("QInfo:0x%" PRIx64 "-%p exec completed, free handle:%d", qId, *handle, *freeHandle);
} }
} else { } else {
SRetrieveTableRsp *pRsp = (SRetrieveTableRsp *)rpcMallocCont(sizeof(SRetrieveTableRsp)); SRetrieveTableRsp *pRsp = (SRetrieveTableRsp *)rpcMallocCont(sizeof(SRetrieveTableRsp));
...@@ -203,7 +76,7 @@ static int32_t vnodeDumpQueryResult(SRspRet *pRet, void *pVnode, uint64_t qId, v ...@@ -203,7 +76,7 @@ static int32_t vnodeDumpQueryResult(SRspRet *pRet, void *pVnode, uint64_t qId, v
return code; return code;
} }
static void vnodeBuildNoResultQueryRsp(SRspRet *pRet) { static void vnodeBuildNoResultQueryRsp(SVnRsp *pRet) {
pRet->rsp = (SRetrieveTableRsp *)rpcMallocCont(sizeof(SRetrieveTableRsp)); pRet->rsp = (SRetrieveTableRsp *)rpcMallocCont(sizeof(SRetrieveTableRsp));
pRet->len = sizeof(SRetrieveTableRsp); pRet->len = sizeof(SRetrieveTableRsp);
...@@ -212,15 +85,16 @@ static void vnodeBuildNoResultQueryRsp(SRspRet *pRet) { ...@@ -212,15 +85,16 @@ static void vnodeBuildNoResultQueryRsp(SRspRet *pRet) {
pRsp->completed = true; pRsp->completed = true;
} }
#endif
int32_t vnodeProcessQueryMsg(SVnode *pVnode, SReadMsg *pRead) {
static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { #if 0
void * pCont = pRead->pCont; void * pCont = pRead->pCont;
int32_t contLen = pRead->contLen; int32_t contLen = pRead->contLen;
SRspRet *pRet = &pRead->rspRet; SVnRsp *pRet = &pRead->rspRet;
SQueryTableMsg *pQueryTableMsg = (SQueryTableMsg *)pCont; SQueryTableMsg *pQueryTableMsg = (SQueryTableMsg *)pCont;
memset(pRet, 0, sizeof(SRspRet)); memset(pRet, 0, sizeof(SVnRsp));
// qHandle needs to be freed correctly // qHandle needs to be freed correctly
if (pRead->code == TSDB_CODE_RPC_NETWORK_UNAVAIL) { if (pRead->code == TSDB_CODE_RPC_NETWORK_UNAVAIL) {
...@@ -231,13 +105,13 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -231,13 +105,13 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
void ** handle = NULL; void ** handle = NULL;
if (contLen != 0) { if (contLen != 0) {
qinfo_t pQInfo = NULL; qinfo_t pQInfo = NULL;
uint64_t qId = genQueryId(); uint64_t qId = genQueryId();
code = qCreateQueryInfo(pVnode->tsdb, pVnode->vgId, pQueryTableMsg, &pQInfo, qId); code = qCreateQueryInfo(pVnode->tsdb, pVnode->vgId, pQueryTableMsg, &pQInfo, qId);
SQueryTableRsp *pRsp = (SQueryTableRsp *)rpcMallocCont(sizeof(SQueryTableRsp)); SQueryTableRsp *pRsp = (SQueryTableRsp *)rpcMallocCont(sizeof(SQueryTableRsp));
pRsp->code = code; pRsp->code = code;
pRsp->qId = 0; pRsp->qId = 0;
pRet->len = sizeof(SQueryTableRsp); pRet->len = sizeof(SQueryTableRsp);
pRet->rsp = pRsp; pRet->rsp = pRsp;
...@@ -250,8 +124,8 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -250,8 +124,8 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
pRsp->code = terrno; pRsp->code = terrno;
terrno = 0; terrno = 0;
vError("vgId:%d, QInfo:0x%"PRIx64 "-%p register qhandle failed, return to app, code:%s,", pVnode->vgId, qId, (void *)pQInfo, vError("vgId:%d, QInfo:0x%" PRIx64 "-%p register qhandle failed, return to app, code:%s,", pVnode->vgId, qId,
tstrerror(pRsp->code)); (void *)pQInfo, tstrerror(pRsp->code));
qDestroyQueryInfo(pQInfo); // destroy it directly qDestroyQueryInfo(pQInfo); // destroy it directly
return pRsp->code; return pRsp->code;
} else { } else {
...@@ -261,7 +135,7 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -261,7 +135,7 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
if (handle != NULL && if (handle != NULL &&
vnodeNotifyCurrentQhandle(pRead->rpcHandle, qId, *handle, pVnode->vgId) != TSDB_CODE_SUCCESS) { vnodeNotifyCurrentQhandle(pRead->rpcHandle, qId, *handle, pVnode->vgId) != TSDB_CODE_SUCCESS) {
vError("vgId:%d, QInfo:0x%"PRIx64 "-%p, query discarded since link is broken, %p", pVnode->vgId, qId, *handle, vError("vgId:%d, QInfo:0x%" PRIx64 "-%p, query discarded since link is broken, %p", pVnode->vgId, qId, *handle,
pRead->rpcHandle); pRead->rpcHandle);
pRsp->code = TSDB_CODE_RPC_NETWORK_UNAVAIL; pRsp->code = TSDB_CODE_RPC_NETWORK_UNAVAIL;
...@@ -274,8 +148,9 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -274,8 +148,9 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
} }
if (handle != NULL) { if (handle != NULL) {
vTrace("vgId:%d, QInfo:0x%"PRIx64 "-%p, dnode query msg disposed, create qhandle and returns to app", vgId, qId, *handle); vTrace("vgId:%d, QInfo:0x%" PRIx64 "-%p, query msg disposed, create qhandle and returns to app", vgId, qId,
code = vnodePutItemIntoReadQueue(pVnode, handle, pRead->rpcHandle); *handle);
code = vnodeReputPutToRQueue(pVnode, handle, pRead->rpcHandle);
if (code != TSDB_CODE_SUCCESS) { if (code != TSDB_CODE_SUCCESS) {
pRsp->code = code; pRsp->code = code;
qReleaseQInfo(pVnode->qMgmt, (void **)&handle, true); qReleaseQInfo(pVnode->qMgmt, (void **)&handle, true);
...@@ -283,14 +158,14 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -283,14 +158,14 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
} }
} }
int32_t remain = atomic_add_fetch_32(&vNumOfExistedQHandle, 1); int32_t remain = atomic_add_fetch_32(&pVnode->numOfExistQHandle, 1);
vTrace("vgId:%d, new qhandle created, total qhandle:%d", pVnode->vgId, remain); vTrace("vgId:%d, new qhandle created, total qhandle:%d", pVnode->vgId, remain);
} else { } else {
assert(pCont != NULL); assert(pCont != NULL);
void **qhandle = (void **)pRead->qhandle; void ** qhandle = (void **)pRead->qhandle;
uint64_t qId = 0; uint64_t qId = 0;
vTrace("vgId:%d, QInfo:%p, dnode continues to exec query", pVnode->vgId, *qhandle); vTrace("vgId:%d, QInfo:%p, continues to exec query", pVnode->vgId, *qhandle);
// In the retrieve blocking model, only 50% CPU will be used in query processing // In the retrieve blocking model, only 50% CPU will be used in query processing
if (tsRetrieveBlockingModel) { if (tsRetrieveBlockingModel) {
...@@ -315,10 +190,11 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -315,10 +190,11 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
// NOTE: set return code to be TSDB_CODE_QRY_HAS_RSP to notify dnode to return msg to client // NOTE: set return code to be TSDB_CODE_QRY_HAS_RSP to notify dnode to return msg to client
code = TSDB_CODE_QRY_HAS_RSP; code = TSDB_CODE_QRY_HAS_RSP;
} else { } else {
//void *h1 = qGetResultRetrieveMsg(*qhandle); // void *h1 = qGetResultRetrieveMsg(*qhandle);
/* remove this assert, one possible case that will cause h1 not NULL: query thread unlock pQInfo->lock, and then FETCH thread execute twice before query thread reach here */ /* remove this assert, one possible case that will cause h1 not NULL: query thread unlock pQInfo->lock, and then
//assert(h1 == NULL); * FETCH thread execute twice before query thread reach here */
// assert(h1 == NULL);
freehandle = qQueryCompleted(*qhandle); freehandle = qQueryCompleted(*qhandle);
} }
...@@ -327,22 +203,24 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -327,22 +203,24 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
// If the building of result is not required, simply free it. Otherwise, mandatorily free the qhandle // If the building of result is not required, simply free it. Otherwise, mandatorily free the qhandle
if (freehandle || (!buildRes)) { if (freehandle || (!buildRes)) {
if (freehandle) { if (freehandle) {
int32_t remain = atomic_sub_fetch_32(&vNumOfExistedQHandle, 1); int32_t remain = atomic_sub_fetch_32(&pVnode->numOfExistQHandle, 1);
vTrace("vgId:%d, QInfo:%p, start to free qhandle, remain qhandle:%d", pVnode->vgId, *qhandle, remain); vTrace("vgId:%d, QInfo:%p, start to free qhandle, remain qhandle:%d", pVnode->vgId, *qhandle, remain);
} }
qReleaseQInfo(pVnode->qMgmt, (void **)&qhandle, freehandle); qReleaseQInfo(pVnode->qMgmt, (void **)&qhandle, freehandle);
} }
} }
} }
return code; return code;
#endif
return 0;
} }
static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { int32_t vnodeProcessFetchMsg(SVnode *pVnode, SReadMsg *pRead) {
void *pCont = pRead->pCont; #if 0
SRspRet *pRet = &pRead->rspRet; void * pCont = pRead->pCont;
SVnRsp *pRet = &pRead->rspRet;
SRetrieveTableMsg *pRetrieve = pCont; SRetrieveTableMsg *pRetrieve = pCont;
pRetrieve->free = htons(pRetrieve->free); pRetrieve->free = htons(pRetrieve->free);
...@@ -351,7 +229,7 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -351,7 +229,7 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
vTrace("vgId:%d, qId:0x%" PRIx64 ", retrieve msg is disposed, free:%d, conn:%p", pVnode->vgId, pRetrieve->qId, vTrace("vgId:%d, qId:0x%" PRIx64 ", retrieve msg is disposed, free:%d, conn:%p", pVnode->vgId, pRetrieve->qId,
pRetrieve->free, pRead->rpcHandle); pRetrieve->free, pRead->rpcHandle);
memset(pRet, 0, sizeof(SRspRet)); memset(pRet, 0, sizeof(SVnRsp));
terrno = TSDB_CODE_SUCCESS; terrno = TSDB_CODE_SUCCESS;
int32_t code = TSDB_CODE_SUCCESS; int32_t code = TSDB_CODE_SUCCESS;
...@@ -364,16 +242,17 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -364,16 +242,17 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
} }
if (code != TSDB_CODE_SUCCESS) { if (code != TSDB_CODE_SUCCESS) {
vError("vgId:%d, invalid qId in retrieving result, code:%s, QInfo:%" PRIu64, pVnode->vgId, tstrerror(code), pRetrieve->qId); vError("vgId:%d, invalid qId in retrieving result, code:%s, QInfo:%" PRIu64, pVnode->vgId, tstrerror(code),
pRetrieve->qId);
vnodeBuildNoResultQueryRsp(pRet); vnodeBuildNoResultQueryRsp(pRet);
return code; return code;
} }
// kill current query and free corresponding resources. // kill current query and free corresponding resources.
if (pRetrieve->free == 1) { if (pRetrieve->free == 1) {
int32_t remain = atomic_sub_fetch_32(&vNumOfExistedQHandle, 1); int32_t remain = atomic_sub_fetch_32(&pVnode->numOfExistQHandle, 1);
vWarn("vgId:%d, QInfo:%"PRIx64 "-%p, retrieve msg received to kill query and free qhandle, remain qhandle:%d", pVnode->vgId, pRetrieve->qId, vWarn("vgId:%d, QInfo:%" PRIx64 "-%p, retrieve msg received to kill query and free qhandle, remain qhandle:%d",
*handle, remain); pVnode->vgId, pRetrieve->qId, *handle, remain);
qKillQuery(*handle); qKillQuery(*handle);
qReleaseQInfo(pVnode->qMgmt, (void **)&handle, true); qReleaseQInfo(pVnode->qMgmt, (void **)&handle, true);
...@@ -385,9 +264,9 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -385,9 +264,9 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
// register the qhandle to connect to quit query immediate if connection is broken // register the qhandle to connect to quit query immediate if connection is broken
if (vnodeNotifyCurrentQhandle(pRead->rpcHandle, pRetrieve->qId, *handle, pVnode->vgId) != TSDB_CODE_SUCCESS) { if (vnodeNotifyCurrentQhandle(pRead->rpcHandle, pRetrieve->qId, *handle, pVnode->vgId) != TSDB_CODE_SUCCESS) {
int32_t remain = atomic_sub_fetch_32(&vNumOfExistedQHandle, 1); int32_t remain = atomic_sub_fetch_32(&pVnode->numOfExistQHandle, 1);
vError("vgId:%d, QInfo:%"PRIu64 "-%p, retrieve discarded since link is broken, conn:%p, remain qhandle:%d", pVnode->vgId, pRetrieve->qhandle, vError("vgId:%d, QInfo:%" PRIu64 "-%p, retrieve discarded since link is broken, conn:%p, remain qhandle:%d",
*handle, pRead->rpcHandle, remain); pVnode->vgId, pRetrieve->qhandle, *handle, pRead->rpcHandle, remain);
code = TSDB_CODE_RPC_NETWORK_UNAVAIL; code = TSDB_CODE_RPC_NETWORK_UNAVAIL;
qKillQuery(*handle); qKillQuery(*handle);
...@@ -422,29 +301,13 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { ...@@ -422,29 +301,13 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
// If qhandle is not added into vread queue, the query should be completed already or paused with error. // If qhandle is not added into vread queue, the query should be completed already or paused with error.
