Merge remote-tracking branch 'origin/3.0' into feature/3.0_wxy

example/src/tstream.c

@@ -63,7 +63,7 @@ int32_t init_env() {
 }
 
 int32_t create_stream() {
-  printf("create topic\n");
+  printf("create stream\n");
   TAOS_RES* pRes;
   TAOS*     pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   if (pConn == NULL) {
@@ -77,7 +77,9 @@ int32_t create_stream() {
   }
   taos_free_result(pRes);
 
-  const char* sql = "select ts,k from tu1";
+  /*const char* sql = "select min(k), max(k), sum(k) from tu1";*/
+  const char* sql = "select min(k), max(k), sum(k) as sum_of_k from st1";
+  /*const char* sql = "select sum(k) from tu1 interval(10m)";*/
   pRes = tmq_create_stream(pConn, "stream1", "out1", sql);
   if (taos_errno(pRes) != 0) {
     printf("failed to create stream out1, reason:%s\n", taos_errstr(pRes));

include/common/taosdef.h

@@ -63,9 +63,12 @@ typedef enum {
 } ETsdbStatisStatus;
 
 typedef enum {
-  TSDB_SMA_STAT_OK = 0,       // ready to provide service
-  TSDB_SMA_STAT_EXPIRED = 1,  // not ready or expired
-} ETsdbSmaStat;
+  TSDB_SMA_STAT_UNKNOWN = -1,  // unknown
+  TSDB_SMA_STAT_OK = 0,        // ready to provide service
+  TSDB_SMA_STAT_EXPIRED = 1,   // not ready or expired
+  TSDB_SMA_STAT_DROPPED = 2,   // sma dropped
+} ETsdbSmaStat;                // bit operation
+
 
 typedef enum {
   TSDB_SMA_TYPE_BLOCK = 0,       // Block-wise SMA
@@ -75,6 +78,8 @@ typedef enum {
 
 extern char *qtypeStr[];
 
+#define TSDB_PORT_DNODEDNODE 5
+#define TSDB_PORT_SYNC      10
 #define TSDB_PORT_HTTP      11
 
 #ifdef __cplusplus

include/common/tmsg.h

@@ -23,8 +23,8 @@
 #include "tencode.h"
 #include "thash.h"
 #include "tlist.h"
-#include "trow.h"
 #include "tname.h"
+#include "trow.h"
 #include "tuuid.h"
 
 #ifdef __cplusplus
@@ -109,6 +109,7 @@ typedef enum _mgmt_table {
   TSDB_MGMT_TABLE_STREAMTABLES,
   TSDB_MGMT_TABLE_TP,
   TSDB_MGMT_TABLE_FUNC,
+  TSDB_MGMT_TABLE_INDEX,
   TSDB_MGMT_TABLE_MAX,
 } EShowType;
 
@@ -268,11 +269,15 @@ typedef struct SSchema {
 typedef struct {
   char    name[TSDB_TABLE_FNAME_LEN];
   int8_t  igExists;
+  float   xFilesFactor;
+  int32_t aggregationMethod;
+  int32_t delay;
   int32_t numOfColumns;
   int32_t numOfTags;
+  int32_t commentLen;
   SArray* pColumns;
   SArray* pTags;
-  char    comment[TSDB_STB_COMMENT_LEN];
+  char*   comment;
 } SMCreateStbReq;
 
 int32_t tSerializeSMCreateStbReq(void* buf, int32_t bufLen, SMCreateStbReq* pReq);
@@ -462,7 +467,8 @@ typedef struct {
   int32_t tz;  // query client timezone
   char    intervalUnit;
   char    slidingUnit;
-  char    offsetUnit;
+  char    offsetUnit;   // TODO Remove it, the offset is the number of precision tickle, and it must be a immutable duration.
+  int8_t  precision;
   int64_t interval;
   int64_t sliding;
   int64_t offset;
@@ -472,10 +478,9 @@ typedef struct {
   int32_t code;
 } SQueryTableRsp;
 
-int32_t tSerializeSQueryTableRsp(void *buf, int32_t bufLen, SQueryTableRsp *pRsp);
-
-int32_t tDeserializeSQueryTableRsp(void *buf, int32_t bufLen, SQueryTableRsp *pRsp);
+int32_t tSerializeSQueryTableRsp(void* buf, int32_t bufLen, SQueryTableRsp* pRsp);
 
+int32_t tDeserializeSQueryTableRsp(void* buf, int32_t bufLen, SQueryTableRsp* pRsp);
 
 typedef struct {
   char    db[TSDB_DB_FNAME_LEN];
@@ -888,14 +893,14 @@ typedef struct {
 } SRetrieveTableRsp;
 
 typedef struct {
-  int64_t  handle;
-  int64_t  useconds;
-  int8_t   completed;  // all results are returned to client
-  int8_t   precision;
-  int8_t   compressed;
-  int32_t  compLen;
-  int32_t  numOfRows;
-  char     data[];
+  int64_t handle;
+  int64_t useconds;
+  int8_t  completed;  // all results are returned to client
+  int8_t  precision;
+  int8_t  compressed;
+  int32_t compLen;
+  int32_t numOfRows;
+  char    data[];
 } SRetrieveMetaTableRsp;
 
 typedef struct {
@@ -1425,12 +1430,11 @@ int32_t tSerializeSVCreateTbBatchReq(void** buf, SVCreateTbBatchReq* pReq);
 void*   tDeserializeSVCreateTbBatchReq(void* buf, SVCreateTbBatchReq* pReq);
 
 typedef struct {
-  SArray* rspList; // SArray<SVCreateTbRsp>
+  SArray* rspList;  // SArray<SVCreateTbRsp>
 } SVCreateTbBatchRsp;
 
