提交 acd35eaf 编写于 作者: H Haojun Liao

[TD-225] merge develop

......@@ -20,6 +20,7 @@
#include "tlog.h"
#include "tutil.h"
#include "ttimer.h"
#include "tref.h"
#include "tsocket.h"
#include "tglobal.h"
#include "taoserror.h"
......@@ -43,6 +44,7 @@ char tsNodeFqdn[TSDB_FQDN_LEN];
static ttpool_h tsTcpPool;
static void * syncTmrCtrl = NULL;
static void * vgIdHash;
static int tsSyncRefId = -1;
// local functions
static void syncProcessSyncRequest(char *pMsg, SSyncPeer *pPeer);
......@@ -54,13 +56,13 @@ static int syncProcessPeerMsg(void *param, void *buffer);
static void syncProcessIncommingConnection(int connFd, uint32_t sourceIp);
static void syncRemovePeer(SSyncPeer *pPeer);
static void syncAddArbitrator(SSyncNode *pNode);
static void syncAddNodeRef(SSyncNode *pNode);
static void syncDecNodeRef(SSyncNode *pNode);
static void syncFreeNode(void *);
static void syncRemoveConfirmedFwdInfo(SSyncNode *pNode);
static void syncMonitorFwdInfos(void *param, void *tmrId);
static void syncProcessFwdAck(SSyncNode *pNode, SFwdInfo *pFwdInfo, int32_t code);
static void syncSaveFwdInfo(SSyncNode *pNode, uint64_t version, void *mhandle);
static void syncRestartPeer(SSyncPeer *pPeer);
static int32_t syncForwardToPeerImpl(SSyncNode *pNode, void *data, void *mhandle, int qtyp);
static SSyncPeer *syncAddPeer(SSyncNode *pNode, const SNodeInfo *pInfo);
char* syncRole[] = {
......@@ -106,6 +108,12 @@ int32_t syncInit() {
return -1;
}
tsSyncRefId = taosOpenRef(200, syncFreeNode);
if (tsSyncRefId < 0) {
syncCleanUp();
return -1;
}
tstrncpy(tsNodeFqdn, tsLocalFqdn, sizeof(tsNodeFqdn));
sInfo("sync module initialized successfully");
......@@ -128,6 +136,9 @@ void syncCleanUp() {
vgIdHash = NULL;
}
taosCloseRef(tsSyncRefId);
tsSyncRefId = -1;
sInfo("sync module is cleaned up");
}
......@@ -159,6 +170,12 @@ void *syncStart(const SSyncInfo *pInfo) {
pNode->quorum = pCfg->quorum;
if (pNode->quorum > pNode->replica) pNode->quorum = pNode->replica;
int ret = taosAddRef(tsSyncRefId, pNode);
if (ret < 0) {
syncFreeNode(pNode);
return NULL;
}
for (int i = 0; i < pCfg->replica; ++i) {
const SNodeInfo *pNodeInfo = pCfg->nodeInfo + i;
pNode->peerInfo[i] = syncAddPeer(pNode, pNodeInfo);
......@@ -167,8 +184,6 @@ void *syncStart(const SSyncInfo *pInfo) {
}
}
syncAddNodeRef(pNode);
if (pNode->selfIndex < 0) {
sInfo("vgId:%d, this node is not configured", pNode->vgId);
terrno = TSDB_CODE_SYN_INVALID_CONFIG;
......@@ -210,7 +225,9 @@ void syncStop(void *param) {
SSyncNode *pNode = param;
SSyncPeer *pPeer;
if (pNode == NULL) return;
int ret = taosAcquireRef(tsSyncRefId, pNode);
if (ret < 0) return;
sInfo("vgId:%d, cleanup sync", pNode->vgId);
pthread_mutex_lock(&(pNode->mutex));