// Here free qhandle immediately // Here free qhandle immediately
if (freeHandle) { if (freeHandle) {
int32_t remain = atomic_sub_fetch_32(&vNumOfExistedQHandle, 1); int32_t remain = atomic_sub_fetch_32(&pVnode->numOfExistQHandle, 1);
vTrace("vgId:%d, QInfo:%p, start to free qhandle, remain qhandle:%d", pVnode->vgId, *handle, remain); vTrace("vgId:%d, QInfo:%p, start to free qhandle, remain qhandle:%d", pVnode->vgId, *handle, remain);
qReleaseQInfo(pVnode->qMgmt, (void **)&handle, true); qReleaseQInfo(pVnode->qMgmt, (void **)&handle, true);
} }
return code; return code;
#endif
return 0;
} }
// notify connection(handle) that current qhandle is created, if current connection from
// client is broken, the query needs to be killed immediately.
int32_t vnodeNotifyCurrentQhandle(void *handle, uint64_t qId, void *qhandle, int32_t vgId) {
SRetrieveTableMsg *pMsg = rpcMallocCont(sizeof(SRetrieveTableMsg));
pMsg->qId = htobe64(qId);
pMsg->header.vgId = htonl(vgId);
pMsg->header.contLen = htonl(sizeof(SRetrieveTableMsg));
vTrace("QInfo:0x%"PRIx64"-%p register qhandle to connect:%p", qId, qhandle, handle);
return rpcReportProgress(handle, (char *)pMsg, sizeof(SRetrieveTableMsg));
}
void vnodeWaitReadCompleted(SVnodeObj *pVnode) {
while (pVnode->queuedRMsg > 0) {
vTrace("vgId:%d, queued rmsg num:%d", pVnode->vgId, pVnode->queuedRMsg);
taosMsleep(10);
}
}
...@@ -16,9 +16,9 @@ ...@@ -16,9 +16,9 @@
#define _DEFAULT_SOURCE #define _DEFAULT_SOURCE
#include "os.h" #include "os.h"
#include "taosmsg.h" #include "taosmsg.h"
#include "query.h" // #include "query.h"
#include "vnodeStatus.h"
#include "vnodeRead.h" #include "vnodeRead.h"
#include "vnodeStatus.h"
#include "vnodeWrite.h" #include "vnodeWrite.h"
char* vnodeStatus[] = { char* vnodeStatus[] = {
...@@ -29,30 +29,32 @@ char* vnodeStatus[] = { ...@@ -29,30 +29,32 @@ char* vnodeStatus[] = {
"reset" "reset"
}; };
bool vnodeSetInitStatus(SVnodeObj* pVnode) { bool vnodeSetInitStatus(SVnode* pVnode) {
pthread_mutex_lock(&pVnode->statusMutex); pthread_mutex_lock(&pVnode->statusMutex);
pVnode->status = TAOS_VN_STATUS_INIT; pVnode->status = TAOS_VN_STATUS_INIT;
pthread_mutex_unlock(&pVnode->statusMutex); pthread_mutex_unlock(&pVnode->statusMutex);
return true; return true;
} }
bool vnodeSetReadyStatus(SVnodeObj* pVnode) { bool vnodeSetReadyStatus(SVnode* pVnode) {
bool set = false; bool set = false;
pthread_mutex_lock(&pVnode->statusMutex); pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_READY || if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_READY ||
pVnode->status == TAOS_VN_STATUS_UPDATING || pVnode->status == TAOS_VN_STATUS_RESET) { pVnode->status == TAOS_VN_STATUS_UPDATING) {
pVnode->status = TAOS_VN_STATUS_READY; pVnode->status = TAOS_VN_STATUS_READY;
set = true; set = true;
} }
#if 0
qQueryMgmtReOpen(pVnode->qMgmt); qQueryMgmtReOpen(pVnode->qMgmt);
#endif
pthread_mutex_unlock(&pVnode->statusMutex); pthread_mutex_unlock(&pVnode->statusMutex);
return set; return set;
} }
static bool vnodeSetClosingStatusImp(SVnodeObj* pVnode) { static bool vnodeSetClosingStatusImp(SVnode* pVnode) {
bool set = false; bool set = false;
pthread_mutex_lock(&pVnode->statusMutex); pthread_mutex_lock(&pVnode->statusMutex);
...@@ -65,7 +67,7 @@ static bool vnodeSetClosingStatusImp(SVnodeObj* pVnode) { ...@@ -65,7 +67,7 @@ static bool vnodeSetClosingStatusImp(SVnodeObj* pVnode) {
return set; return set;
} }
bool vnodeSetClosingStatus(SVnodeObj* pVnode) { bool vnodeSetClosingStatus(SVnode* pVnode) {
if (pVnode->status == TAOS_VN_STATUS_CLOSING) if (pVnode->status == TAOS_VN_STATUS_CLOSING)
return true; return true;
...@@ -73,15 +75,17 @@ bool vnodeSetClosingStatus(SVnodeObj* pVnode) { ...@@ -73,15 +75,17 @@ bool vnodeSetClosingStatus(SVnodeObj* pVnode) {
taosMsleep(1); taosMsleep(1);
} }
#if 0
// release local resources only after cutting off outside connections // release local resources only after cutting off outside connections
qQueryMgmtNotifyClosed(pVnode->qMgmt); qQueryMgmtNotifyClosed(pVnode->qMgmt);
#endif
vnodeWaitReadCompleted(pVnode); vnodeWaitReadCompleted(pVnode);
vnodeWaitWriteCompleted(pVnode); vnodeWaitWriteCompleted(pVnode);
return true; return true;
} }
bool vnodeSetUpdatingStatus(SVnodeObj* pVnode) { bool vnodeSetUpdatingStatus(SVnode* pVnode) {
bool set = false; bool set = false;
pthread_mutex_lock(&pVnode->statusMutex); pthread_mutex_lock(&pVnode->statusMutex);
...@@ -94,35 +98,7 @@ bool vnodeSetUpdatingStatus(SVnodeObj* pVnode) { ...@@ -94,35 +98,7 @@ bool vnodeSetUpdatingStatus(SVnodeObj* pVnode) {
return set; return set;
} }
static bool vnodeSetResetStatusImp(SVnodeObj* pVnode) { bool vnodeInInitStatus(SVnode* pVnode) {
bool set = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_READY || pVnode->status == TAOS_VN_STATUS_INIT) {
pVnode->status = TAOS_VN_STATUS_RESET;
set = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return set;
}
bool vnodeSetResetStatus(SVnodeObj* pVnode) {
while (!vnodeSetResetStatusImp(pVnode)) {
taosMsleep(1);
}
vInfo("vgId:%d, set to reset status", pVnode->vgId);
// release local resources only after cutting off outside connections
qQueryMgmtNotifyClosed(pVnode->qMgmt);
vnodeWaitReadCompleted(pVnode);
vnodeWaitWriteCompleted(pVnode);
return true;
}
bool vnodeInInitStatus(SVnodeObj* pVnode) {
bool in = false; bool in = false;
pthread_mutex_lock(&pVnode->statusMutex); pthread_mutex_lock(&pVnode->statusMutex);
...@@ -134,7 +110,7 @@ bool vnodeInInitStatus(SVnodeObj* pVnode) { ...@@ -134,7 +110,7 @@ bool vnodeInInitStatus(SVnodeObj* pVnode) {
return in; return in;
} }
bool vnodeInReadyStatus(SVnodeObj* pVnode) { bool vnodeInReadyStatus(SVnode* pVnode) {
bool in = false; bool in = false;
pthread_mutex_lock(&pVnode->statusMutex); pthread_mutex_lock(&pVnode->statusMutex);
...@@ -146,19 +122,7 @@ bool vnodeInReadyStatus(SVnodeObj* pVnode) { ...@@ -146,19 +122,7 @@ bool vnodeInReadyStatus(SVnodeObj* pVnode) {
return in; return in;
} }
bool vnodeInReadyOrUpdatingStatus(SVnodeObj* pVnode) { bool vnodeInClosingStatus(SVnode* pVnode) {
bool in = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_READY || pVnode->status == TAOS_VN_STATUS_UPDATING) {
in = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return in;
}
bool vnodeInClosingStatus(SVnodeObj* pVnode) {
bool in = false; bool in = false;
pthread_mutex_lock(&pVnode->statusMutex); pthread_mutex_lock(&pVnode->statusMutex);
...@@ -170,14 +134,3 @@ bool vnodeInClosingStatus(SVnodeObj* pVnode) { ...@@ -170,14 +134,3 @@ bool vnodeInClosingStatus(SVnodeObj* pVnode) {
return in; return in;
} }
bool vnodeInResetStatus(SVnodeObj* pVnode) {
bool in = false;
pthread_mutex_lock(&pVnode->statusMutex);
if (pVnode->status == TAOS_VN_STATUS_RESET) {
in = true;
}
pthread_mutex_unlock(&pVnode->statusMutex);
return in;
}
...@@ -19,7 +19,7 @@ ...@@ -19,7 +19,7 @@
#include "tglobal.h" #include "tglobal.h"
#include "vnodeVersion.h" #include "vnodeVersion.h"
int32_t vnodeReadVersion(SVnodeObj *pVnode) { int32_t vnodeReadVersion(SVnode *pVnode) {
int32_t len = 0; int32_t len = 0;
int32_t maxLen = 100; int32_t maxLen = 100;
char * content = calloc(1, maxLen + 1); char * content = calloc(1, maxLen + 1);
...@@ -71,7 +71,7 @@ PARSE_VER_ERROR: ...@@ -71,7 +71,7 @@ PARSE_VER_ERROR:
return terrno; return terrno;
} }
int32_t vnodeSaveVersion(SVnodeObj *pVnode) { int32_t vnodeSaveVersion(SVnode *pVnode) {
char file[TSDB_FILENAME_LEN + 30] = {0}; char file[TSDB_FILENAME_LEN + 30] = {0};
sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId); sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId);
...@@ -90,7 +90,7 @@ int32_t vnodeSaveVersion(SVnodeObj *pVnode) { ...@@ -90,7 +90,7 @@ int32_t vnodeSaveVersion(SVnodeObj *pVnode) {
len += snprintf(content + len, maxLen - len, "}\n"); len += snprintf(content + len, maxLen - len, "}\n");
fwrite(content, 1, len, fp); fwrite(content, 1, len, fp);
taosFsync(fileno(fp)); taosFsyncFile(fileno(fp));
fclose(fp); fclose(fp);
free(content); free(content);
terrno = 0; terrno = 0;
......