-int32_t tSerializeSVCreateTbBatchRsp(void *buf, int32_t bufLen, SVCreateTbBatchRsp *pRsp);
-int32_t tDeserializeSVCreateTbBatchRsp(void *buf, int32_t bufLen, SVCreateTbBatchRsp *pRsp);
-
+int32_t tSerializeSVCreateTbBatchRsp(void* buf, int32_t bufLen, SVCreateTbBatchRsp* pRsp);
+int32_t tDeserializeSVCreateTbBatchRsp(void* buf, int32_t bufLen, SVCreateTbBatchRsp* pRsp);
 
 typedef struct {
   int64_t  ver;
@@ -1940,12 +1944,47 @@ static FORCE_INLINE void* tDecodeSSchemaWrapper(void* buf, SSchemaWrapper* pSW)
   }
   return buf;
 }
+
+typedef struct {
+  char    name[TSDB_TABLE_FNAME_LEN];
+  char    stb[TSDB_TABLE_FNAME_LEN];
+  int8_t  igExists;
+  int8_t  intervalUnit;
+  int8_t  slidingUnit;
+  int8_t  timezone;
+  int32_t dstVgId;  // for stream
+  int64_t interval;
+  int64_t offset;
+  int64_t sliding;
+  int32_t exprLen;        // strlen + 1
+  int32_t tagsFilterLen;  // strlen + 1
+  int32_t sqlLen;         // strlen + 1
+  int32_t astLen;         // strlen + 1
+  char*   expr;  
+  char*   tagsFilter;
+  char*   sql;
+  char*   ast;
+} SMCreateSmaReq;
+
+int32_t tSerializeSMCreateSmaReq(void* buf, int32_t bufLen, SMCreateSmaReq* pReq);
+int32_t tDeserializeSMCreateSmaReq(void* buf, int32_t bufLen, SMCreateSmaReq* pReq);
+void    tFreeSMCreateSmaReq(SMCreateSmaReq* pReq);
+
+typedef struct {
+  char   name[TSDB_TABLE_FNAME_LEN];
+  int8_t igNotExists;
+} SMDropSmaReq;
+
+int32_t tSerializeSMDropSmaReq(void* buf, int32_t bufLen, SMDropSmaReq* pReq);
+int32_t tDeserializeSMDropSmaReq(void* buf, int32_t bufLen, SMDropSmaReq* pReq);
+
 typedef struct {
   int8_t   version;       // for compatibility(default 0)
   int8_t   intervalUnit;  // MACRO: TIME_UNIT_XXX
   int8_t   slidingUnit;   // MACRO: TIME_UNIT_XXX
+  int8_t   timezoneInt;      // sma data expired if timezone changes.
   char     indexName[TSDB_INDEX_NAME_LEN];
-  char     timezone[TD_TIMEZONE_LEN];  // sma data expired if timezone changes.
+  char     timezone[TD_TIMEZONE_LEN];  
   int32_t  exprLen;
   int32_t  tagsFilterLen;
   int64_t  indexUid;
@@ -2052,8 +2091,8 @@ static FORCE_INLINE int32_t tEncodeTSma(void** buf, const STSma* pSma) {
   tlen += taosEncodeFixedI8(buf, pSma->version);
   tlen += taosEncodeFixedI8(buf, pSma->intervalUnit);
   tlen += taosEncodeFixedI8(buf, pSma->slidingUnit);
+  tlen += taosEncodeFixedI8(buf, pSma->timezoneInt);
   tlen += taosEncodeString(buf, pSma->indexName);
-  tlen += taosEncodeString(buf, pSma->timezone);
   tlen += taosEncodeFixedI32(buf, pSma->exprLen);
   tlen += taosEncodeFixedI32(buf, pSma->tagsFilterLen);
   tlen += taosEncodeFixedI64(buf, pSma->indexUid);
@@ -2087,8 +2126,8 @@ static FORCE_INLINE void* tDecodeTSma(void* buf, STSma* pSma) {
   buf = taosDecodeFixedI8(buf, &pSma->version);
   buf = taosDecodeFixedI8(buf, &pSma->intervalUnit);
   buf = taosDecodeFixedI8(buf, &pSma->slidingUnit);
+  buf = taosDecodeFixedI8(buf, &pSma->timezoneInt);
   buf = taosDecodeStringTo(buf, pSma->indexName);
-  buf = taosDecodeStringTo(buf, pSma->timezone);
   buf = taosDecodeFixedI32(buf, &pSma->exprLen);
   buf = taosDecodeFixedI32(buf, &pSma->tagsFilterLen);
   buf = taosDecodeFixedI64(buf, &pSma->indexUid);
@@ -2312,6 +2351,11 @@ enum {
   STREAM_TASK_STATUS__STOP,
 };
 
+enum {
+  STREAM_NEXT_OP_DST__VND = 1,
+  STREAM_NEXT_OP_DST__SND,
+};
+
 typedef struct {
   void* inputHandle;
   void* executor;
@@ -2326,6 +2370,7 @@ typedef struct {
   int8_t  pipeSink;
   int8_t  numOfRunners;
   int8_t  parallelizable;
+  int8_t  nextOpDst;  // vnode or snode
   SEpSet  NextOpEp;
   char*   qmsg;
   // not applied to encoder and decoder
@@ -2341,6 +2386,8 @@ static FORCE_INLINE SStreamTask* streamTaskNew(int64_t streamId, int32_t level)
     return NULL;
   }
   pTask->taskId = tGenIdPI32();
+  pTask->streamId = streamId;
+  pTask->level = level;
   pTask->status = STREAM_TASK_STATUS__RUNNING;
   pTask->qmsg = NULL;
   return pTask;
@@ -2368,6 +2415,13 @@ typedef struct {
   int32_t reserved;
 } SStreamTaskExecRsp;
 
+typedef struct {
+  SMsgHead head;
+  int64_t  streamId;
+  int64_t  version;
+  SArray*  res;  // SArray<SSDataBlock>
+} SStreamSmaSinkReq;
+
 #pragma pack(pop)
 