......@@ -228,14 +245,17 @@ void syncStop(void *param) {
pthread_mutex_unlock(&(pNode->mutex));
syncDecNodeRef(pNode);
taosReleaseRef(tsSyncRefId, pNode);
taosRemoveRef(tsSyncRefId, pNode);
}
int32_t syncReconfig(void *param, const SSyncCfg *pNewCfg) {
SSyncNode *pNode = param;
int i, j;
if (pNode == NULL) return TSDB_CODE_SYN_INVALID_CONFIG;
int ret = taosAcquireRef(tsSyncRefId, pNode);
if (ret < 0) return TSDB_CODE_SYN_INVALID_CONFIG;
sInfo("vgId:%d, reconfig, role:%s replica:%d old:%d", pNode->vgId, syncRole[nodeRole], pNewCfg->replica,
pNode->replica);
......@@ -298,105 +318,63 @@ int32_t syncReconfig(void *param, const SSyncCfg *pNewCfg) {
syncRole[nodeRole]);
syncBroadcastStatus(pNode);
taosReleaseRef(tsSyncRefId, pNode);
return 0;
}
int32_t syncForwardToPeer(void *param, void *data, void *mhandle, int qtype) {
SSyncNode *pNode = param;
SSyncPeer *pPeer;
SSyncHead *pSyncHead;
SWalHead * pWalHead = data;
int fwdLen;
int code = 0;
if (pNode == NULL) return 0;
if (nodeRole == TAOS_SYNC_ROLE_SLAVE && pWalHead->version != nodeVersion + 1) {
sError("vgId:%d, received ver:%" PRIu64 ", inconsistent with last ver:%" PRIu64 ", restart connection", pNode->vgId,
pWalHead->version, nodeVersion);
for (int i = 0; i < pNode->replica; ++i) {
pPeer = pNode->peerInfo[i];
syncRestartConnection(pPeer);
}
return TSDB_CODE_SYN_INVALID_VERSION;
}
// always update version
nodeVersion = pWalHead->version;
sDebug("vgId:%d, replica:%d nodeRole:%s qtype:%d ver:%" PRIu64, pNode->vgId, pNode->replica, syncRole[nodeRole],
qtype, pWalHead->version);
if (pNode->replica == 1 || nodeRole != TAOS_SYNC_ROLE_MASTER) return 0;
// only pkt from RPC or CQ can be forwarded
if (qtype != TAOS_QTYPE_RPC && qtype != TAOS_QTYPE_CQ) return 0;
// a hacker way to improve the performance
pSyncHead = (SSyncHead *)(((char *)pWalHead) - sizeof(SSyncHead));
pSyncHead->type = TAOS_SMSG_FORWARD;
pSyncHead->pversion = 0;
pSyncHead->len = sizeof(SWalHead) + pWalHead->len;
fwdLen = pSyncHead->len + sizeof(SSyncHead); // include the WAL and SYNC head
int ret = taosAcquireRef(tsSyncRefId, pNode);
if (ret < 0) return 0;
pthread_mutex_lock(&(pNode->mutex));
int32_t code = syncForwardToPeerImpl(pNode, data, mhandle, qtype);
for (int i = 0; i < pNode->replica; ++i) {
pPeer = pNode->peerInfo[i];
if (pPeer == NULL || pPeer->peerFd < 0) continue;
if (pPeer->role != TAOS_SYNC_ROLE_SLAVE && pPeer->sstatus != TAOS_SYNC_STATUS_CACHE) continue;
if (pNode->quorum > 1 && code == 0) {
syncSaveFwdInfo(pNode, pWalHead->version, mhandle);
code = 1;
}
int retLen = write(pPeer->peerFd, pSyncHead, fwdLen);
if (retLen == fwdLen) {
sDebug("%s, forward is sent, ver:%" PRIu64 " contLen:%d", pPeer->id, pWalHead->version, pWalHead->len);
} else {