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "vnodeMain.h"
#include "vnodeWorker.h"
enum { CLEANUP_TASK = 0, DESTROY_TASK = 1, BACKUP_TASK = 2 };
typedef struct {
int32_t vgId;
int32_t code;
int32_t type;
void * rpcHandle;
SVnode *pVnode;
} SVnTask;
static struct {
SWorkerPool pool;
taos_queue pQueue;
} tsVworker = {0};
static void vnodeProcessTaskStart(void *unused, SVnTask *pTask, int32_t qtype) {
pTask->code = 0;
switch (pTask->type) {
case CLEANUP_TASK:
vnodeCleanUp(pTask->pVnode);
break;
case DESTROY_TASK:
vnodeDestroy(pTask->pVnode);
break;
case BACKUP_TASK:
vnodeBackup(pTask->vgId);
break;
default:
break;
}
}
static void vnodeProcessTaskEnd(void *unused, SVnTask *pTask, int32_t qtype, int32_t code) {
if (pTask->rpcHandle != NULL) {
SRpcMsg rpcRsp = {.handle = pTask->rpcHandle, .code = pTask->code};
rpcSendResponse(&rpcRsp);
}
taosFreeQitem(pTask);
}
static int32_t vnodeWriteIntoTaskQueue(SVnode *pVnode, int32_t type, void *rpcHandle) {
SVnTask *pTask = taosAllocateQitem(sizeof(SVnTask));
if (pTask == NULL) return TSDB_CODE_VND_OUT_OF_MEMORY;
pTask->vgId = pVnode->vgId;
pTask->pVnode = pVnode;
pTask->rpcHandle = rpcHandle;
pTask->type = type;
return taosWriteQitem(tsVworker.pQueue, TAOS_QTYPE_RPC, pTask);
}
void vnodeProcessCleanupTask(SVnode *pVnode) {
vnodeWriteIntoTaskQueue(pVnode, CLEANUP_TASK, NULL);
}
void vnodeProcessDestroyTask(SVnode *pVnode) {
vnodeWriteIntoTaskQueue(pVnode, DESTROY_TASK, NULL);
}
void vnodeProcessBackupTask(SVnode *pVnode) {
vnodeWriteIntoTaskQueue(pVnode, BACKUP_TASK, NULL);
}
int32_t vnodeInitWorker() {
SWorkerPool *pPool = &tsVworker.pool;
pPool->name = "vworker";
pPool->startFp = (ProcessStartFp)vnodeProcessTaskStart;
pPool->endFp = (ProcessEndFp)vnodeProcessTaskEnd;
pPool->min = 0;
pPool->max = 1;
if (tWorkerInit(pPool) != 0) {
return TSDB_CODE_VND_OUT_OF_MEMORY;
}
tsVworker.pQueue = tWorkerAllocQueue(pPool, NULL);
vInfo("vworker is initialized, max worker %d", pPool->max);
return TSDB_CODE_SUCCESS;
}
void vnodeCleanupWorker() {
tWorkerFreeQueue(&tsVworker.pool, tsVworker.pQueue);
tWorkerCleanup(&tsVworker.pool);
tsVworker.pQueue = NULL;
vInfo("vworker is closed");
}
...@@ -13,55 +13,224 @@ ...@@ -13,55 +13,224 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>. * along with this program. If not, see <http://www.gnu.org/licenses/>.
*/ */
#include "vnodeInt.h" #define _DEFAULT_SOURCE
#include "os.h"
#include "tglobal.h"
#include "tqueue.h"
#include "tworker.h"
#include "taosmsg.h"
#include "vnodeMain.h"
#include "vnodeStatus.h"
#include "vnodeWrite.h"
#include "vnodeWriteMsg.h"
int vnodeProcessSubmitReq(SVnode *pVnode, SSubmitReq *pReq, SSubmitRsp *pRsp) { typedef int32_t (*WriteMsgFp)(SVnode *, void *pCont, SVnRsp *);
// TODO: Check inputs
#if 0 typedef struct {
void *pMem = NULL; int32_t code;
if ((pMem = amalloc(pVnode->allocator, REQ_SIZE(pReq))) == NULL) { int8_t qtype;
// No more memory to allocate, schedule an async commit SVnode * pVnode;
// and continue SRpcMsg rpcMsg;
vnodeAsyncCommit(pVnode); SVnRsp rspRet;
char reserveForSync[24];
// Reset allocator and allocat more SWalHead walHead;
vnodeResetAllocator(pVnode); } SVnWriteMsg;
pMem = amalloc(pVnode->allocator, REQ_SIZE(pReq));
if (pMem == NULL) { static struct {
// TODO: handle the error SWriteWorkerPool pool;
} int64_t queuedBytes;
int32_t queuedMsgs;
} tsVwrite = {0};
void vnodeStartWrite(SVnode *pVnode) {}
void vnodeStoprite(SVnode *pVnode) {}
void vnodeWaitWriteCompleted(SVnode *pVnode) {
while (pVnode->queuedWMsg > 0) {
vTrace("vgId:%d, queued wmsg num:%d", pVnode->vgId, pVnode->queuedWMsg);
taosMsleep(10);
}
}
static int32_t vnodeWriteToWQueue(SVnode *pVnode, SWalHead *pHead, int32_t qtype, SRpcMsg *pRpcMsg) {
if (!(pVnode->accessState & TSDB_VN_WRITE_ACCCESS)) {
vWarn("vgId:%d, no write auth", pVnode->vgId);
return TSDB_CODE_VND_NO_WRITE_AUTH;
} }
// TODO: if SSubmitReq is compressed or encoded, we need to decode the request if (tsAvailDataDirGB <= tsMinimalDataDirGB) {
memcpy(pMem, pReq, REQ_SIZE(pReq)); vWarn("vgId:%d, failed to write into vwqueue since no diskspace, avail:%fGB", pVnode->vgId, tsAvailDataDirGB);
return TSDB_CODE_VND_NO_DISKSPACE;
}
if (tqPushMsg((SSubmitReq *)pReq) < 0) { if (pHead->len > TSDB_MAX_WAL_SIZE) {
// TODO: handle error vError("vgId:%d, wal len:%d exceeds limit, hver:%" PRIu64, pVnode->vgId, pHead->len, pHead->version);
return TSDB_CODE_WAL_SIZE_LIMIT;
} }
SSubmitReqReader reader; if (!vnodeInReadyStatus(pVnode)) {
taosInitSubmitReqReader(&reader, (SSubmitReq *)pMem); vError("vgId:%d, failed to write into vwqueue, vstatus is %s", pVnode->vgId, vnodeStatus[pVnode->status]);
return TSDB_CODE_APP_NOT_READY;
}
if (tsdbInsert(pVnode->pTsdb, (SSubmitReq *)pMem) < 0) { if (tsVwrite.queuedBytes > tsMaxVnodeQueuedBytes) {
// TODO: handler error vDebug("vgId:%d, too many bytes:%" PRId64 " in vwqueue, flow control", pVnode->vgId, tsVwrite.queuedBytes);
return TSDB_CODE_VND_IS_FLOWCTRL;
} }
#endif
return 0; int32_t size = sizeof(SVnWriteMsg) + pHead->len;
SVnWriteMsg *pWrite = taosAllocateQitem(size);
if (pWrite == NULL) {
return TSDB_CODE_VND_OUT_OF_MEMORY;
}
if (pRpcMsg != NULL) {
pWrite->rpcMsg = *pRpcMsg;
}
memcpy(&pWrite->walHead, pHead, sizeof(SWalHead) + pHead->len);
pWrite->pVnode = pVnode;
pWrite->qtype = qtype;
atomic_add_fetch_64(&tsVwrite.queuedBytes, size);
atomic_add_fetch_32(&tsVwrite.queuedMsgs, 1);
atomic_add_fetch_32(&pVnode->refCount, 1);
atomic_add_fetch_32(&pVnode->queuedWMsg, 1);
taosWriteQitem(pVnode->wqueue, pWrite->qtype, pWrite);
return TSDB_CODE_SUCCESS;
}
static void vnodeFreeFromWQueue(SVnode *pVnode, SVnWriteMsg *pWrite) {
int64_t size = sizeof(SVnWriteMsg) + pWrite->walHead.len;
atomic_sub_fetch_64(&tsVwrite.queuedBytes, size);
atomic_sub_fetch_32(&tsVwrite.queuedMsgs, 1);
atomic_sub_fetch_32(&pVnode->queuedWMsg, 1);
taosFreeQitem(pWrite);
vnodeRelease(pVnode);
}
int32_t vnodeProcessWalMsg(SVnode *pVnode, SWalHead *pHead) {
return vnodeWriteToWQueue(pVnode, pHead, TAOS_QTYPE_WAL, NULL);
}
void vnodeProcessWriteMsg(SRpcMsg *pRpcMsg) {
int32_t code;
SMsgHead *pMsg = pRpcMsg->pCont;
pMsg->vgId = htonl(pMsg->vgId);
pMsg->contLen = htonl(pMsg->contLen);
SVnode *pVnode = vnodeAcquireNotClose(pMsg->vgId);
if (pVnode == NULL) {
code = TSDB_CODE_VND_INVALID_VGROUP_ID;
} else {
SWalHead *pHead = (SWalHead *)((char *)pRpcMsg->pCont - sizeof(SWalHead));
pHead->msgType = pRpcMsg->msgType;
pHead->version = 0;
pHead->len = pMsg->contLen;
code = vnodeWriteToWQueue(pVnode, pHead, TAOS_QTYPE_RPC, pRpcMsg);
}
if (code != TSDB_CODE_SUCCESS) {
SRpcMsg rpcRsp = {.handle = pRpcMsg->handle, .code = code};
rpcSendResponse(&rpcRsp);
}
vnodeRelease(pVnode);
rpcFreeCont(pRpcMsg->pCont);
}
static bool vnodeProcessWriteStart(SVnode *pVnode, SVnWriteMsg *pWrite, int32_t qtype) {
SWalHead *pHead = &pWrite->walHead;
SVnRsp * pRet = &pWrite->rspRet;
int32_t msgType = pHead->msgType;
vTrace("vgId:%d, msg:%s will be processed, hver:%" PRIu64, pVnode->vgId, taosMsg[pHead->msgType], pHead->version);
// write into WAL
#if 0
pWrite->code = walWrite(pVnode->wal, pHead);
if (pWrite->code < 0) return false;
#endif
pVnode->version = pHead->version;
// write data locally
switch (msgType) {
case TSDB_MSG_TYPE_SUBMIT:
pRet->len = sizeof(SSubmitRsp);
pRet->rsp = rpcMallocCont(pRet->len);
pWrite->code = vnodeProcessSubmitReq(pVnode, (void*)pHead->cont, pRet->rsp);
break;
case TSDB_MSG_TYPE_MD_CREATE_TABLE:
pWrite->code = vnodeProcessCreateTableReq(pVnode, (void*)pHead->cont, NULL);
break;
case TSDB_MSG_TYPE_MD_DROP_TABLE:
pWrite->code = vnodeProcessDropTableReq(pVnode, (void*)pHead->cont, NULL);
break;
case TSDB_MSG_TYPE_MD_ALTER_TABLE:
pWrite->code = vnodeProcessAlterTableReq(pVnode, (void*)pHead->cont, NULL);
break;
case TSDB_MSG_TYPE_MD_DROP_STABLE:
pWrite->code = vnodeProcessDropStableReq(pVnode, (void*)pHead->cont, NULL);
break;
case TSDB_MSG_TYPE_UPDATE_TAG_VAL:
pWrite->code = vnodeProcessUpdateTagValReq(pVnode, (void*)pHead->cont, NULL);
break;
default:
pWrite->code = TSDB_CODE_VND_MSG_NOT_PROCESSED;
break;
}
if (pWrite->code < 0) return false;
// update fync
return (pWrite->code == 0 && msgType != TSDB_MSG_TYPE_SUBMIT);
}
static void vnodeFsync(SVnode *pVnode, bool fsync) {
#if 0
walFsync(pVnode->wal, fsync);
#endif
} }
int vnodeProcessCreateTableReq(SVnode *pVnode, SCreateTableReq *pReq, SCreateTableRsp *pRsp) { static void vnodeProcessWriteEnd(SVnode *pVnode, SVnWriteMsg *pWrite, int32_t qtype, int32_t code) {
// TODO if (qtype == TAOS_QTYPE_RPC) {
return 0; SRpcMsg rpcRsp = {
.handle = pWrite->rpcMsg.handle,
.pCont = pWrite->rspRet.rsp,
.contLen = pWrite->rspRet.len,
.code = pWrite->code,
};
rpcSendResponse(&rpcRsp);
} else {
if (pWrite->rspRet.rsp) {
rpcFreeCont(pWrite->rspRet.rsp);
}
}
vnodeFreeFromWQueue(pVnode, pWrite);
} }
int vnodeProcessDropTableReq(SVnode *pVnode, SDropTableReq *pReq, SDropTableRsp *pRsp) { int32_t vnodeInitWrite() {
// TODO SWriteWorkerPool *pPool = &tsVwrite.pool;
return 0; pPool->name = "vwrite";
pPool->max = tsNumOfCores;
pPool->startFp = (ProcessWriteStartFp)vnodeProcessWriteStart;
pPool->syncFp = (ProcessWriteSyncFp)vnodeFsync;
pPool->endFp = (ProcessWriteEndFp)vnodeProcessWriteEnd;
if (tWriteWorkerInit(pPool) != 0) return -1;
vInfo("vwrite is initialized, max worker %d", pPool->max);
return TSDB_CODE_SUCCESS;
} }
int vnodeProcessAlterTableReq(SVnode *pVnode, SAlterTableReq *pReq, SAlterTableRsp *pRsp) { void vnodeCleanupWrite() {
// TODO tWriteWorkerCleanup(&tsVwrite.pool);
return 0; vInfo("vwrite is closed");
} }
taos_queue vnodeAllocWriteQueue(SVnode *pVnode) { return tWriteWorkerAllocQueue(&tsVwrite.pool, pVnode); }
void vnodeFreeWriteQueue(taos_queue pQueue) { tWriteWorkerFreeQueue(&tsVwrite.pool, pQueue); }