 #ifdef __cplusplus

include/common/tmsgdef.h

@@ -127,6 +127,8 @@ enum {
   TD_DEF_MSG_TYPE(TDMT_MND_CREATE_STB, "mnode-create-stb", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_MND_ALTER_STB, "mnode-alter-stb", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_MND_DROP_STB, "mnode-drop-stb", NULL, NULL)
+  TD_DEF_MSG_TYPE(TDMT_MND_CREATE_SMA, "mnode-create-sma", NULL, NULL)
+  TD_DEF_MSG_TYPE(TDMT_MND_DROP_SMA, "mnode-drop-sma", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_MND_TABLE_META, "mnode-table-meta", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_MND_VGROUP_LIST, "mnode-vgroup-list", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_MND_QNODE_LIST, "mnode-qnode-list", NULL, NULL)
@@ -191,6 +193,7 @@ enum {
   TD_DEF_MSG_TYPE(TDMT_VND_CONSUME, "vnode-consume", SMqCVConsumeReq, SMqCVConsumeRsp)
   TD_DEF_MSG_TYPE(TDMT_VND_TASK_DEPLOY, "vnode-task-deploy", SStreamTaskDeployReq, SStreamTaskDeployRsp)
   TD_DEF_MSG_TYPE(TDMT_VND_TASK_EXEC, "vnode-task-exec", SStreamTaskExecReq, SStreamTaskExecRsp)
+  TD_DEF_MSG_TYPE(TDMT_VND_STREAM_TRIGGER, "vnode-stream-trigger", NULL, NULL)
 
   TD_DEF_MSG_TYPE(TDMT_VND_CREATE_SMA, "vnode-create-sma", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_CANCEL_SMA, "vnode-cancel-sma", NULL, NULL)

include/common/trow.h

@@ -166,6 +166,7 @@ typedef struct {
 #define TD_ROW_HEAD_LEN  (sizeof(STSRow))
 #define TD_ROW_NCOLS_LEN (sizeof(col_id_t))
 
+#define TD_ROW_INFO(r)     ((r)->info)
 #define TD_ROW_TYPE(r)     ((r)->type)
 #define TD_ROW_DELETE(r)   ((r)->del)
 #define TD_ROW_ENDIAN(r)   ((r)->endian)
@@ -180,6 +181,7 @@ typedef struct {
 // (int32_t)ceil((double)nCols/TD_VTYPE_PARTS) should be added if TD_SUPPORT_BITMAP defined.
 #define TD_ROW_MAX_BYTES_FROM_SCHEMA(s) (schemaTLen(s) + TD_ROW_HEAD_LEN)
 
+#define TD_ROW_SET_INFO(r, i)  (TD_ROW_INFO(r) = (i))
 #define TD_ROW_SET_TYPE(r, t)  (TD_ROW_TYPE(r) = (t))
 #define TD_ROW_SET_DELETE(r)   (TD_ROW_DELETE(r) = 1)
 #define TD_ROW_SET_SVER(r, v)  (TD_ROW_SVER(r) = (v))
@@ -473,6 +475,7 @@ static int32_t tdSRowResetBuf(SRowBuilder *pBuilder, void *pBuf) {
     return terrno;
   }
 
+  TD_ROW_SET_INFO(pBuilder->pBuf, 0);
   TD_ROW_SET_TYPE(pBuilder->pBuf, pBuilder->rowType);
 
   uint32_t len = 0;

include/dnode/mnode/sdb/sdb.h

@@ -119,10 +119,11 @@ typedef enum {
   SDB_CONSUMER = 13,
   SDB_TOPIC = 14,
   SDB_VGROUP = 15,
-  SDB_STB = 16,
-  SDB_DB = 17,
-  SDB_FUNC = 18,
-  SDB_MAX = 19
+  SDB_SMA = 16,
+  SDB_STB = 17,
+  SDB_DB = 18,
+  SDB_FUNC = 19,
+  SDB_MAX = 20
 } ESdbType;
 
 typedef struct SSdb SSdb;

include/libs/sync/sync.h

@@ -35,6 +35,7 @@ typedef enum {
   TAOS_SYNC_STATE_FOLLOWER = 100,
   TAOS_SYNC_STATE_CANDIDATE = 101,
   TAOS_SYNC_STATE_LEADER = 102,
+  TAOS_SYNC_STATE_ERROR = 103,
 } ESyncState;
 
 typedef struct SSyncBuffer {
@@ -68,17 +69,20 @@ typedef struct SSnapshot {
 typedef struct SSyncFSM {
   void* data;
 
-  // when value in pBuf finish a raft flow, FpCommitCb is called, code indicates the result
+  // when value in pMsg finish a raft flow, FpCommitCb is called, code indicates the result
   // user can do something according to the code and isWeak. for example, write data into tsdb
-  void (*FpCommitCb)(struct SSyncFSM* pFsm, const SRpcMsg* pBuf, SyncIndex index, bool isWeak, int32_t code);
+  void (*FpCommitCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SyncIndex index, bool isWeak, int32_t code,
+                     ESyncState state);
 
-  // when value in pBuf has been written into local log store, FpPreCommitCb is called, code indicates the result
+  // when value in pMsg has been written into local log store, FpPreCommitCb is called, code indicates the result
   // user can do something according to the code and isWeak. for example, write data into tsdb
-  void (*FpPreCommitCb)(struct SSyncFSM* pFsm, const SRpcMsg* pBuf, SyncIndex index, bool isWeak, int32_t code);
+  void (*FpPreCommitCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SyncIndex index, bool isWeak, int32_t code,
+                        ESyncState state);
 
   // when log entry is updated by a new one, FpRollBackCb is called
   // user can do something to roll back. for example, delete data from tsdb, or just ignore it
-  void (*FpRollBackCb)(struct SSyncFSM* pFsm, const SRpcMsg* pBuf, SyncIndex index, bool isWeak, int32_t code);
+  void (*FpRollBackCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SyncIndex index, bool isWeak, int32_t code,
+                       ESyncState state);
 