sError("%s, failed to forward, ver:%" PRIu64 " retLen:%d", pPeer->id, pWalHead->version, retLen);
syncRestartConnection(pPeer);
}
}
pthread_mutex_unlock(&(pNode->mutex));
taosReleaseRef(tsSyncRefId, pNode);
return code;
}
void syncConfirmForward(void *param, uint64_t version, int32_t code) {
SSyncNode *pNode = param;
if (pNode == NULL) return;
if (pNode->quorum <= 1) return;
SSyncPeer *pPeer = pNode->pMaster;
if (pPeer == NULL) return;
int ret = taosAcquireRef(tsSyncRefId, pNode);
if (ret < 0) return;
char msg[sizeof(SSyncHead) + sizeof(SFwdRsp)] = {0};
SSyncPeer *pPeer = pNode->pMaster;
if (pPeer && pNode->quorum > 1) {
char msg[sizeof(SSyncHead) + sizeof(SFwdRsp)] = {0};
SSyncHead *pHead = (SSyncHead *)msg;
pHead->type = TAOS_SMSG_FORWARD_RSP;
pHead->len = sizeof(SFwdRsp);
SSyncHead *pHead = (SSyncHead *)msg;
pHead->type = TAOS_SMSG_FORWARD_RSP;
pHead->len = sizeof(SFwdRsp);
SFwdRsp *pFwdRsp = (SFwdRsp *)(msg + sizeof(SSyncHead));
pFwdRsp->version = version;
pFwdRsp->code = code;
SFwdRsp *pFwdRsp = (SFwdRsp *)(msg + sizeof(SSyncHead));
pFwdRsp->version = version;
pFwdRsp->code = code;
int msgLen = sizeof(SSyncHead) + sizeof(SFwdRsp);
int retLen = write(pPeer->peerFd, msg, msgLen);
int msgLen = sizeof(SSyncHead) + sizeof(SFwdRsp);
int retLen = write(pPeer->peerFd, msg, msgLen);
if (retLen == msgLen) {
sDebug("%s, forward-rsp is sent, ver:%" PRIu64, pPeer->id, version);
} else {
sDebug("%s, failed to send forward ack, restart", pPeer->id);
syncRestartConnection(pPeer);
if (retLen == msgLen) {
sDebug("%s, forward-rsp is sent, ver:%" PRIu64, pPeer->id, version);
} else {
sDebug("%s, failed to send forward ack, restart", pPeer->id);
syncRestartConnection(pPeer);
}
}
taosReleaseRef(tsSyncRefId, pNode);
}
void syncRecover(void *param) {
SSyncNode *pNode = param;
SSyncPeer *pPeer;
int ret = taosAcquireRef(tsSyncRefId, pNode);
if (ret < 0) return;
// to do: add a few lines to check if recover is OK
// if take this node to unsync state, the whole system may not work
......@@ -414,17 +392,24 @@ void syncRecover(void *param) {
}
pthread_mutex_unlock(&(pNode->mutex));
taosReleaseRef(tsSyncRefId, pNode);
}
int syncGetNodesRole(void *param, SNodesRole *pNodesRole) {
SSyncNode *pNode = param;
int ret = taosAcquireRef(tsSyncRefId, pNode);
if (ret < 0) return -1;
pNodesRole->selfIndex = pNode->selfIndex;
for (int i = 0; i < pNode->replica; ++i) {
pNodesRole->nodeId[i] = pNode->peerInfo[i]->nodeId;
pNodesRole->role[i] = pNode->peerInfo[i]->role;
}
taosReleaseRef(tsSyncRefId, pNode);
return 0;
}
......@@ -457,22 +442,20 @@ static void syncAddArbitrator(SSyncNode *pNode) {
pNode->peerInfo[TAOS_SYNC_MAX_REPLICA] = syncAddPeer(pNode, &nodeInfo);
}