\ No newline at end of file
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "vnodeWriteMsg.h"
int32_t vnodeProcessSubmitReq(SVnode *pVnode, SSubmitReq *pReq, SSubmitRsp *pRsp) {
// TODO: Check inputs
#if 0
void *pMem = NULL;
if ((pMem = amalloc(pVnode->allocator, REQ_SIZE(pReq))) == NULL) {
// No more memory to allocate, schedule an async commit
// and continue
vnodeAsyncCommit(pVnode);
// Reset allocator and allocat more
vnodeResetAllocator(pVnode);
pMem = amalloc(pVnode->allocator, REQ_SIZE(pReq));
if (pMem == NULL) {
// TODO: handle the error
}
}
// TODO: if SSubmitReq is compressed or encoded, we need to decode the request
memcpy(pMem, pReq, REQ_SIZE(pReq));
if (tqPushMsg((SSubmitReq *)pReq) < 0) {
// TODO: handle error
}
SSubmitReqReader reader;
taosInitSubmitReqReader(&reader, (SSubmitReq *)pMem);
if (tsdbInsert(pVnode->pTsdb, (SSubmitReq *)pMem) < 0) {
// TODO: handler error
}
#endif
return 0;
}
int32_t vnodeProcessCreateTableReq(SVnode *pVnode, SCreateTableReq *pReq, SCreateTableRsp *pRsp) {
// TODO
return 0;
}
int32_t vnodeProcessDropTableReq(SVnode *pVnode, SDropTableReq *pReq, SDropTableRsp *pRsp) {
// TODO
return 0;
}
int32_t vnodeProcessAlterTableReq(SVnode *pVnode, SAlterTableReq *pReq, SAlterTableRsp *pRsp) {
// TODO
return 0;
}
int32_t vnodeProcessDropStableReq(SVnode *pVnode, SDropStableReq *pReq, SDropStableRsp *pRsp) {
// TODO
return 0;
}
int32_t vnodeProcessUpdateTagValReq(SVnode *pVnode, SUpdateTagValReq *pReq, SUpdateTagValRsp *pRsp) {
// TODO
return 0;
}
...@@ -20,15 +20,14 @@ ...@@ -20,15 +20,14 @@
typedef struct SStepObj { typedef struct SStepObj {
char * name; char * name;
void ** self;
InitFp initFp; InitFp initFp;
CleanupFp cleanupFp; CleanupFp cleanupFp;
} SStep; } SStep;
typedef struct SSteps { typedef struct SSteps {
int32_t cursize; int32_t cursize;
int32_t maxsize; int32_t maxsize;
SStep * steps; SStep * steps;
ReportFp reportFp; ReportFp reportFp;
} SSteps; } SSteps;
...@@ -44,14 +43,14 @@ SSteps *taosStepInit(int32_t maxsize, ReportFp fp) { ...@@ -44,14 +43,14 @@ SSteps *taosStepInit(int32_t maxsize, ReportFp fp) {
return steps; return steps;
} }
int32_t taosStepAdd(struct SSteps *steps, char *name, void **obj, InitFp initFp, CleanupFp cleanupFp) { int32_t taosStepAdd(struct SSteps *steps, char *name, InitFp initFp, CleanupFp cleanupFp) {
if (steps == NULL) return -1; if (steps == NULL) return -1;
if (steps->cursize >= steps->maxsize) { if (steps->cursize >= steps->maxsize) {
uError("failed to add step since up to the maxsize"); uError("failed to add step since up to the maxsize");
return -1; return -1;
} }
SStep step = {.name = name, .self = obj, .initFp = initFp, .cleanupFp = cleanupFp}; SStep step = {.name = name, .initFp = initFp, .cleanupFp = cleanupFp};
steps->steps[steps->cursize++] = step; steps->steps[steps->cursize++] = step;
return 0; return 0;
} }
...@@ -61,7 +60,7 @@ static void taosStepCleanupImp(SSteps *steps, int32_t pos) { ...@@ -61,7 +60,7 @@ static void taosStepCleanupImp(SSteps *steps, int32_t pos) {
SStep *step = steps->steps + s; SStep *step = steps->steps + s;
uDebug("step:%s will cleanup", step->name); uDebug("step:%s will cleanup", step->name);
if (step->cleanupFp != NULL) { if (step->cleanupFp != NULL) {
(*step->cleanupFp)(step->self); (*step->cleanupFp)();
} }
} }
} }
...@@ -77,7 +76,7 @@ int32_t taosStepExec(SSteps *steps) { ...@@ -77,7 +76,7 @@ int32_t taosStepExec(SSteps *steps) {
(*steps->reportFp)(step->name, "start initialize"); (*steps->reportFp)(step->name, "start initialize");
} }
int32_t code = (*step->initFp)(step->self); int32_t code = (*step->initFp)();
if (code != 0) { if (code != 0) {
uDebug("step:%s will cleanup", step->name); uDebug("step:%s will cleanup", step->name);
taosStepCleanupImp(steps, s); taosStepCleanupImp(steps, s);
......
...@@ -19,37 +19,16 @@ ...@@ -19,37 +19,16 @@
#include "tqueue.h" #include "tqueue.h"
#include "tworker.h" #include "tworker.h"
static void *taosWorkerThreadFp(void *wparam) { typedef void* (*ThreadFp)(void *param);
SWorker * worker = wparam;
SWorkerPool *pool = worker->pool;
void * msg = NULL;
int32_t qtype = 0;
void * ahandle = NULL;
int32_t code = 0;
setThreadName(pool->name);
while (1) {
if (taosReadQitemFromQset(pool->qset, &qtype, (void **)&msg, &ahandle) == 0) {
uDebug("pool:%s, worker:%d qset:%p, got no message and exiting", pool->name, worker->id, pool->qset);
break;
}
code = (*pool->reqFp)(ahandle, msg);
(*pool->rspFp)(ahandle, msg, qtype, code);
}
return NULL;
}
int32_t tWorkerInit(SWorkerPool *pool) { int32_t tWorkerInit(SWorkerPool *pool) {
pool->qset = taosOpenQset(); pool->qset = taosOpenQset();
pool->workers = calloc(sizeof(SWorker), pool->max); pool->workers = calloc(sizeof(SWorker), pool->max);
pthread_mutex_init(&pool->mutex, NULL); pthread_mutex_init(&pool->mutex, NULL);
for (int i = 0; i < pool->max; ++i) { for (int i = 0; i < pool->max; ++i) {
SWorker *pWorker = pool->workers + i; SWorker *worker = pool->workers + i;
pWorker->id = i; worker->id = i;
pWorker->pool = pool; worker->pool = pool;
} }
uInfo("worker:%s is initialized, min:%d max:%d", pool->name, pool->min, pool->max); uInfo("worker:%s is initialized, min:%d max:%d", pool->name, pool->min, pool->max);
...@@ -58,16 +37,16 @@ int32_t tWorkerInit(SWorkerPool *pool) { ...@@ -58,16 +37,16 @@ int32_t tWorkerInit(SWorkerPool *pool) {
void tWorkerCleanup(SWorkerPool *pool) { void tWorkerCleanup(SWorkerPool *pool) {
for (int i = 0; i < pool->max; ++i) { for (int i = 0; i < pool->max; ++i) {
SWorker *pWorker = pool->workers + i; SWorker *worker = pool->workers + i;
if(taosCheckPthreadValid(pWorker->thread)) { if (taosCheckPthreadValid(worker->thread)) {
taosQsetThreadResume(pool->qset); taosQsetThreadResume(pool->qset);
} }
} }
for (int i = 0; i < pool->max; ++i) { for (int i = 0; i < pool->max; ++i) {
SWorker *pWorker = pool->workers + i; SWorker *worker = pool->workers + i;
if (taosCheckPthreadValid(pWorker->thread)) { if (taosCheckPthreadValid(worker->thread)) {
pthread_join(pWorker->thread, NULL); pthread_join(worker->thread, NULL);
} }
} }
...@@ -78,42 +57,204 @@ void tWorkerCleanup(SWorkerPool *pool) { ...@@ -78,42 +57,204 @@ void tWorkerCleanup(SWorkerPool *pool) {
uInfo("worker:%s is closed", pool->name); uInfo("worker:%s is closed", pool->name);
} }
void *tWorkerAllocQueue(SWorkerPool *pool, void *ahandle) { static void *tWorkerThreadFp(SWorker *worker) {
SWorkerPool *pool = worker->pool;
void * msg = NULL;
void * ahandle = NULL;
int32_t qtype = 0;
int32_t code = 0;
taosBlockSIGPIPE();
setThreadName(pool->name);
uDebug("worker:%s:%d is running", pool->name, worker->id);
while (1) {
if (taosReadQitemFromQset(pool->qset, &qtype, (void **)&msg, &ahandle) == 0) {
uDebug("worker:%s:%d qset:%p, got no message and exiting", pool->name, worker->id, pool->qset);
break;
}
code = (*pool->startFp)(ahandle, msg, qtype);
(*pool->endFp)(ahandle, msg, qtype, code);
}
return NULL;
}
taos_queue tWorkerAllocQueue(SWorkerPool *pool, void *ahandle) {
pthread_mutex_lock(&pool->mutex); pthread_mutex_lock(&pool->mutex);
taos_queue pQueue = taosOpenQueue(); taos_queue queue = taosOpenQueue();
if (pQueue == NULL) { if (queue == NULL) {
pthread_mutex_unlock(&pool->mutex); pthread_mutex_unlock(&pool->mutex);
return NULL; return NULL;
} }
taosAddIntoQset(pool->qset, pQueue, ahandle); taosAddIntoQset(pool->qset, queue, ahandle);
// spawn a thread to process queue // spawn a thread to process queue
if (pool->num < pool->max) { if (pool->num < pool->max) {
do { do {
SWorker *pWorker = pool->workers + pool->num; SWorker *worker = pool->workers + pool->num;
pthread_attr_t thAttr; pthread_attr_t thAttr;
pthread_attr_init(&thAttr); pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&pWorker->thread, &thAttr, taosWorkerThreadFp, pWorker) != 0) { if (pthread_create(&worker->thread, &thAttr, (ThreadFp)tWorkerThreadFp, worker) != 0) {
uError("workers:%s:%d failed to create thread to process since %s", pool->name, pWorker->id, strerror(errno)); uError("worker:%s:%d failed to create thread to process since %s", pool->name, worker->id, strerror(errno));
} }
pthread_attr_destroy(&thAttr); pthread_attr_destroy(&thAttr);
pool->num++; pool->num++;
uDebug("workers:%s:%d is launched, total:%d", pool->name, pWorker->id, pool->num); uDebug("worker:%s:%d is launched, total:%d", pool->name, worker->id, pool->num);
} while (pool->num < pool->min); } while (pool->num < pool->min);
} }
pthread_mutex_unlock(&pool->mutex); pthread_mutex_unlock(&pool->mutex);
uDebug("workers:%s, queue:%p is allocated, ahandle:%p", pool->name, pQueue, ahandle); uDebug("worker:%s, queue:%p is allocated, ahandle:%p", pool->name, queue, ahandle);
return queue;
}
void tWorkerFreeQueue(SWorkerPool *pool, void *queue) {
taosCloseQueue(queue);
uDebug("worker:%s, queue:%p is freed", pool->name, queue);
}
int32_t tWriteWorkerInit(SWriteWorkerPool *pool) {
pool->nextId = 0;
pool->workers = calloc(sizeof(SWriteWorker), pool->max);
if (pool->workers == NULL) return -1;
pthread_mutex_init(&pool->mutex, NULL);
for (int32_t i = 0; i < pool->max; ++i) {
SWriteWorker *worker = pool->workers + i;
worker->id = i;
worker->qall = NULL;
worker->qset = NULL;
worker->pool = pool;
}
uInfo("worker:%s is initialized, max:%d", pool->name, pool->max);
return 0;
}
void tWriteWorkerCleanup(SWriteWorkerPool *pool) {
for (int32_t i = 0; i < pool->max; ++i) {
SWriteWorker *worker = pool->workers + i;
if (taosCheckPthreadValid(worker->thread)) {
if (worker->qset) taosQsetThreadResume(worker->qset);
}
}
for (int32_t i = 0; i < pool->max; ++i) {
SWriteWorker *worker = pool->workers + i;
if (taosCheckPthreadValid(worker->thread)) {