   // user should implement this function, use "data" to take snapshot into "snapshot"
   int32_t (*FpTakeSnapshot)(SSnapshot* snapshot);
@@ -157,10 +161,14 @@ void    syncCleanUp();
 int64_t    syncStart(const SSyncInfo* pSyncInfo);
 void       syncStop(int64_t rid);
 int32_t    syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg);
-int32_t    syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak);        // use this function
-int32_t    syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak);  // just for compatibility
+int32_t    syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak);
 ESyncState syncGetMyRole(int64_t rid);
-void       syncGetNodesRole(int64_t rid, SNodesRole* pNodeRole);
+
+// propose with sequence number, to implement linearizable semantics
+int32_t syncPropose2(int64_t rid, const SRpcMsg* pMsg, bool isWeak, uint64_t seqNum);
+
+// for compatibility, the same as syncPropose
+int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak);
 
 extern int32_t sDebugFlag;
 

include/os/osSocket.h

@@ -51,10 +51,31 @@
 extern "C" {
 #endif
 
-#if (defined(_TD_WINDOWS_64) || defined(_TD_WINDOWS_32)) 
+#if defined(WINDOWS)
   #define htobe64 htonll
 #endif
 
+#if defined(WINDOWS)
+  #define TAOS_EPOLL_WAIT_TIME 100
+  typedef SOCKET eventfd_t;
+  #define eventfd(a, b) -1
+  typedef SOCKET EpollFd;
+  #define EpollClose(pollFd) epoll_close(pollFd)
+  #ifndef EPOLLWAKEUP
+    #define EPOLLWAKEUP (1u << 29)
+  #endif
+#elif defined(_TD_DARWIN_64)
+  #define TAOS_EPOLL_WAIT_TIME 500
+  typedef int32_t SOCKET;
+  typedef SOCKET EpollFd;
+  #define EpollClose(pollFd) epoll_close(pollFd)
+#else
+  #define TAOS_EPOLL_WAIT_TIME 500
+  typedef int32_t SOCKET;
+  typedef SOCKET EpollFd;
+  #define EpollClose(pollFd) taosCloseSocket(pollFd)
+#endif
+
 #if defined(_TD_DARWIN_64)
 //  #define htobe64 htonll
 
@@ -64,12 +85,12 @@ extern "C" {
 #	define htole16(x) OSSwapHostToLittleInt16(x)
 #	define be16toh(x) OSSwapBigToHostInt16(x)
 #	define le16toh(x) OSSwapLittleToHostInt16(x)
- 
+
 #	define htobe32(x) OSSwapHostToBigInt32(x)
 #	define htole32(x) OSSwapHostToLittleInt32(x)
 #	define be32toh(x) OSSwapBigToHostInt32(x)
 #	define le32toh(x) OSSwapLittleToHostInt32(x)
- 
+
 #	define htobe64(x) OSSwapHostToBigInt64(x)
 #	define htole64(x) OSSwapHostToLittleInt64(x)
 #	define be64toh(x) OSSwapBigToHostInt64(x)
@@ -83,6 +104,17 @@ extern "C" {
 
 #define TAOS_EPOLL_WAIT_TIME 500
 
+typedef int32_t SocketFd;
+typedef SocketFd  EpollFd;
+
+typedef struct TdSocket {
+#if SOCKET_WITH_LOCK
+  TdThreadRwlock rwlock;
+#endif
+  int      refId;
+  SocketFd fd;
+} *TdSocketPtr, TdSocket;
+
 typedef struct TdSocketServer *TdSocketServerPtr;
 typedef struct TdSocket *TdSocketPtr;
 typedef struct TdEpoll *TdEpollPtr;
@@ -91,6 +123,7 @@ int32_t taosSendto(TdSocketPtr pSocket, void * msg, int len, unsigned int flags,
 int32_t taosWriteSocket(TdSocketPtr pSocket, void *msg, int len);
 int32_t taosReadSocket(TdSocketPtr pSocket, void *msg, int len);
 int32_t taosReadFromSocket(TdSocketPtr pSocket, void *buf, int32_t len, int32_t flags, struct sockaddr *destAddr, socklen_t *addrLen);
+int32_t taosCloseSocketNoCheck1(SocketFd fd);
 int32_t taosCloseSocket(TdSocketPtr *ppSocket);
 int32_t taosCloseSocketServer(TdSocketServerPtr *ppSocketServer);
 int32_t taosShutDownSocketRD(TdSocketPtr pSocket);

include/util/taoserror.h

@@ -274,18 +274,23 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_MND_STREAM_NOT_EXIST          TAOS_DEF_ERROR_CODE(0, 0x03F1)
 #define TSDB_CODE_MND_INVALID_STREAM_OPTION     TAOS_DEF_ERROR_CODE(0, 0x03F2)
 