static void syncAddNodeRef(SSyncNode *pNode) { atomic_add_fetch_8(&pNode->refCount, 1); }
static void syncFreeNode(void *param) {
SSyncNode *pNode = param;
static void syncDecNodeRef(SSyncNode *pNode) {
if (atomic_sub_fetch_8(&pNode->refCount, 1) == 0) {
pthread_mutex_destroy(&pNode->mutex);
taosTFree(pNode->pRecv);
taosTFree(pNode->pSyncFwds);
taosTFree(pNode);
}
pthread_mutex_destroy(&pNode->mutex);
taosTFree(pNode->pRecv);
taosTFree(pNode->pSyncFwds);
taosTFree(pNode);
}
void syncAddPeerRef(SSyncPeer *pPeer) { atomic_add_fetch_8(&pPeer->refCount, 1); }
int syncDecPeerRef(SSyncPeer *pPeer) {
if (atomic_sub_fetch_8(&pPeer->refCount, 1) == 0) {
syncDecNodeRef(pPeer->pSyncNode);
taosReleaseRef(tsSyncRefId, pPeer->pSyncNode);
sDebug("%s, resource is freed", pPeer->id);
taosTFree(pPeer->watchFd);
......@@ -529,7 +512,7 @@ static SSyncPeer *syncAddPeer(SSyncNode *pNode, const SNodeInfo *pInfo) {
taosTmrReset(syncCheckPeerConnection, checkMs, pPeer, syncTmrCtrl, &pPeer->timer);
}
syncAddNodeRef(pNode);
taosAcquireRef(tsSyncRefId, pNode);
return pPeer;
}
......@@ -1122,7 +1105,7 @@ static void syncProcessBrokenLink(void *param) {
SSyncPeer *pPeer = param;
SSyncNode *pNode = pPeer->pSyncNode;
syncAddNodeRef(pNode);
if (taosAcquireRef(tsSyncRefId, pNode) < 0) return;
pthread_mutex_lock(&(pNode->mutex));
sDebug("%s, TCP link is broken(%s)", pPeer->id, strerror(errno));
......@@ -1133,7 +1116,7 @@ static void syncProcessBrokenLink(void *param) {
}
pthread_mutex_unlock(&(pNode->mutex));
syncDecNodeRef(pNode);
taosReleaseRef(tsSyncRefId, pNode);
}
static void syncSaveFwdInfo(SSyncNode *pNode, uint64_t version, void *mhandle) {
......@@ -1202,22 +1185,90 @@ static void syncProcessFwdAck(SSyncNode *pNode, SFwdInfo *pFwdInfo, int32_t code
static void syncMonitorFwdInfos(void *param, void *tmrId) {
SSyncNode *pNode = param;
int ret = taosAcquireRef(tsSyncRefId, pNode);
if ( ret < 0) return;
SSyncFwds *pSyncFwds = pNode->pSyncFwds;
if (pSyncFwds == NULL) return;
uint64_t time = taosGetTimestampMs();
if (pSyncFwds) {;
uint64_t time = taosGetTimestampMs();
if (pSyncFwds->fwds > 0) {
pthread_mutex_lock(&(pNode->mutex));
for (int i = 0; i < pSyncFwds->fwds; ++i) {
SFwdInfo *pFwdInfo = pSyncFwds->fwdInfo + (pSyncFwds->first + i) % tsMaxFwdInfo;
if (time - pFwdInfo->time < 2000) break;
syncProcessFwdAck(pNode, pFwdInfo, TSDB_CODE_RPC_NETWORK_UNAVAIL);
if (pSyncFwds->fwds > 0) {
pthread_mutex_lock(&(pNode->mutex));
for (int i = 0; i < pSyncFwds->fwds; ++i) {
SFwdInfo *pFwdInfo = pSyncFwds->fwdInfo + (pSyncFwds->first + i) % tsMaxFwdInfo;
if (time - pFwdInfo->time < 2000) break;
syncProcessFwdAck(pNode, pFwdInfo, TSDB_CODE_RPC_NETWORK_UNAVAIL);
}
syncRemoveConfirmedFwdInfo(pNode);
pthread_mutex_unlock(&(pNode->mutex));