pthread_join(worker->thread, NULL);
taosFreeQall(worker->qall);
taosCloseQset(worker->qset);
}
}
free(pool->workers);
pthread_mutex_destroy(&pool->mutex);
uInfo("worker:%s is closed", pool->name);
}
static void *tWriteWorkerThreadFp(SWriteWorker *worker) {
SWriteWorkerPool *pool = worker->pool;
void * msg = NULL;
void * ahandle = NULL;
int32_t numOfMsgs = 0;
int32_t qtype = 0;
taosBlockSIGPIPE();
setThreadName(pool->name);
uDebug("worker:%s:%d is running", pool->name, worker->id);
while (1) {
numOfMsgs = taosReadAllQitemsFromQset(worker->qset, worker->qall, &ahandle);
if (numOfMsgs == 0) {
uDebug("worker:%s:%d qset:%p, got no message and exiting", pool->name, worker->id, worker->qset);
break;
}
bool fsync = false;
for (int32_t i = 0; i < numOfMsgs; ++i) {
taosGetQitem(worker->qall, &qtype, (void **)&msg);
fsync = fsync | (*pool->startFp)(ahandle, msg, qtype);
}
(*pool->syncFp)(ahandle, fsync);
// browse all items, and process them one by one
taosResetQitems(worker->qall);
for (int32_t i = 0; i < numOfMsgs; ++i) {
taosGetQitem(worker->qall, &qtype, (void **)&msg);
(*pool->endFp)(ahandle, msg, qtype);
}
}
return NULL;
}
taos_queue tWriteWorkerAllocQueue(SWriteWorkerPool *pool, void *ahandle) {
pthread_mutex_lock(&pool->mutex);
SWriteWorker *worker = pool->workers + pool->nextId;
taos_queue *queue = taosOpenQueue();
if (queue == NULL) {
pthread_mutex_unlock(&pool->mutex);
return NULL;
}
if (worker->qset == NULL) {
worker->qset = taosOpenQset();
if (worker->qset == NULL) {
taosCloseQueue(queue);
pthread_mutex_unlock(&pool->mutex);
return NULL;
}
taosAddIntoQset(worker->qset, queue, ahandle);
worker->qall = taosAllocateQall();
if (worker->qall == NULL) {
taosCloseQset(worker->qset);
taosCloseQueue(queue);
pthread_mutex_unlock(&pool->mutex);
return NULL;
}
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&worker->thread, &thAttr, (ThreadFp)tWriteWorkerThreadFp, worker) != 0) {
uError("worker:%s:%d failed to create thread to process since %s", pool->name, worker->id, strerror(errno));
taosFreeQall(worker->qall);
taosCloseQset(worker->qset);
taosCloseQueue(queue);
queue = NULL;
} else {
uDebug("worker:%s:%d is launched, max:%d", pool->name, worker->id, pool->max);
pool->nextId = (pool->nextId + 1) % pool->max;
}
pthread_attr_destroy(&thAttr);
} else {
taosAddIntoQset(worker->qset, queue, ahandle);
pool->nextId = (pool->nextId + 1) % pool->max;
}
pthread_mutex_unlock(&pool->mutex);
uDebug("worker:%s, queue:%p is allocated, ahandle:%p", pool->name, queue, ahandle);
return pQueue; return queue;
} }
void tWorkerFreeQueue(SWorkerPool *pool, void *pQueue) { void tWriteWorkerFreeQueue(SWriteWorkerPool *pool, taos_queue queue) {
taosCloseQueue(pQueue); taosCloseQueue(queue);
uDebug("workers:%s, queue:%p is freed", pool->name, pQueue); uDebug("worker:%s, queue:%p is freed", pool->name, queue);
} }
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef _TD_WAL_H_
#define _TD_WAL_H_
#ifdef __cplusplus
extern "C" {
#endif
typedef enum {
TAOS_WAL_NOLOG = 0,
TAOS_WAL_WRITE = 1,
TAOS_WAL_FSYNC = 2
} EWalType;
typedef enum {
TAOS_WAL_NOT_KEEP = 0,
TAOS_WAL_KEEP = 1
} EWalKeep;
typedef struct {
int8_t msgType;
int8_t sver; // sver 2 for WAL SDataRow/SMemRow compatibility
int8_t reserved[2];
int32_t len;
uint64_t version;
uint32_t signature;
uint32_t cksum;
char cont[];
} SWalHead;
typedef struct {
int32_t vgId;
int32_t fsyncPeriod; // millisecond
EWalType walLevel; // wal level
EWalKeep keep; // keep the wal file when closed
} SWalCfg;
typedef void * twalh; // WAL HANDLE
typedef int32_t FWalWrite(void *ahandle, void *pHead, int32_t qtype, void *pMsg);
int32_t walInit();
void walCleanUp();
twalh walOpen(char *path, SWalCfg *pCfg);
int32_t walAlter(twalh pWal, SWalCfg *pCfg);
void walStop(twalh);
void walClose(twalh);
int32_t walRenew(twalh);
void walRemoveOneOldFile(twalh);
void walRemoveAllOldFiles(twalh);
int32_t walWrite(twalh, SWalHead *);
void walFsync(twalh, bool forceFsync);
int32_t walRestore(twalh, void *pVnode, FWalWrite writeFp);
int32_t walGetWalFile(twalh, char *fileName, int64_t *fileId);
uint64_t walGetVersion(twalh);
void walResetVersion(twalh, uint64_t newVer);
#ifdef __cplusplus
}
#endif
#endif // _TD_WAL_H_
aux_source_directory(source RAFT_SRC)
add_library(raft ${RAFT_SRC})
target_include_directories(
raft
PUBLIC "${CMAKE_SOURCE_DIR}/include/raft"
PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/include"
)
\ No newline at end of file
CMAKE_MINIMUM_REQUIRED(VERSION 2.8...3.20)
PROJECT(TDengine)
INCLUDE_DIRECTORIES(${TD_COMMUNITY_DIR}/deps/cJson/inc)
INCLUDE_DIRECTORIES(${TD_COMMUNITY_DIR}/src/query/inc)
INCLUDE_DIRECTORIES(${TD_COMMUNITY_DIR}/src/tsdb/inc)
INCLUDE_DIRECTORIES(${TD_COMMUNITY_DIR}/src/dnode/inc)
INCLUDE_DIRECTORIES(${TD_COMMUNITY_DIR}/src/sync/inc)
INCLUDE_DIRECTORIES(${TD_ENTERPRISE_DIR}/src/inc)
INCLUDE_DIRECTORIES(inc)
AUX_SOURCE_DIRECTORY(src SRC)
ADD_LIBRARY(vnode ${SRC})
TARGET_LINK_LIBRARIES(vnode tsdb tcq common)
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_VNODE_INT_H
#define TDENGINE_VNODE_INT_H
#ifdef __cplusplus
extern "C" {
#endif
#include "tlog.h"
#include "tsync.h"
#include "tcq.h"
#include "tsdb.h"
#include "vnode.h"
extern int32_t vDebugFlag;
extern int32_t vNumOfExistedQHandle; // current initialized and existed query handle in current dnode
#define vFatal(...) { if (vDebugFlag & DEBUG_FATAL) { taosPrintLog("VND FATAL ", 255, __VA_ARGS__); }}
#define vError(...) { if (vDebugFlag & DEBUG_ERROR) { taosPrintLog("VND ERROR ", 255, __VA_ARGS__); }}
#define vWarn(...) { if (vDebugFlag & DEBUG_WARN) { taosPrintLog("VND WARN ", 255, __VA_ARGS__); }}
#define vInfo(...) { if (vDebugFlag & DEBUG_INFO) { taosPrintLog("VND ", 255, __VA_ARGS__); }}
#define vDebug(...) { if (vDebugFlag & DEBUG_DEBUG) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
#define vTrace(...) { if (vDebugFlag & DEBUG_TRACE) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
typedef struct {
int32_t vgId; // global vnode group ID
int32_t refCount; // reference count
int64_t queuedWMsgSize;
int32_t queuedWMsg;
int32_t queuedRMsg;
int32_t flowctrlLevel;
int8_t preClose; // drop and close switch
int8_t reserved[3];
int64_t sequence; // for topic
int8_t status;
int8_t role;
int8_t accessState;
int8_t isFull;
int8_t isCommiting;
int8_t dbReplica;
int8_t dropped;
int8_t dbType;
uint64_t version; // current version
uint64_t cversion; // version while commit start
uint64_t fversion; // version on saved data file
void * wqueue; // write queue
void * qqueue; // read query queue
void * fqueue; // read fetch/cancel queue
void * wal;
void * tsdb;
int64_t sync;
void * events;
void * cq; // continuous query
int32_t dbCfgVersion;
int32_t vgCfgVersion;
STsdbCfg tsdbCfg;
SSyncCfg syncCfg;
SWalCfg walCfg;
void * qMgmt;
char * rootDir;
tsem_t sem;
char db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN];
pthread_mutex_t statusMutex;
} SVnodeObj;
#ifdef __cplusplus
}
#endif
#endif
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TDENGINE_VNODE_SYNC_H
#define TDENGINE_VNODE_SYNC_H
#ifdef __cplusplus
extern "C" {
#endif
#include "vnodeInt.h"
uint32_t vnodeGetFileInfo(int32_t vgId, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fver);
int32_t vnodeGetWalInfo(int32_t vgId, char *fileName, int64_t *fileId);
void vnodeNotifyRole(int32_t vgId, int8_t role);
void vnodeCtrlFlow(int32_t vgId, int32_t level);
void vnodeStartSyncFile(int32_t vgId);
void vnodeStopSyncFile(int32_t vgId, uint64_t fversion);
void vnodeConfirmForard(int32_t vgId, void *wparam, int32_t code);
int32_t vnodeWriteToCache(int32_t vgId, void *wparam, int32_t qtype, void *rparam);
int32_t vnodeGetVersion(int32_t vgId, uint64_t *fver, uint64_t *wver);
void vnodeConfirmForward(void *pVnode, uint64_t version, int32_t code, bool force);
#ifdef __cplusplus
}
#endif
#endif
\ No newline at end of file
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "tutil.h"
#include "tqueue.h"
#include "tglobal.h"
#include "tfs.h"
#include "vnodeBackup.h"
#include "vnodeMain.h"
typedef struct {
int32_t vgId;
} SVBackupMsg;
typedef struct {
pthread_t thread;
int32_t workerId;
} SVBackupWorker;
typedef struct {
int32_t num;
SVBackupWorker *worker;
} SVBackupWorkerPool;
static SVBackupWorkerPool tsVBackupPool;
static taos_qset tsVBackupQset;
static taos_queue tsVBackupQueue;
static void vnodeProcessBackupMsg(SVBackupMsg *pMsg) {
int32_t vgId = pMsg->vgId;
char newDir[TSDB_FILENAME_LEN] = {0};
char stagingDir[TSDB_FILENAME_LEN] = {0};
sprintf(newDir, "%s/vnode%d", "vnode_bak", vgId);
sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId);
if (tsEnableVnodeBak) {
tfsRmdir(newDir);
tfsRename(stagingDir, newDir);
} else {
vInfo("vgId:%d, vnode backup not enabled", vgId);
tfsRmdir(stagingDir);
}
}
static void *vnodeBackupFunc(void *param) {
setThreadName("vnodeBackup");
while (1) {
SVBackupMsg *pMsg = NULL;
if (taosReadQitemFromQset(tsVBackupQset, NULL, (void **)&pMsg, NULL) == 0) {
vDebug("qset:%p, vbackup got no message from qset, exiting", tsVBackupQset);
break;
}
vTrace("vgId:%d, will be processed in vbackup queue", pMsg->vgId);
vnodeProcessBackupMsg(pMsg);
vTrace("vgId:%d, disposed in vbackup worker", pMsg->vgId);
taosFreeQitem(pMsg);
}
return NULL;
}
static int32_t vnodeStartBackup() {
tsVBackupQueue = taosOpenQueue();
if (tsVBackupQueue == NULL) return TSDB_CODE_DND_OUT_OF_MEMORY;
taosAddIntoQset(tsVBackupQset, tsVBackupQueue, NULL);
for (int32_t i = 0; i < tsVBackupPool.num; ++i) {
SVBackupWorker *pWorker = tsVBackupPool.worker + i;
pWorker->workerId = i;
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&pWorker->thread, &thAttr, vnodeBackupFunc, pWorker) != 0) {