+// mnode-sma
+#define TSDB_CODE_MND_SMA_ALREADY_EXIST         TAOS_DEF_ERROR_CODE(0, 0x0400)
+#define TSDB_CODE_MND_SMA_NOT_EXIST             TAOS_DEF_ERROR_CODE(0, 0x0401)
+#define TSDB_CODE_MND_INVALID_SMA_OPTION        TAOS_DEF_ERROR_CODE(0, 0x0402)
+
 // dnode
-#define TSDB_CODE_DND_ACTION_IN_PROGRESS        TAOS_DEF_ERROR_CODE(0, 0x0400)
-#define TSDB_CODE_DND_OFFLINE                   TAOS_DEF_ERROR_CODE(0, 0x0401)
-#define TSDB_CODE_DND_INVALID_MSG_LEN           TAOS_DEF_ERROR_CODE(0, 0x0402)
-#define TSDB_CODE_NODE_ALREADY_DEPLOYED         TAOS_DEF_ERROR_CODE(0, 0x0403)
-#define TSDB_CODE_NODE_NOT_DEPLOYED             TAOS_DEF_ERROR_CODE(0, 0x0404)
-#define TSDB_CODE_NODE_PARSE_FILE_ERROR         TAOS_DEF_ERROR_CODE(0, 0x0405)
-#define TSDB_CODE_NODE_INVALID_OPTION           TAOS_DEF_ERROR_CODE(0, 0x0406)
-#define TSDB_CODE_DND_VNODE_ALREADY_DEPLOYED    TAOS_DEF_ERROR_CODE(0, 0x0410)
-#define TSDB_CODE_DND_VNODE_NOT_DEPLOYED        TAOS_DEF_ERROR_CODE(0, 0x0411)
-#define TSDB_CODE_DND_VNODE_INVALID_OPTION      TAOS_DEF_ERROR_CODE(0, 0x0412)
-#define TSDB_CODE_DND_VNODE_TOO_MANY_VNODES     TAOS_DEF_ERROR_CODE(0, 0x0413)
+#define TSDB_CODE_DND_ACTION_IN_PROGRESS        TAOS_DEF_ERROR_CODE(0, 0x04A0)
+#define TSDB_CODE_DND_OFFLINE                   TAOS_DEF_ERROR_CODE(0, 0x04A1)
+#define TSDB_CODE_DND_INVALID_MSG_LEN           TAOS_DEF_ERROR_CODE(0, 0x04A2)
+#define TSDB_CODE_NODE_ALREADY_DEPLOYED         TAOS_DEF_ERROR_CODE(0, 0x04A3)
+#define TSDB_CODE_NODE_NOT_DEPLOYED             TAOS_DEF_ERROR_CODE(0, 0x04A4)
+#define TSDB_CODE_NODE_PARSE_FILE_ERROR         TAOS_DEF_ERROR_CODE(0, 0x04A5)
+#define TSDB_CODE_NODE_INVALID_OPTION           TAOS_DEF_ERROR_CODE(0, 0x04A6)
+#define TSDB_CODE_DND_VNODE_ALREADY_DEPLOYED    TAOS_DEF_ERROR_CODE(0, 0x04A7)
+#define TSDB_CODE_DND_VNODE_NOT_DEPLOYED        TAOS_DEF_ERROR_CODE(0, 0x04A8)
+#define TSDB_CODE_DND_VNODE_INVALID_OPTION      TAOS_DEF_ERROR_CODE(0, 0x04A9)
+#define TSDB_CODE_DND_VNODE_TOO_MANY_VNODES     TAOS_DEF_ERROR_CODE(0, 0x04AA)
 
 // vnode
 #define TSDB_CODE_VND_ACTION_IN_PROGRESS        TAOS_DEF_ERROR_CODE(0, 0x0500)
@@ -309,7 +314,8 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_VND_IS_SYNCING                TAOS_DEF_ERROR_CODE(0, 0x0513)
 #define TSDB_CODE_VND_INVALID_TSDB_STATE        TAOS_DEF_ERROR_CODE(0, 0x0514)
 #define TSDB_CODE_VND_TB_NOT_EXIST              TAOS_DEF_ERROR_CODE(0, 0x0515)
-#define TSDB_CODE_VND_HASH_MISMATCH             TAOS_DEF_ERROR_CODE(0, 0x0516)
+#define TSDB_CODE_VND_SMA_NOT_EXIST             TAOS_DEF_ERROR_CODE(0, 0x0516)
+#define TSDB_CODE_VND_HASH_MISMATCH             TAOS_DEF_ERROR_CODE(0, 0x0517)
 
 // tsdb
 #define TSDB_CODE_TDB_INVALID_TABLE_ID          TAOS_DEF_ERROR_CODE(0, 0x0600)
@@ -336,8 +342,9 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_TDB_IVLD_TAG_VAL              TAOS_DEF_ERROR_CODE(0, 0x0615)
 #define TSDB_CODE_TDB_NO_CACHE_LAST_ROW         TAOS_DEF_ERROR_CODE(0, 0x0616)
 #define TSDB_CODE_TDB_TABLE_RECREATED           TAOS_DEF_ERROR_CODE(0, 0x0617)
-#define TSDB_CODE_TDB_NO_SMA_INDEX_IN_META      TAOS_DEF_ERROR_CODE(0, 0x0618)
-#define TSDB_CODE_TDB_TDB_ENV_OPEN_ERROR        TAOS_DEF_ERROR_CODE(0, 0x0619)
+#define TSDB_CODE_TDB_TDB_ENV_OPEN_ERROR        TAOS_DEF_ERROR_CODE(0, 0x0618)
+#define TSDB_CODE_TDB_NO_SMA_INDEX_IN_META      TAOS_DEF_ERROR_CODE(0, 0x0619)
+#define TSDB_CODE_TDB_INVALID_SMA_STAT          TAOS_DEF_ERROR_CODE(0, 0x0620)
 
 // query
 #define TSDB_CODE_QRY_INVALID_QHANDLE           TAOS_DEF_ERROR_CODE(0, 0x0700)

source/common/src/tdataformat.c

@@ -421,6 +421,7 @@ SDataCols *tdFreeDataCols(SDataCols *pCols) {
   return NULL;
 }
 
+#if 0
 SDataCols *tdDupDataCols(SDataCols *pDataCols, bool keepData) {
   SDataCols *pRet = tdNewDataCols(pDataCols->maxCols, pDataCols->maxPoints);
   if (pRet == NULL) return NULL;
@@ -454,6 +455,7 @@ SDataCols *tdDupDataCols(SDataCols *pDataCols, bool keepData) {
 
   return pRet;
 }
+#endif
 
 void tdResetDataCols(SDataCols *pCols) {
   if (pCols != NULL) {

source/common/src/tglobal.c

@@ -132,6 +132,9 @@ bool tsdbForceKeepFile = false;
 int32_t  tsDiskCfgNum = 0;
 SDiskCfg tsDiskCfg[TFS_MAX_DISKS] = {0};
 