}
syncRemoveConfirmedFwdInfo(pNode);
pthread_mutex_unlock(&(pNode->mutex));
pNode->pFwdTimer = taosTmrStart(syncMonitorFwdInfos, 300, pNode, syncTmrCtrl);
}
pNode->pFwdTimer = taosTmrStart(syncMonitorFwdInfos, 300, pNode, syncTmrCtrl);
taosReleaseRef(tsSyncRefId, pNode);
}
static int32_t syncForwardToPeerImpl(SSyncNode *pNode, void *data, void *mhandle, int qtype) {
SSyncPeer *pPeer;
SSyncHead *pSyncHead;
SWalHead * pWalHead = data;
int fwdLen;
int32_t code = 0;
if (nodeRole == TAOS_SYNC_ROLE_SLAVE && pWalHead->version != nodeVersion + 1) {
sError("vgId:%d, received ver:%" PRIu64 ", inconsistent with last ver:%" PRIu64 ", restart connection", pNode->vgId,
pWalHead->version, nodeVersion);
for (int i = 0; i < pNode->replica; ++i) {
pPeer = pNode->peerInfo[i];
syncRestartConnection(pPeer);
}
return TSDB_CODE_SYN_INVALID_VERSION;
}
// always update version
nodeVersion = pWalHead->version;
sDebug("vgId:%d, replica:%d nodeRole:%s qtype:%d ver:%" PRIu64, pNode->vgId, pNode->replica, syncRole[nodeRole],
qtype, pWalHead->version);
if (pNode->replica == 1 || nodeRole != TAOS_SYNC_ROLE_MASTER) return 0;
// only pkt from RPC or CQ can be forwarded
if (qtype != TAOS_QTYPE_RPC && qtype != TAOS_QTYPE_CQ) return 0;
// a hacker way to improve the performance
pSyncHead = (SSyncHead *)(((char *)pWalHead) - sizeof(SSyncHead));
pSyncHead->type = TAOS_SMSG_FORWARD;
pSyncHead->pversion = 0;
pSyncHead->len = sizeof(SWalHead) + pWalHead->len;
fwdLen = pSyncHead->len + sizeof(SSyncHead); // include the WAL and SYNC head
pthread_mutex_lock(&(pNode->mutex));
for (int i = 0; i < pNode->replica; ++i) {
pPeer = pNode->peerInfo[i];
if (pPeer == NULL || pPeer->peerFd < 0) continue;
if (pPeer->role != TAOS_SYNC_ROLE_SLAVE && pPeer->sstatus != TAOS_SYNC_STATUS_CACHE) continue;
if (pNode->quorum > 1 && code == 0) {
syncSaveFwdInfo(pNode, pWalHead->version, mhandle);
code = 1;
}
int retLen = write(pPeer->peerFd, pSyncHead, fwdLen);
if (retLen == fwdLen) {
sDebug("%s, forward is sent, ver:%" PRIu64 " contLen:%d", pPeer->id, pWalHead->version, pWalHead->len);
} else {
sError("%s, failed to forward, ver:%" PRIu64 " retLen:%d", pPeer->id, pWalHead->version, retLen);
syncRestartConnection(pPeer);
}
}
pthread_mutex_unlock(&(pNode->mutex));
return code;
}
......@@ -562,12 +562,12 @@ int tsdbUnlockRepoMeta(STsdbRepo *pRepo) {
void tsdbRefTable(STable *pTable) {
int32_t ref = T_REF_INC(pTable);
UNUSED(ref);
// tsdbDebug("ref table %"PRIu64", tid:%d, refCount:%d", TABLE_UID(pTable), TABLE_TID(pTable), ref);
tsdbDebug("ref table %s uid %" PRIu64 " tid:%d, refCount:%d", TABLE_CHAR_NAME(pTable), TABLE_UID(pTable), TABLE_TID(pTable), ref);
}
void tsdbUnRefTable(STable *pTable) {