vError("failed to create thread to process vbackup queue, reason:%s", strerror(errno));
}
pthread_attr_destroy(&thAttr);
vDebug("vbackup:%d is launched, total:%d", pWorker->workerId, tsVBackupPool.num);
}
vDebug("vbackup queue:%p is allocated", tsVBackupQueue);
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeWriteIntoBackupWorker(int32_t vgId) {
SVBackupMsg *pMsg = taosAllocateQitem(sizeof(SVBackupMsg));
if (pMsg == NULL) return TSDB_CODE_VND_OUT_OF_MEMORY;
pMsg->vgId = vgId;
int32_t code = taosWriteQitem(tsVBackupQueue, TAOS_QTYPE_RPC, pMsg);
if (code == 0) code = TSDB_CODE_DND_ACTION_IN_PROGRESS;
return code;
}
int32_t vnodeBackup(int32_t vgId) {
vTrace("vgId:%d, will backup", vgId);
return vnodeWriteIntoBackupWorker(vgId);
}
int32_t vnodeInitBackup() {
tsVBackupQset = taosOpenQset();
tsVBackupPool.num = 1;
tsVBackupPool.worker = calloc(sizeof(SVBackupWorker), tsVBackupPool.num);
if (tsVBackupPool.worker == NULL) return -1;
for (int32_t i = 0; i < tsVBackupPool.num; ++i) {
SVBackupWorker *pWorker = tsVBackupPool.worker + i;
pWorker->workerId = i;
vDebug("vbackup:%d is created", i);
}
vDebug("vbackup is initialized, num:%d qset:%p", tsVBackupPool.num, tsVBackupQset);
return vnodeStartBackup();
}
void vnodeCleanupBackup() {
for (int32_t i = 0; i < tsVBackupPool.num; ++i) {
SVBackupWorker *pWorker = tsVBackupPool.worker + i;
if (taosCheckPthreadValid(pWorker->thread)) {
taosQsetThreadResume(tsVBackupQset);
}
vDebug("vbackup:%d is closed", i);
}
for (int32_t i = 0; i < tsVBackupPool.num; ++i) {
SVBackupWorker *pWorker = tsVBackupPool.worker + i;
vDebug("vbackup:%d start to join", i);
if (taosCheckPthreadValid(pWorker->thread)) {
pthread_join(pWorker->thread, NULL);
}
vDebug("vbackup:%d join success", i);
}
vDebug("vbackup is closed, qset:%p", tsVBackupQset);
taosCloseQset(tsVBackupQset);
tsVBackupQset = NULL;
tfree(tsVBackupPool.worker);
vDebug("vbackup queue:%p is freed", tsVBackupQueue);
taosCloseQueue(tsVBackupQueue);
tsVBackupQueue = NULL;
}
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "dnode.h"
#include "vnodeStatus.h"
#include "vnodeBackup.h"
#include "vnodeWorker.h"
#include "vnodeRead.h"
#include "vnodeWrite.h"
#include "vnodeMain.h"
static SHashObj *tsVnodesHash = NULL;
static int32_t vnodeInitHash(void);
static void vnodeCleanupHash(void);
static void vnodeIncRef(void *ptNode);
static SStep tsVnodeSteps[] = {
{"vnode-backup", vnodeInitBackup, vnodeCleanupBackup},
{"vnode-worker", vnodeInitMWorker, vnodeCleanupMWorker},
{"vnode-write", vnodeInitWrite, vnodeCleanupWrite},
{"vnode-read", vnodeInitRead, vnodeCleanupRead},
{"vnode-hash", vnodeInitHash, vnodeCleanupHash},
{"tsdb-queue", tsdbInitCommitQueue, tsdbDestroyCommitQueue}
};
int32_t vnodeInitMgmt() {
int32_t stepSize = sizeof(tsVnodeSteps) / sizeof(SStep);
return dnodeStepInit(tsVnodeSteps, stepSize);
}
void vnodeCleanupMgmt() {
int32_t stepSize = sizeof(tsVnodeSteps) / sizeof(SStep);
dnodeStepCleanup(tsVnodeSteps, stepSize);
}
static int32_t vnodeInitHash() {
tsVnodesHash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK);
if (tsVnodesHash == NULL) {
vError("failed to init vnode mgmt");
return -1;
}
return 0;
}
static void vnodeCleanupHash() {
if (tsVnodesHash != NULL) {
vDebug("vnode mgmt is cleanup");
taosHashCleanup(tsVnodesHash);
tsVnodesHash = NULL;
}
}
void *vnodeGetWal(void *pVnode) {
return ((SVnodeObj *)pVnode)->wal;
}
void vnodeAddIntoHash(SVnodeObj *pVnode) {
taosHashPut(tsVnodesHash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnodeObj *));
}
void vnodeRemoveFromHash(SVnodeObj *pVnode) {
taosHashRemove(tsVnodesHash, &pVnode->vgId, sizeof(int32_t));
}
static void vnodeIncRef(void *ptNode) {
assert(ptNode != NULL);
SVnodeObj **ppVnode = (SVnodeObj **)ptNode;
assert(ppVnode);
assert(*ppVnode);
SVnodeObj *pVnode = *ppVnode;
atomic_add_fetch_32(&pVnode->refCount, 1);
vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
}
void *vnodeAcquire(int32_t vgId) {
SVnodeObj *pVnode = NULL;
if (tsVnodesHash != NULL) {
taosHashGetClone(tsVnodesHash, &vgId, sizeof(int32_t), vnodeIncRef, &pVnode);
}
if (pVnode == NULL) {
terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
vDebug("vgId:%d, not exist", vgId);
return NULL;
}
return pVnode;
}
void vnodeRelease(void *vparam) {
SVnodeObj *pVnode = vparam;
if (vparam == NULL) return;
int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1);
int32_t vgId = pVnode->vgId;
vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", vgId, refCount, pVnode);
assert(refCount >= 0);
if (refCount > 0) {
if (vnodeInResetStatus(pVnode) && refCount <= 3) {
tsem_post(&pVnode->sem);
}
} else {
vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", vgId, refCount, pVnode);
vnodeDestroyInMWorker(pVnode);
int32_t count = taosHashGetSize(tsVnodesHash);
vDebug("vgId:%d, vnode is destroyed, vnodes:%d", vgId, count);
}
}
void *vnodeAcquireNotClose(int32_t vgId) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode != NULL && pVnode->preClose == 1) {
vnodeRelease(pVnode);
terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
vDebug("vgId:%d, not exist, pre closing", vgId);
return NULL;
}
return pVnode;
}
static void vnodeBuildVloadMsg(SVnodeObj *pVnode, SStatusMsg *pStatus) {
int64_t totalStorage = 0;
int64_t compStorage = 0;
int64_t pointsWritten = 0;
if (vnodeInClosingStatus(pVnode)) return;
if (pStatus->openVnodes >= TSDB_MAX_VNODES) return;
if (pVnode->tsdb) {
tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage);
}
SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++];
pLoad->vgId = htonl(pVnode->vgId);
pLoad->dbCfgVersion = htonl(pVnode->dbCfgVersion);
pLoad->vgCfgVersion = htonl(pVnode->vgCfgVersion);
pLoad->totalStorage = htobe64(totalStorage);
pLoad->compStorage = htobe64(compStorage);
pLoad->pointsWritten = htobe64(pointsWritten);
pLoad->vnodeVersion = htobe64(pVnode->version);
pLoad->status = pVnode->status;
pLoad->role = pVnode->role;
pLoad->replica = pVnode->syncCfg.replica;
pLoad->compact = (pVnode->tsdb != NULL) ? tsdbGetCompactState(pVnode->tsdb) : 0;
}
int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) {
void *pIter = taosHashIterate(tsVnodesHash, NULL);
while (pIter) {
SVnodeObj **pVnode = pIter;
if (*pVnode) {
(*numOfVnodes)++;
if (*numOfVnodes >= TSDB_MAX_VNODES) {
vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES);
continue;
} else {
vnodeList[*numOfVnodes - 1] = (*pVnode)->vgId;
}
}
pIter = taosHashIterate(tsVnodesHash, pIter);
}
return TSDB_CODE_SUCCESS;
}
void vnodeBuildStatusMsg(void *param) {
SStatusMsg *pStatus = param;
void *pIter = taosHashIterate(tsVnodesHash, NULL);
while (pIter) {
SVnodeObj **pVnode = pIter;
if (*pVnode) {
vnodeBuildVloadMsg(*pVnode, pStatus);
}
pIter = taosHashIterate(tsVnodesHash, pIter);
}
}
void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes) {
for (int32_t i = 0; i < numOfVnodes; ++i) {
pAccess[i].vgId = htonl(pAccess[i].vgId);
SVnodeObj *pVnode = vnodeAcquireNotClose(pAccess[i].vgId);
if (pVnode != NULL) {
pVnode->accessState = pAccess[i].accessState;
if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) {
vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState);
}
vnodeRelease(pVnode);
}
}
}
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "taosmsg.h"
#include "query.h"
#include "dnode.h"
#include "vnodeVersion.h"
#include "vnodeMain.h"
#include "vnodeStatus.h"
uint32_t vnodeGetFileInfo(int32_t vgId, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fver) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
vError("vgId:%d, vnode not found while get file info", vgId);
return 0;
}
*fver = pVnode->fversion;
uint32_t ret = tsdbGetFileInfo(pVnode->tsdb, name, index, eindex, size);
vnodeRelease(pVnode);
return ret;
}
int32_t vnodeGetWalInfo(int32_t vgId, char *fileName, int64_t *fileId) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
vError("vgId:%d, vnode not found while get wal info", vgId);
return -1;
}
int32_t code = walGetWalFile(pVnode->wal, fileName, fileId);
vnodeRelease(pVnode);
return code;
}
void vnodeNotifyRole(int32_t vgId, int8_t role) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
vTrace("vgId:%d, vnode not found while notify role", vgId);
return;
}
if (pVnode->dropped) {
vTrace("vgId:%d, vnode dropped while notify role", vgId);
vnodeRelease(pVnode);
return;
}
vInfo("vgId:%d, sync role changed from %s to %s", pVnode->vgId, syncRole[pVnode->role], syncRole[role]);
pVnode->role = role;
dnodeSendStatusMsgToMnode();
if (pVnode->role == TAOS_SYNC_ROLE_MASTER) {
cqStart(pVnode->cq);
} else {
cqStop(pVnode->cq);
}
vnodeRelease(pVnode);
}
void vnodeCtrlFlow(int32_t vgId, int32_t level) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
vTrace("vgId:%d, vnode not found while flow ctrl", vgId);
return;
}
if (pVnode->dropped) {
vTrace("vgId:%d, vnode dropped while flow ctrl", vgId);
vnodeRelease(pVnode);
return;
}
if (pVnode->flowctrlLevel != level) {
vDebug("vgId:%d, set flowctrl level from %d to %d", pVnode->vgId, pVnode->flowctrlLevel, level);
pVnode->flowctrlLevel = level;
}
vnodeRelease(pVnode);
}
void vnodeStartSyncFile(int32_t vgId) {
SVnodeObj *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) {
vError("vgId:%d, vnode not found while start filesync", vgId);
return;
}
vInfo("vgId:%d, datafile will be synced", vgId);
vnodeSetResetStatus(pVnode);
vnodeRelease(pVnode);
}
void vnodeStopSyncFile(int32_t vgId, uint64_t fversion) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
vError("vgId:%d, vnode not found while stop filesync", vgId);
return;
}
pVnode->fversion = fversion;
pVnode->version = fversion;
vnodeSaveVersion(pVnode);
walResetVersion(pVnode->wal, fversion);
vInfo("vgId:%d, datafile is synced, fver:%" PRIu64 " vver:%" PRIu64, vgId, fversion, fversion);