+// stream scheduler
+bool tsStreamSchedV = true;
+
 /*
  * minimum scale for whole system, millisecond by default
  * for TSDB_TIME_PRECISION_MILLI: 86400000L
@@ -585,4 +588,4 @@ void taosCfgDynamicOptions(const char *option, const char *value) {
     taosResetLog();
     cfgDumpCfg(tsCfg, 1, false);
   }
-}
+}
\ No newline at end of file

source/common/src/tmsg.c

@@ -314,12 +314,12 @@ int32_t tSerializeSVCreateTbReq(void **buf, SVCreateTbReq *pReq) {
       for (col_id_t i = 0; i < pReq->stbCfg.nBSmaCols; ++i) {
         tlen += taosEncodeFixedI16(buf, pReq->stbCfg.pBSmaCols[i]);
       }
-      if(pReq->rollup && NULL != pReq->stbCfg.pRSmaParam) {
+      if(pReq->rollup && pReq->stbCfg.pRSmaParam) {
         SRSmaParam *param = pReq->stbCfg.pRSmaParam;
         tlen += taosEncodeFixedU32(buf, (uint32_t)param->xFilesFactor);
         tlen += taosEncodeFixedI8(buf, param->delayUnit);
         tlen += taosEncodeFixedI8(buf, param->nFuncIds);
-        for(int8_t i=0; i< param->nFuncIds; ++i) {
+        for (int8_t i = 0; i < param->nFuncIds; ++i) {
           tlen += taosEncodeFixedI32(buf, param->pFuncIds[i]);
         }
         tlen += taosEncodeFixedI64(buf, param->delay);
@@ -337,16 +337,16 @@ int32_t tSerializeSVCreateTbReq(void **buf, SVCreateTbReq *pReq) {
         tlen += taosEncodeFixedI32(buf, pReq->ntbCfg.pSchema[i].bytes);
         tlen += taosEncodeString(buf, pReq->ntbCfg.pSchema[i].name);
       }
-      tlen += taosEncodeFixedI16(buf, pReq->stbCfg.nBSmaCols);
-      for (col_id_t i = 0; i < pReq->stbCfg.nBSmaCols; ++i) {
-        tlen += taosEncodeFixedI16(buf, pReq->stbCfg.pBSmaCols[i]);
+      tlen += taosEncodeFixedI16(buf, pReq->ntbCfg.nBSmaCols);
+      for (col_id_t i = 0; i < pReq->ntbCfg.nBSmaCols; ++i) {
+        tlen += taosEncodeFixedI16(buf, pReq->ntbCfg.pBSmaCols[i]);
       }
-      if(pReq->rollup && NULL != pReq->stbCfg.pRSmaParam) {
-        SRSmaParam *param = pReq->stbCfg.pRSmaParam;
+      if(pReq->rollup && pReq->ntbCfg.pRSmaParam) {
+        SRSmaParam *param = pReq->ntbCfg.pRSmaParam;
         tlen += taosEncodeFixedU32(buf, (uint32_t)param->xFilesFactor);
         tlen += taosEncodeFixedI8(buf, param->delayUnit);
         tlen += taosEncodeFixedI8(buf, param->nFuncIds);
-        for(int8_t i=0; i< param->nFuncIds; ++i) {
+        for (int8_t i = 0; i < param->nFuncIds; ++i) {
           tlen += taosEncodeFixedI32(buf, param->pFuncIds[i]);
         }
         tlen += taosEncodeFixedI64(buf, param->delay);
@@ -387,7 +387,7 @@ void *tDeserializeSVCreateTbReq(void *buf, SVCreateTbReq *pReq) {
         buf = taosDecodeStringTo(buf, pReq->stbCfg.pTagSchema[i].name);
       }
       buf = taosDecodeFixedI16(buf, &(pReq->stbCfg.nBSmaCols));
-      if(pReq->stbCfg.nBSmaCols > 0) {
+      if (pReq->stbCfg.nBSmaCols > 0) {
         pReq->stbCfg.pBSmaCols = (col_id_t *)malloc(pReq->stbCfg.nBSmaCols * sizeof(col_id_t));
         for (col_id_t i = 0; i < pReq->stbCfg.nBSmaCols; ++i) {
           buf = taosDecodeFixedI16(buf, pReq->stbCfg.pBSmaCols + i);
@@ -395,14 +395,14 @@ void *tDeserializeSVCreateTbReq(void *buf, SVCreateTbReq *pReq) {
       } else {
         pReq->stbCfg.pBSmaCols = NULL;
       }
-      if(pReq->rollup) {
+      if (pReq->rollup) {
         pReq->stbCfg.pRSmaParam = (SRSmaParam *)malloc(sizeof(SRSmaParam));
         SRSmaParam *param = pReq->stbCfg.pRSmaParam;
-        buf = taosDecodeFixedU32(buf, (uint32_t*)&param->xFilesFactor);
+        buf = taosDecodeFixedU32(buf, (uint32_t *)&param->xFilesFactor);
         buf = taosDecodeFixedI8(buf, &param->delayUnit);
         buf = taosDecodeFixedI8(buf, &param->nFuncIds);
-        if(param->nFuncIds > 0) {
-          for (int8_t i = 0; i< param->nFuncIds; ++i) {
+        if (param->nFuncIds > 0) {
+          for (int8_t i = 0; i < param->nFuncIds; ++i) {
             buf = taosDecodeFixedI32(buf, param->pFuncIds + i);
           }
         } else {
@@ -426,23 +426,23 @@ void *tDeserializeSVCreateTbReq(void *buf, SVCreateTbReq *pReq) {
         buf = taosDecodeFixedI32(buf, &pReq->ntbCfg.pSchema[i].bytes);
         buf = taosDecodeStringTo(buf, pReq->ntbCfg.pSchema[i].name);
       }
-      buf = taosDecodeFixedI16(buf, &(pReq->stbCfg.nBSmaCols));
-      if(pReq->stbCfg.nBSmaCols > 0) {
-        pReq->stbCfg.pBSmaCols = (col_id_t *)malloc(pReq->stbCfg.nBSmaCols * sizeof(col_id_t));
-        for (col_id_t i = 0; i < pReq->stbCfg.nBSmaCols; ++i) {
-          buf = taosDecodeFixedI16(buf, pReq->stbCfg.pBSmaCols + i);
+      buf = taosDecodeFixedI16(buf, &(pReq->ntbCfg.nBSmaCols));
+      if(pReq->ntbCfg.nBSmaCols > 0) {
+        pReq->ntbCfg.pBSmaCols = (col_id_t *)malloc(pReq->ntbCfg.nBSmaCols * sizeof(col_id_t));
+        for (col_id_t i = 0; i < pReq->ntbCfg.nBSmaCols; ++i) {
+          buf = taosDecodeFixedI16(buf, pReq->ntbCfg.pBSmaCols + i);
         }
       } else {
-        pReq->stbCfg.pBSmaCols = NULL;
+        pReq->ntbCfg.pBSmaCols = NULL;
       }
       if(pReq->rollup) {
-        pReq->stbCfg.pRSmaParam = (SRSmaParam *)malloc(sizeof(SRSmaParam));
-        SRSmaParam *param = pReq->stbCfg.pRSmaParam;
+        pReq->ntbCfg.pRSmaParam = (SRSmaParam *)malloc(sizeof(SRSmaParam));
+        SRSmaParam *param = pReq->ntbCfg.pRSmaParam;
         buf = taosDecodeFixedU32(buf, (uint32_t*)&param->xFilesFactor);
         buf = taosDecodeFixedI8(buf, &param->delayUnit);
         buf = taosDecodeFixedI8(buf, &param->nFuncIds);
-        if(param->nFuncIds > 0) {
-          for (int8_t i = 0; i< param->nFuncIds; ++i) {
+        if (param->nFuncIds > 0) {
+          for (int8_t i = 0; i < param->nFuncIds; ++i) {
             buf = taosDecodeFixedI32(buf, param->pFuncIds + i);
           }
         } else {
@@ -450,7 +450,7 @@ void *tDeserializeSVCreateTbReq(void *buf, SVCreateTbReq *pReq) {
         }
         buf = taosDecodeFixedI64(buf, &param->delay);
       } else {
-        pReq->stbCfg.pRSmaParam = NULL;
+        pReq->ntbCfg.pRSmaParam = NULL;
       }
       break;
     default:
@@ -510,8 +510,12 @@ int32_t tSerializeSMCreateStbReq(void *buf, int32_t bufLen, SMCreateStbReq *pReq
   if (tStartEncode(&encoder) < 0) return -1;
   if (tEncodeCStr(&encoder, pReq->name) < 0) return -1;
   if (tEncodeI8(&encoder, pReq->igExists) < 0) return -1;
+  if (tEncodeFloat(&encoder, pReq->xFilesFactor) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->aggregationMethod) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->delay) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->numOfColumns) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->numOfTags) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->commentLen) < 0) return -1;
 