int32_t ref = T_REF_DEC(pTable);
tsdbDebug("unref table uid:%"PRIu64", tid:%d, refCount:%d", TABLE_UID(pTable), TABLE_TID(pTable), ref);
tsdbDebug("unref table %s uid:%"PRIu64" tid:%d, refCount:%d", TABLE_CHAR_NAME(pTable), TABLE_UID(pTable), TABLE_TID(pTable), ref);
if (ref == 0) {
// tsdbDebug("destory table name:%s uid:%"PRIu64", tid:%d", TABLE_CHAR_NAME(pTable), TABLE_UID(pTable), TABLE_TID(pTable));
......@@ -745,7 +745,7 @@ static STable *tsdbCreateTableFromCfg(STableCfg *pCfg, bool isSuper) {
T_REF_INC(pTable);
tsdbTrace("table %s tid %d uid %" PRIu64 " is created", TABLE_CHAR_NAME(pTable), TABLE_TID(pTable),
tsdbDebug("table %s tid %d uid %" PRIu64 " is created", TABLE_CHAR_NAME(pTable), TABLE_TID(pTable),
TABLE_UID(pTable));
return pTable;
......@@ -889,7 +889,7 @@ static void tsdbRemoveTableFromMeta(STsdbRepo *pRepo, STable *pTable, bool rmFro
}
if (lock) tsdbUnlockRepoMeta(pRepo);
tsdbDebug("vgId:%d table %s is removed from meta", REPO_ID(pRepo), TABLE_CHAR_NAME(pTable));
tsdbDebug("vgId:%d table %s uid %" PRIu64 " is removed from meta", REPO_ID(pRepo), TABLE_CHAR_NAME(pTable), TABLE_UID(pTable));
tsdbUnRefTable(pTable);
}
......
......@@ -2122,8 +2122,8 @@ STimeWindow changeTableGroupByLastrow(STableGroupInfo *groupList) {
}
// clear current group, unref unused table
for(int32_t i = 0; i < numOfTables; ++i) {
STableKeyInfo* pKeyInfo = (STableKeyInfo*) taosArrayGet(pGroup, i);
for (int32_t i = 0; i < numOfTables; ++i) {
STableKeyInfo* pKeyInfo = (STableKeyInfo*)taosArrayGet(pGroup, i);
// keyInfo.pTable may be NULL here.
if (pKeyInfo->pTable != keyInfo.pTable) {
......
......@@ -21,15 +21,42 @@
extern "C" {
#endif
int taosOpenRef(int max, void (*fp)(void *)); // return refId which will be used by other APIs
// open an instance, return refId which will be used by other APIs
int taosOpenRef(int max, void (*fp)(void *));
// close the Ref instance
void taosCloseRef(int refId);
int taosListRef(); // return the number of references in system
// add ref, p is the pointer to resource or pointer ID
int taosAddRef(int refId, void *p);
#define taosRemoveRef taosReleaseRef
// acquire ref, p is the pointer to resource or pointer ID
int taosAcquireRef(int refId, void *p);
// release ref, p is the pointer to resource or pinter ID
void taosReleaseRef(int refId, void *p);
#define taosRemoveRef taosReleaseRef
// return the first if p is null, otherwise return the next after p
void *taosIterateRef(int refId, void *p);
// return the number of references in system
int taosListRef();
/* sample code to iterate the refs
void demoIterateRefs(int refId) {
void *p = taosIterateRef(refId, NULL);
while (p) {
// process P
p = taosIterateRef(refId, p);
}
}
*/
#ifdef __cplusplus
}
......