vnodeSetReadyStatus(pVnode);
vnodeRelease(pVnode);
}
void vnodeConfirmForard(int32_t vgId, void *wparam, int32_t code) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
vError("vgId:%d, vnode not found while confirm forward", vgId);
}
if (code == TSDB_CODE_SYN_CONFIRM_EXPIRED && pVnode->status == TAOS_VN_STATUS_CLOSING) {
vDebug("vgId:%d, db:%s, vnode is closing while confirm forward", vgId, pVnode->db);
code = TSDB_CODE_VND_IS_CLOSING;
}
dnodeSendRpcVWriteRsp(pVnode, wparam, code);
vnodeRelease(pVnode);
}
int32_t vnodeWriteToCache(int32_t vgId, void *wparam, int32_t qtype, void *rparam) {
SVnodeObj *pVnode = vnodeAcquire(vgId);
if (pVnode == NULL) {
vError("vgId:%d, vnode not found while write to cache", vgId);
vnodeRelease(pVnode);
return TSDB_CODE_VND_INVALID_VGROUP_ID;
}
int32_t code = vnodeWriteToWQueue(pVnode, wparam, qtype, rparam);
vnodeRelease(pVnode);
return code;
}
int32_t vnodeGetVersion(int32_t vgId, uint64_t *fver, uint64_t *wver) {
SVnodeObj *pVnode = vnodeAcquireNotClose(vgId);
if (pVnode == NULL) {
vError("vgId:%d, vnode not found while write to cache", vgId);
return -1;
}
int32_t code = 0;
if (pVnode->isCommiting) {
vInfo("vgId:%d, vnode is commiting while get version", vgId);
code = -1;
} else {
*fver = pVnode->fversion;
*wver = pVnode->version;
}
vnodeRelease(pVnode);
return code;
}
void vnodeConfirmForward(void *vparam, uint64_t version, int32_t code, bool force) {
SVnodeObj *pVnode = vparam;
syncConfirmForward(pVnode->sync, version, code, force);
}
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "taoserror.h"
#include "taosmsg.h"
#include "tutil.h"
#include "tqueue.h"
#include "tglobal.h"
#include "vnodeWorker.h"
#include "vnodeMain.h"
typedef enum {
VNODE_WORKER_ACTION_CLEANUP,
VNODE_WORKER_ACTION_DESTROY
} EVMWorkerAction;
typedef struct {
int32_t vgId;
int32_t code;
void * rpcHandle;
SVnodeObj *pVnode;
EVMWorkerAction action;
} SVMWorkerMsg;
typedef struct {
pthread_t thread;
int32_t workerId;
} SVMWorker;
typedef struct {
int32_t curNum;
int32_t maxNum;
SVMWorker *worker;
} SVMWorkerPool;
static SVMWorkerPool tsVMWorkerPool;
static taos_qset tsVMWorkerQset;
static taos_queue tsVMWorkerQueue;
static void *vnodeMWorkerFunc(void *param);
static int32_t vnodeStartMWorker() {
tsVMWorkerQueue = taosOpenQueue();
if (tsVMWorkerQueue == NULL) return TSDB_CODE_DND_OUT_OF_MEMORY;
taosAddIntoQset(tsVMWorkerQset, tsVMWorkerQueue, NULL);
for (int32_t i = tsVMWorkerPool.curNum; i < tsVMWorkerPool.maxNum; ++i) {
SVMWorker *pWorker = tsVMWorkerPool.worker + i;
pWorker->workerId = i;
pthread_attr_t thAttr;
pthread_attr_init(&thAttr);
pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
if (pthread_create(&pWorker->thread, &thAttr, vnodeMWorkerFunc, pWorker) != 0) {
vError("failed to create thread to process vmworker queue, reason:%s", strerror(errno));
}
pthread_attr_destroy(&thAttr);
tsVMWorkerPool.curNum = i + 1;
vDebug("vmworker:%d is launched, total:%d", pWorker->workerId, tsVMWorkerPool.maxNum);
}
vDebug("vmworker queue:%p is allocated", tsVMWorkerQueue);
return TSDB_CODE_SUCCESS;
}
int32_t vnodeInitMWorker() {
tsVMWorkerQset = taosOpenQset();
tsVMWorkerPool.maxNum = 1;
tsVMWorkerPool.curNum = 0;
tsVMWorkerPool.worker = calloc(sizeof(SVMWorker), tsVMWorkerPool.maxNum);
if (tsVMWorkerPool.worker == NULL) return -1;
for (int32_t i = 0; i < tsVMWorkerPool.maxNum; ++i) {
SVMWorker *pWorker = tsVMWorkerPool.worker + i;
pWorker->workerId = i;
vDebug("vmworker:%d is created", i);
}
vDebug("vmworker is initialized, num:%d qset:%p", tsVMWorkerPool.maxNum, tsVMWorkerQset);
return vnodeStartMWorker();
}
static void vnodeStopMWorker() {
vDebug("vmworker queue:%p is freed", tsVMWorkerQueue);
taosCloseQueue(tsVMWorkerQueue);
tsVMWorkerQueue = NULL;
}
void vnodeCleanupMWorker() {
for (int32_t i = 0; i < tsVMWorkerPool.maxNum; ++i) {
SVMWorker *pWorker = tsVMWorkerPool.worker + i;
if (taosCheckPthreadValid(pWorker->thread)) {
taosQsetThreadResume(tsVMWorkerQset);
}
vDebug("vmworker:%d is closed", i);
}
for (int32_t i = 0; i < tsVMWorkerPool.maxNum; ++i) {
SVMWorker *pWorker = tsVMWorkerPool.worker + i;
vDebug("vmworker:%d start to join", i);
if (taosCheckPthreadValid(pWorker->thread)) {
pthread_join(pWorker->thread, NULL);
}
vDebug("vmworker:%d join success", i);
}
vDebug("vmworker is closed, qset:%p", tsVMWorkerQset);
taosCloseQset(tsVMWorkerQset);
tsVMWorkerQset = NULL;
tfree(tsVMWorkerPool.worker);
vnodeStopMWorker();
}
static int32_t vnodeWriteIntoMWorker(SVnodeObj *pVnode, EVMWorkerAction action, void *rpcHandle) {
SVMWorkerMsg *pMsg = taosAllocateQitem(sizeof(SVMWorkerMsg));
if (pMsg == NULL) return TSDB_CODE_VND_OUT_OF_MEMORY;
pMsg->vgId = pVnode->vgId;
pMsg->pVnode = pVnode;
pMsg->rpcHandle = rpcHandle;
pMsg->action = action;
int32_t code = taosWriteQitem(tsVMWorkerQueue, TAOS_QTYPE_RPC, pMsg);
if (code == 0) code = TSDB_CODE_DND_ACTION_IN_PROGRESS;
return code;
}
int32_t vnodeCleanupInMWorker(SVnodeObj *pVnode) {
vTrace("vgId:%d, will cleanup in vmworker", pVnode->vgId);
return vnodeWriteIntoMWorker(pVnode, VNODE_WORKER_ACTION_CLEANUP, NULL);
}
int32_t vnodeDestroyInMWorker(SVnodeObj *pVnode) {
vTrace("vgId:%d, will destroy in vmworker", pVnode->vgId);
return vnodeWriteIntoMWorker(pVnode, VNODE_WORKER_ACTION_DESTROY, NULL);
}
static void vnodeFreeMWorkerMsg(SVMWorkerMsg *pMsg) {
vTrace("vgId:%d, disposed in vmworker", pMsg->vgId);
taosFreeQitem(pMsg);
}
static void vnodeSendVMWorkerRpcRsp(SVMWorkerMsg *pMsg) {
if (pMsg->rpcHandle != NULL) {
SRpcMsg rpcRsp = {.handle = pMsg->rpcHandle, .code = pMsg->code};
rpcSendResponse(&rpcRsp);
}
vnodeFreeMWorkerMsg(pMsg);
}
static void vnodeProcessMWorkerMsg(SVMWorkerMsg *pMsg) {
pMsg->code = 0;
switch (pMsg->action) {
case VNODE_WORKER_ACTION_CLEANUP:
vnodeCleanUp(pMsg->pVnode);
break;
case VNODE_WORKER_ACTION_DESTROY:
vnodeDestroy(pMsg->pVnode);
break;
default:
break;
}
}
static void *vnodeMWorkerFunc(void *param) {
setThreadName("vnodeMWorker");
while (1) {
SVMWorkerMsg *pMsg = NULL;
if (taosReadQitemFromQset(tsVMWorkerQset, NULL, (void **)&pMsg, NULL) == 0) {
vDebug("qset:%p, vmworker got no message from qset, exiting", tsVMWorkerQset);
break;
}
vTrace("vgId:%d, action:%d will be processed in vmworker queue", pMsg->vgId, pMsg->action);
vnodeProcessMWorkerMsg(pMsg);
vnodeSendVMWorkerRpcRsp(pMsg);
}
return NULL;
}
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* This program is free software: you can use, redistribute, and/or modify
* it under the terms of the GNU Affero General Public License, version 3
* or later ("AGPL"), as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "os.h"
#include "tp.h"
#include "taosmsg.h"
#include "taoserror.h"
#include "tglobal.h"
#include "tqueue.h"
#include "ttimer.h"
#include "dnode.h"
#include "vnodeStatus.h"
#define MAX_QUEUED_MSG_NUM 100000
#define MAX_QUEUED_MSG_SIZE 1024*1024*1024 //1GB
extern void * tsDnodeTmr;
static int32_t (*vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MAX])(SVnodeObj *, void *pCont, SRspRet *);
static int32_t vnodeProcessSubmitMsg(SVnodeObj *pVnode, void *pCont, SRspRet *);
static int32_t vnodeProcessCreateTableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *);
static int32_t vnodeProcessDropTableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *);
static int32_t vnodeProcessAlterTableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *);
static int32_t vnodeProcessDropStableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *);
static int32_t vnodeProcessUpdateTagValMsg(SVnodeObj *pVnode, void *pCont, SRspRet *);
static int32_t vnodePerformFlowCtrl(SVWriteMsg *pWrite);
int32_t vnodeInitWrite(void) {
vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessSubmitMsg;
vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessCreateTableMsg;
vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessDropTableMsg;
vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessAlterTableMsg;
vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessDropStableMsg;
vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessUpdateTagValMsg;
return 0;
}
void vnodeCleanupWrite() {}
int32_t vnodeProcessWrite(void *vparam, void *wparam, int32_t qtype, void *rparam) {
int32_t code = 0;
SVnodeObj *pVnode = vparam;
SWalHead * pHead = wparam;
SVWriteMsg*pWrite = rparam;
SRspRet *pRspRet = NULL;
if (pWrite != NULL) pRspRet = &pWrite->rspRet;
if (vnodeProcessWriteMsgFp[pHead->msgType] == NULL) {
vError("vgId:%d, msg:%s not processed since no handle, qtype:%s hver:%" PRIu64, pVnode->vgId,
taosMsg[pHead->msgType], qtypeStr[qtype], pHead->version);
return TSDB_CODE_VND_MSG_NOT_PROCESSED;
}
vTrace("vgId:%d, msg:%s will be processed in vnode, qtype:%s hver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId,
taosMsg[pHead->msgType], qtypeStr[qtype], pHead->version, pVnode->version);
if (pHead->version == 0) { // from client or CQ
if (!vnodeInReadyStatus(pVnode)) {
vDebug("vgId:%d, msg:%s not processed since vstatus:%d, qtype:%s hver:%" PRIu64, pVnode->vgId,
taosMsg[pHead->msgType], pVnode->status, qtypeStr[qtype], pHead->version);
return TSDB_CODE_APP_NOT_READY; // it may be in deleting or closing state
}
if (pVnode->role != TAOS_SYNC_ROLE_MASTER) {
vDebug("vgId:%d, msg:%s not processed since replica:%d role:%s, qtype:%s hver:%" PRIu64, pVnode->vgId,
taosMsg[pHead->msgType], pVnode->syncCfg.replica, syncRole[pVnode->role], qtypeStr[qtype], pHead->version);
return TSDB_CODE_APP_NOT_READY;
}
// assign version
pHead->version = pVnode->version + 1;
} else { // from wal or forward
// for data from WAL or forward, version may be smaller
if (pHead->version <= pVnode->version) return 0;
}