   for (int32_t i = 0; i < pReq->numOfColumns; ++i) {
     SField *pField = taosArrayGet(pReq->pColumns, i);
@@ -527,7 +531,7 @@ int32_t tSerializeSMCreateStbReq(void *buf, int32_t bufLen, SMCreateStbReq *pReq
     if (tEncodeCStr(&encoder, pField->name) < 0) return -1;
   }
 
-  if (tEncodeCStr(&encoder, pReq->comment) < 0) return -1;
+  if (tEncodeBinary(&encoder, pReq->comment, pReq->commentLen) < 0) return -1;
   tEndEncode(&encoder);
 
   int32_t tlen = encoder.pos;
@@ -542,8 +546,12 @@ int32_t tDeserializeSMCreateStbReq(void *buf, int32_t bufLen, SMCreateStbReq *pR
   if (tStartDecode(&decoder) < 0) return -1;
   if (tDecodeCStrTo(&decoder, pReq->name) < 0) return -1;
   if (tDecodeI8(&decoder, &pReq->igExists) < 0) return -1;
+  if (tDecodeFloat(&decoder, &pReq->xFilesFactor) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->aggregationMethod) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->delay) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->numOfColumns) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->numOfTags) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->commentLen) < 0) return -1;
 
   pReq->pColumns = taosArrayInit(pReq->numOfColumns, sizeof(SField));
   pReq->pTags = taosArrayInit(pReq->numOfTags, sizeof(SField));
@@ -574,6 +582,12 @@ int32_t tDeserializeSMCreateStbReq(void *buf, int32_t bufLen, SMCreateStbReq *pR
     }
   }
 
+  if (pReq->commentLen > 0) {
+    pReq->comment = malloc(pReq->commentLen);
+    if (pReq->comment == NULL) return -1;
+    if (tDecodeCStrTo(&decoder, pReq->comment) < 0) return -1;
+  }
+
   if (tDecodeCStrTo(&decoder, pReq->comment) < 0) return -1;
   tEndDecode(&decoder);
 
@@ -671,6 +685,123 @@ void tFreeSMAltertbReq(SMAltertbReq *pReq) {
   pReq->pFields = NULL;
 }
 