......@@ -143,8 +143,6 @@ int taosAddRef(int refId, void *p)
return TSDB_CODE_REF_INVALID_ID;
}
uTrace("refId:%d p:%p try to add", refId, p);
pSet = tsRefSetList + refId;
taosIncRefCount(pSet);
if (pSet->state != TSDB_REF_STATE_ACTIVE) {
......@@ -203,8 +201,6 @@ int taosAcquireRef(int refId, void *p)
return TSDB_CODE_REF_INVALID_ID;
}
uTrace("refId:%d p:%p try to acquire", refId, p);
pSet = tsRefSetList + refId;
taosIncRefCount(pSet);
if (pSet->state != TSDB_REF_STATE_ACTIVE) {
......@@ -254,8 +250,6 @@ void taosReleaseRef(int refId, void *p)
return;
}
uTrace("refId:%d p:%p try to release", refId, p);
pSet = tsRefSetList + refId;
if (pSet->state == TSDB_REF_STATE_EMPTY) {
uTrace("refId:%d p:%p failed to release, cleaned", refId, p);
......@@ -305,6 +299,75 @@ void taosReleaseRef(int refId, void *p)
if (released) taosDecRefCount(pSet);
}
// if p is NULL, return the first p in hash list, otherwise, return the next after p
void *taosIterateRef(int refId, void *p) {
SRefNode *pNode = NULL;
SRefSet *pSet;
if (refId < 0 || refId >= TSDB_REF_OBJECTS) {
uTrace("refId:%d p:%p failed to iterate, refId not valid", refId, p);
return NULL;
}
pSet = tsRefSetList + refId;
taosIncRefCount(pSet);
if (pSet->state != TSDB_REF_STATE_ACTIVE) {
uTrace("refId:%d p:%p failed to iterate, not active", refId, p);
taosDecRefCount(pSet);
return NULL;
}
int hash = 0;
if (p) {
hash = taosHashRef(pSet, p);
taosLockList(pSet->lockedBy+hash);
pNode = pSet->nodeList[hash];
while (pNode) {
if (pNode->p == p) break;
pNode = pNode->next;
}
if (pNode == NULL) {
uError("refId:%d p:%p not there, quit", refId, p);
taosUnlockList(pSet->lockedBy+hash);
return NULL;
}
// p is there
pNode = pNode->next;
if (pNode == NULL) {
taosUnlockList(pSet->lockedBy+hash);
hash++;
}
}
if (pNode == NULL) {
for (; hash < pSet->max; ++hash) {
taosLockList(pSet->lockedBy+hash);
pNode = pSet->nodeList[hash];
if (pNode) break;
taosUnlockList(pSet->lockedBy+hash);
}
}
void *newP = NULL;
if (pNode) {
pNode->count++; // acquire it
newP = pNode->p;
taosUnlockList(pSet->lockedBy+hash);
uTrace("refId:%d p:%p is returned", refId, p);
} else {
uTrace("refId:%d p:%p the list is over", refId, p);
}
if (p) taosReleaseRef(refId, p); // release the current one
taosDecRefCount(pSet);
return newP;
}
int taosListRef() {
SRefSet *pSet;
SRefNode *pNode;
......
......@@ -17,6 +17,19 @@ typedef struct {
void **p;
} SRefSpace;
void iterateRefs(int refId) {
int count = 0;
void *p = taosIterateRef(refId, NULL);
while (p) {
// process P
count++;
p = taosIterateRef(refId, p);
}
printf(" %d ", count);
}
void *takeRefActions(void *param) {
SRefSpace *pSpace = (SRefSpace *)param;
int code, id;
......@@ -44,6 +57,9 @@ void *takeRefActions(void *param) {
usleep(id % 5 + 1);
taosReleaseRef(pSpace->refId, pSpace->p[id]);
}
id = random() % pSpace->refNum;
iterateRefs(id);
}
for (int i=0; i < pSpace->refNum; ++i) {
......@@ -63,7 +79,7 @@ void *openRefSpace(void *param) {
SRefSpace *pSpace = (SRefSpace *)param;
printf("c");
pSpace->refId = taosOpenRef(10000, myfree);
pSpace->refId = taosOpenRef(50, myfree);
if (pSpace->refId < 0) {
printf("failed to open ref, reson:%s\n", tstrerror(pSpace->refId));
......
......@@ -154,6 +154,7 @@ python3 ./test.py -f query/queryConnection.py
python3 ./test.py -f query/queryCountCSVData.py
python3 ./test.py -f query/natualInterval.py
python3 ./test.py -f query/bug1471.py
python3 ./test.py -f query/dataLossTest.py
#stream
python3 ./test.py -f stream/metric_1.py
......