// forward to peers, even it is WAL/FWD, it shall be called to update version in sync
int32_t syncCode = 0;
bool force = (pWrite == NULL ? false : pWrite->walHead.msgType != TSDB_MSG_TYPE_SUBMIT);
syncCode = syncForwardToPeer(pVnode->sync, pHead, pWrite, qtype, force);
if (syncCode < 0) {
pHead->version = 0;
return syncCode;
}
// write into WAL
code = walWrite(pVnode->wal, pHead);
if (code < 0) {
if (syncCode > 0) atomic_sub_fetch_32(&pWrite->processedCount, 1);
vError("vgId:%d, hver:%" PRIu64 " vver:%" PRIu64 " code:0x%x", pVnode->vgId, pHead->version, pVnode->version, code);
pHead->version = 0;
return code;
}
pVnode->version = pHead->version;
// write data locally
code = (*vnodeProcessWriteMsgFp[pHead->msgType])(pVnode, pHead->cont, pRspRet);
if (code < 0) {
if (syncCode > 0) atomic_sub_fetch_32(&pWrite->processedCount, 1);
return code;
}
return syncCode;
}
static int32_t vnodeCheckWrite(SVnodeObj *pVnode) {
if (!(pVnode->accessState & TSDB_VN_WRITE_ACCCESS)) {
vDebug("vgId:%d, no write auth, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
return TSDB_CODE_VND_NO_WRITE_AUTH;
}
if (pVnode->dbReplica != pVnode->syncCfg.replica &&
pVnode->syncCfg.nodeInfo[pVnode->syncCfg.replica - 1].nodeId == dnodeGetDnodeId()) {
vDebug("vgId:%d, vnode is balancing and will be dropped, dbReplica:%d vgReplica:%d, refCount:%d pVnode:%p",
pVnode->vgId, pVnode->dbReplica, pVnode->syncCfg.replica, pVnode->refCount, pVnode);
return TSDB_CODE_VND_IS_BALANCING;
}
// tsdb may be in reset state
if (pVnode->tsdb == NULL) {
vDebug("vgId:%d, tsdb is null, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
return TSDB_CODE_APP_NOT_READY;
}
if (pVnode->isFull) {
vDebug("vgId:%d, vnode is full, refCount:%d", pVnode->vgId, pVnode->refCount);
return TSDB_CODE_VND_IS_FULL;
}
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeProcessSubmitMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) {
int32_t code = TSDB_CODE_SUCCESS;
vTrace("vgId:%d, submit msg is processed", pVnode->vgId);
if (pVnode->dbType == TSDB_DB_TYPE_TOPIC && pVnode->role == TAOS_SYNC_ROLE_MASTER) {
tpUpdateTs(pVnode->vgId, &pVnode->sequence, pCont);
}
// save insert result into item
SShellSubmitRspMsg *pRsp = NULL;
if (pRet) {
pRet->len = sizeof(SShellSubmitRspMsg);
pRet->rsp = rpcMallocCont(pRet->len);
pRsp = pRet->rsp;
}
if (tsdbInsertData(pVnode->tsdb, pCont, pRsp) < 0) code = terrno;
return code;
}
static int32_t vnodeProcessCreateTableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) {
int code = TSDB_CODE_SUCCESS;
STableCfg *pCfg = tsdbCreateTableCfgFromMsg((SMDCreateTableMsg *)pCont);
if (pCfg == NULL) {
ASSERT(terrno != 0);
return terrno;
}
if (tsdbCreateTable(pVnode->tsdb, pCfg) < 0) {
code = terrno;
ASSERT(code != 0);
}
tsdbClearTableCfg(pCfg);
return code;
}
static int32_t vnodeProcessDropTableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) {
SMDDropTableMsg *pTable = pCont;
int32_t code = TSDB_CODE_SUCCESS;
vDebug("vgId:%d, table:%s, start to drop", pVnode->vgId, pTable->tableFname);
STableId tableId = {.uid = htobe64(pTable->uid), .tid = htonl(pTable->tid)};
if (tsdbDropTable(pVnode->tsdb, tableId) < 0) code = terrno;
return code;
}
static int32_t vnodeProcessAlterTableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) {
// TODO: disposed in tsdb
// STableCfg *pCfg = tsdbCreateTableCfgFromMsg((SMDCreateTableMsg *)pCont);
// if (pCfg == NULL) return terrno;
// if (tsdbCreateTable(pVnode->tsdb, pCfg) < 0) code = terrno;
// tsdbClearTableCfg(pCfg);
vDebug("vgId:%d, alter table msg is received", pVnode->vgId);
return TSDB_CODE_SUCCESS;
}
static int32_t vnodeProcessDropStableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) {
SDropSTableMsg *pTable = pCont;
int32_t code = TSDB_CODE_SUCCESS;
vDebug("vgId:%d, stable:%s, start to drop", pVnode->vgId, pTable->tableFname);
STableId stableId = {.uid = htobe64(pTable->uid), .tid = -1};
if (tsdbDropTable(pVnode->tsdb, stableId) < 0) code = terrno;
vDebug("vgId:%d, stable:%s, drop stable result:%s", pVnode->vgId, pTable->tableFname, tstrerror(code));
return code;
}
static int32_t vnodeProcessUpdateTagValMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) {
if (tsdbUpdateTableTagValue(pVnode->tsdb, (SUpdateTableTagValMsg *)pCont) < 0) {
return terrno;
}
return TSDB_CODE_SUCCESS;
}
static SVWriteMsg *vnodeBuildVWriteMsg(SVnodeObj *pVnode, SWalHead *pHead, int32_t qtype, SRpcMsg *pRpcMsg) {
if (pHead->len > TSDB_MAX_WAL_SIZE) {
vError("vgId:%d, wal len:%d exceeds limit, hver:%" PRIu64, pVnode->vgId, pHead->len, pHead->version);
terrno = TSDB_CODE_WAL_SIZE_LIMIT;
return NULL;
}
int32_t size = sizeof(SVWriteMsg) + pHead->len;
SVWriteMsg *pWrite = taosAllocateQitem(size);
if (pWrite == NULL) {
terrno = TSDB_CODE_VND_OUT_OF_MEMORY;
return NULL;
}
if (pRpcMsg != NULL) {
pWrite->rpcMsg = *pRpcMsg;
}
memcpy(&pWrite->walHead, pHead, sizeof(SWalHead) + pHead->len);
pWrite->pVnode = pVnode;
pWrite->qtype = qtype;
atomic_add_fetch_32(&pVnode->refCount, 1);
return pWrite;
}
static int32_t vnodeWriteToWQueueImp(SVWriteMsg *pWrite) {
SVnodeObj *pVnode = pWrite->pVnode;
if (pWrite->qtype == TAOS_QTYPE_RPC) {
int32_t code = vnodeCheckWrite(pVnode);
if (code != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to write into vwqueue since %s", pVnode->vgId, tstrerror(code));
taosFreeQitem(pWrite);
vnodeRelease(pVnode);
return code;
}
}
if (tsAvailDataDirGB <= tsMinimalDataDirGB) {
vError("vgId:%d, failed to write into vwqueue since no diskspace, avail:%fGB", pVnode->vgId, tsAvailDataDirGB);
taosFreeQitem(pWrite);
vnodeRelease(pVnode);
return TSDB_CODE_VND_NO_DISKSPACE;
}
if (!vnodeInReadyOrUpdatingStatus(pVnode)) {
vError("vgId:%d, failed to write into vwqueue, vstatus is %s, refCount:%d pVnode:%p", pVnode->vgId,
vnodeStatus[pVnode->status], pVnode->refCount, pVnode);
taosFreeQitem(pWrite);
vnodeRelease(pVnode);
return TSDB_CODE_APP_NOT_READY;
}
int32_t queued = atomic_add_fetch_32(&pVnode->queuedWMsg, 1);
int64_t queuedSize = atomic_add_fetch_64(&pVnode->queuedWMsgSize, pWrite->walHead.len);
if (queued > MAX_QUEUED_MSG_NUM || queuedSize > MAX_QUEUED_MSG_SIZE) {
int32_t ms = (queued / MAX_QUEUED_MSG_NUM) * 10 + 3;
if (ms > 100) ms = 100;
vDebug("vgId:%d, too many msg:%d in vwqueue, flow control %dms", pVnode->vgId, queued, ms);
taosMsleep(ms);
}
vTrace("vgId:%d, write into vwqueue, refCount:%d queued:%d size:%" PRId64, pVnode->vgId, pVnode->refCount,
pVnode->queuedWMsg, pVnode->queuedWMsgSize);
taosWriteQitem(pVnode->wqueue, pWrite->qtype, pWrite);
return TSDB_CODE_SUCCESS;
}
int32_t vnodeWriteToWQueue(void *vparam, void *wparam, int32_t qtype, void *rparam) {
SVnodeObj *pVnode = vparam;
if (qtype == TAOS_QTYPE_RPC) {
if (!vnodeInReadyStatus(pVnode)) {
return TSDB_CODE_APP_NOT_READY; // it may be in deleting or closing state
}
if (pVnode->role != TAOS_SYNC_ROLE_MASTER) {
return TSDB_CODE_APP_NOT_READY;
}
}
SVWriteMsg *pWrite = vnodeBuildVWriteMsg(vparam, wparam, qtype, rparam);
if (pWrite == NULL) {
assert(terrno != 0);
return terrno;
}
int32_t code = vnodePerformFlowCtrl(pWrite);
if (code != 0) return 0;
return vnodeWriteToWQueueImp(pWrite);
}
void vnodeFreeFromWQueue(void *vparam, SVWriteMsg *pWrite) {
SVnodeObj *pVnode = vparam;
if (pVnode) {
int32_t queued = atomic_sub_fetch_32(&pVnode->queuedWMsg, 1);
int64_t queuedSize = atomic_sub_fetch_64(&pVnode->queuedWMsgSize, pWrite->walHead.len);
vTrace("vgId:%d, msg:%p, app:%p, free from vwqueue, queued:%d size:%" PRId64, pVnode->vgId, pWrite,
pWrite->rpcMsg.ahandle, queued, queuedSize);
}
taosFreeQitem(pWrite);
vnodeRelease(pVnode);
}
static void vnodeFlowCtrlMsgToWQueue(void *param, void *tmrId) {
SVWriteMsg *pWrite = param;
SVnodeObj * pVnode = pWrite->pVnode;
int32_t code = TSDB_CODE_VND_IS_SYNCING;
if (pVnode->flowctrlLevel <= 0) code = TSDB_CODE_VND_IS_FLOWCTRL;
pWrite->processedCount++;
if (pWrite->processedCount >= 100) {
vError("vgId:%d, msg:%p, failed to process since %s, retry:%d", pVnode->vgId, pWrite, tstrerror(code),
pWrite->processedCount);
void *handle = pWrite->rpcMsg.handle;
taosFreeQitem(pWrite);
vnodeRelease(pVnode);
SRpcMsg rpcRsp = {.handle = handle, .code = code};
rpcSendResponse(&rpcRsp);
} else {
code = vnodePerformFlowCtrl(pWrite);
if (code == 0) {
vDebug("vgId:%d, msg:%p, write into vwqueue after flowctrl, retry:%d", pVnode->vgId, pWrite,
pWrite->processedCount);
pWrite->processedCount = 0;
void *handle = pWrite->rpcMsg.handle;
code = vnodeWriteToWQueueImp(pWrite);
if (code != TSDB_CODE_SUCCESS) {
SRpcMsg rpcRsp = {.handle = handle, .code = code};
rpcSendResponse(&rpcRsp);
}
}
}
}
static int32_t vnodePerformFlowCtrl(SVWriteMsg *pWrite) {
SVnodeObj *pVnode = pWrite->pVnode;
if (pWrite->qtype != TAOS_QTYPE_RPC) return 0;
if (pVnode->queuedWMsg < MAX_QUEUED_MSG_NUM && pVnode->queuedWMsgSize < MAX_QUEUED_MSG_SIZE &&
pVnode->flowctrlLevel <= 0)
return 0;
if (tsEnableFlowCtrl == 0) {
int32_t ms = (int32_t)pow(2, pVnode->flowctrlLevel + 2);
if (ms > 100) ms = 100;
vTrace("vgId:%d, msg:%p, app:%p, perform flowctrl for %d ms", pVnode->vgId, pWrite, pWrite->rpcMsg.ahandle, ms);
taosMsleep(ms);
return 0;
} else {
void *unUsedTimerId = NULL;
taosTmrReset(vnodeFlowCtrlMsgToWQueue, 100, pWrite, tsDnodeTmr, &unUsedTimerId);
vTrace("vgId:%d, msg:%p, app:%p, perform flowctrl, retry:%d", pVnode->vgId, pWrite, pWrite->rpcMsg.ahandle,
pWrite->processedCount);
return TSDB_CODE_VND_ACTION_IN_PROGRESS;
}
}
void vnodeWaitWriteCompleted(SVnodeObj *pVnode) {
int32_t extraSleep = 0;
while (pVnode->queuedWMsg > 0) {
vTrace("vgId:%d, queued wmsg num:%d", pVnode->vgId, pVnode->queuedWMsg);
taosMsleep(10);
extraSleep = 1;
}
if (extraSleep)
taosMsleep(900);
}
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册