###################################################################
# Copyright (c) 2016 by TAOS Technologies, Inc.
# All rights reserved.
#
# This file is proprietary and confidential to TAOS Technologies.
# No part of this file may be reproduced, stored, transmitted,
# disclosed or used in any form or by any means other than as
# expressly provided by the written permission from Jianhui Tao
#
###################################################################
# -*- coding: utf-8 -*-
import sys
import taos
import os
from util.log import *
from util.cases import *
from util.sql import *
from util.dnodes import *
import inspect
class TDTestCase:
def init(self, conn, logSql):
tdLog.debug("start to execute %s" % __file__)
tdSql.init(conn.cursor())
self.numberOfTables = 240
self.numberOfRecords = 10000
def run(self):
tdSql.prepare()
os.system("yes | taosdemo -t %d -n %d" % (self.numberOfTables, self.numberOfRecords))
print("==============step1")
tdSql.execute("use test")
sql = "select count(*) from meters"
tdSql.query(sql)
rows = tdSql.getData(0, 0)
print ("number of records: %d" % rows)
newRows = rows
for i in range(10000):
print("kill taosd")
time.sleep(10)
os.system("sudo kill -9 $(pgrep taosd)")
tdDnodes.startWithoutSleep(1)
while True:
try:
tdSql.query(sql)
newRows = tdSql.getData(0, 0)
print("numer of records after kill taosd %d" % newRows)
time.sleep(10)
break
except Exception as e:
pass
continue
if newRows < rows:
caller = inspect.getframeinfo(inspect.stack()[1][0])
args = (caller.filename, caller.lineno, sql, newRows, rows)
tdLog.exit("%s(%d) failed: sql:%s, queryRows:%d != expect:%d" % args)
break
tdSql.query(sql)
tdSql.checkData(0, 0, rows)
def stop(self):
tdSql.close()
tdLog.success("%s successfully executed" % __file__)
tdCases.addWindows(__file__, TDTestCase())
tdCases.addLinux(__file__, TDTestCase())
......@@ -15,6 +15,7 @@ import sys
import os
import os.path
import subprocess
from time import sleep
from util.log import *
......@@ -210,6 +211,7 @@ class TDDnode:
(self.index, self.cfgPath))
def getBuildPath(self):
buildPath = ""
selfPath = os.path.dirname(os.path.realpath(__file__))
if ("community" in selfPath):
......@@ -256,6 +258,35 @@ class TDDnode:
tdLog.debug("wait 5 seconds for the dnode:%d to start." % (self.index))
time.sleep(5)
def startWithoutSleep(self):
buildPath = self.getBuildPath()
if (buildPath == ""):
tdLog.exit("taosd not found!")
else:
tdLog.info("taosd found in %s" % buildPath)
binPath = buildPath + "/build/bin/taosd"
if self.deployed == 0:
tdLog.exit("dnode:%d is not deployed" % (self.index))
if self.valgrind == 0:
cmd = "nohup %s -c %s > /dev/null 2>&1 & " % (
binPath, self.cfgDir)
else:
valgrindCmdline = "valgrind --tool=memcheck --leak-check=full --show-reachable=no --track-origins=yes --show-leak-kinds=all -v --workaround-gcc296-bugs=yes"
cmd = "nohup %s %s -c %s 2>&1 & " % (
valgrindCmdline, binPath, self.cfgDir)
print(cmd)
if os.system(cmd) != 0:
tdLog.exit(cmd)
self.running = 1
tdLog.debug("dnode:%d is running with %s " % (self.index, cmd))
def stop(self):
if self.valgrind == 0:
......@@ -425,6 +456,10 @@ class TDDnodes:
def start(self, index):
self.check(index)
self.dnodes[index - 1].start()
def startWithoutSleep(self, index):
self.check(index)
self.dnodes[index - 1].startWithoutSleep()
def stop(self, index):
self.check(index)
......
......@@ -25,7 +25,7 @@ class TDSql:
self.queryCols = 0
self.affectedRows = 0
def init(self, cursor, log=True):
def init(self, cursor, log=False):
self.cursor = cursor
if (log):
......
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