+int32_t tSerializeSMCreateSmaReq(void *buf, int32_t bufLen, SMCreateSmaReq *pReq) {
+  SCoder encoder = {0};
+  tCoderInit(&encoder, TD_LITTLE_ENDIAN, buf, bufLen, TD_ENCODER);
+
+  if (tStartEncode(&encoder) < 0) return -1;
+  if (tEncodeCStr(&encoder, pReq->name) < 0) return -1;
+  if (tEncodeCStr(&encoder, pReq->stb) < 0) return -1;
+  if (tEncodeI8(&encoder, pReq->igExists) < 0) return -1;
+  if (tEncodeI8(&encoder, pReq->intervalUnit) < 0) return -1;
+  if (tEncodeI8(&encoder, pReq->slidingUnit) < 0) return -1;
+  if (tEncodeI8(&encoder, pReq->timezone) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->dstVgId) < 0) return -1;
+  if (tEncodeI64(&encoder, pReq->interval) < 0) return -1;
+  if (tEncodeI64(&encoder, pReq->offset) < 0) return -1;
+  if (tEncodeI64(&encoder, pReq->sliding) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->exprLen) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->tagsFilterLen) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->sqlLen) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->astLen) < 0) return -1;
+  if (pReq->exprLen > 0) {
+    if (tEncodeBinary(&encoder, pReq->expr, pReq->exprLen) < 0) return -1;
+  }
+  if (pReq->tagsFilterLen > 0) {
+    if (tEncodeBinary(&encoder, pReq->tagsFilter, pReq->tagsFilterLen) < 0) return -1;
+  }
+  if (pReq->sqlLen > 0) {
+    if (tEncodeBinary(&encoder, pReq->sql, pReq->sqlLen) < 0) return -1;
+  }
+  if (pReq->astLen > 0) {
+    if (tEncodeBinary(&encoder, pReq->ast, pReq->astLen) < 0) return -1;
+  }
+  tEndEncode(&encoder);
+
+  int32_t tlen = encoder.pos;
+  tCoderClear(&encoder);
+  return tlen;
+}
+
+int32_t tDeserializeSMCreateSmaReq(void *buf, int32_t bufLen, SMCreateSmaReq *pReq) {
+  SCoder decoder = {0};
+  tCoderInit(&decoder, TD_LITTLE_ENDIAN, buf, bufLen, TD_DECODER);
+
+  if (tStartDecode(&decoder) < 0) return -1;
+  if (tDecodeCStrTo(&decoder, pReq->name) < 0) return -1;
+  if (tDecodeCStrTo(&decoder, pReq->stb) < 0) return -1;
+  if (tDecodeI8(&decoder, &pReq->igExists) < 0) return -1;
+  if (tDecodeI8(&decoder, &pReq->intervalUnit) < 0) return -1;
+  if (tDecodeI8(&decoder, &pReq->slidingUnit) < 0) return -1;
+  if (tDecodeI8(&decoder, &pReq->timezone) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->dstVgId) < 0) return -1;
+  if (tDecodeI64(&decoder, &pReq->interval) < 0) return -1;
+  if (tDecodeI64(&decoder, &pReq->offset) < 0) return -1;
+  if (tDecodeI64(&decoder, &pReq->sliding) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->exprLen) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->tagsFilterLen) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->sqlLen) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->astLen) < 0) return -1;
+  if (pReq->exprLen > 0) {
+    pReq->expr = malloc(pReq->exprLen);
+    if (pReq->expr == NULL) return -1;
+    if (tDecodeCStrTo(&decoder, pReq->expr) < 0) return -1;
+  }
+  if (pReq->tagsFilterLen > 0) {
+    pReq->tagsFilter = malloc(pReq->tagsFilterLen);
+    if (pReq->tagsFilter == NULL) return -1;
+    if (tDecodeCStrTo(&decoder, pReq->tagsFilter) < 0) return -1;
+  }
+  if (pReq->sqlLen > 0) {
+    pReq->sql = malloc(pReq->sqlLen);
+    if (pReq->sql == NULL) return -1;
+    if (tDecodeCStrTo(&decoder, pReq->sql) < 0) return -1;
+  }
+  if (pReq->astLen > 0) {
+    pReq->ast = malloc(pReq->astLen);
+    if (pReq->ast == NULL) return -1;
+    if (tDecodeCStrTo(&decoder, pReq->ast) < 0) return -1;
+  }
+
+  tEndDecode(&decoder);
+  tCoderClear(&decoder);
+  return 0;
+}
+
+void tFreeSMCreateSmaReq(SMCreateSmaReq *pReq) {
+  tfree(pReq->expr);
+  tfree(pReq->tagsFilter);
+  tfree(pReq->sql);
+  tfree(pReq->ast);
+}
+
+int32_t tSerializeSMDropSmaReq(void *buf, int32_t bufLen, SMDropSmaReq *pReq) {
+  SCoder encoder = {0};
+  tCoderInit(&encoder, TD_LITTLE_ENDIAN, buf, bufLen, TD_ENCODER);
+
+  if (tStartEncode(&encoder) < 0) return -1;
+  if (tEncodeCStr(&encoder, pReq->name) < 0) return -1;
+  if (tEncodeI8(&encoder, pReq->igNotExists) < 0) return -1;
+  tEndEncode(&encoder);
+
+  int32_t tlen = encoder.pos;
+  tCoderClear(&encoder);
+  return tlen;
+}
+
+int32_t tDeserializeSMDropSmaReq(void *buf, int32_t bufLen, SMDropSmaReq *pReq) {
+  SCoder decoder = {0};
+  tCoderInit(&decoder, TD_LITTLE_ENDIAN, buf, bufLen, TD_DECODER);
+
+  if (tStartDecode(&decoder) < 0) return -1;
+  if (tDecodeCStrTo(&decoder, pReq->name) < 0) return -1;
+  if (tDecodeI8(&decoder, &pReq->igNotExists) < 0) return -1;
+  tEndDecode(&decoder);
+
+  tCoderClear(&decoder);
+  return 0;
+}
+
 int32_t tSerializeSStatusReq(void *buf, int32_t bufLen, SStatusReq *pReq) {
   SCoder encoder = {0};
   tCoderInit(&encoder, TD_LITTLE_ENDIAN, buf, bufLen, TD_ENCODER);
@@ -2711,7 +2842,7 @@ int32_t tSerializeSQueryTableRsp(void *buf, int32_t bufLen, SQueryTableRsp *pRsp
   tCoderInit(&encoder, TD_LITTLE_ENDIAN, buf, bufLen, TD_ENCODER);
 
   if (tStartEncode(&encoder) < 0) return -1;
-  if (tEncodeI32(&encoder, pRsp->code) < 0) return -1;    
+  if (tEncodeI32(&encoder, pRsp->code) < 0) return -1;
   tEndEncode(&encoder);
 
   int32_t tlen = encoder.pos;
@@ -2738,13 +2869,13 @@ int32_t tSerializeSVCreateTbBatchRsp(void *buf, int32_t bufLen, SVCreateTbBatchR
   if (tStartEncode(&encoder) < 0) return -1;
   if (pRsp->rspList) {
     int32_t num = taosArrayGetSize(pRsp->rspList);
-    if (tEncodeI32(&encoder, num) < 0) return -1;    
+    if (tEncodeI32(&encoder, num) < 0) return -1;
     for (int32_t i = 0; i < num; ++i) {
       SVCreateTbRsp *rsp = taosArrayGet(pRsp->rspList, i);
-      if (tEncodeI32(&encoder, rsp->code) < 0) return -1;    
+      if (tEncodeI32(&encoder, rsp->code) < 0) return -1;
     }
   } else {
-    if (tEncodeI32(&encoder, 0) < 0) return -1;    
+    if (tEncodeI32(&encoder, 0) < 0) return -1;
   }
   tEndEncode(&encoder);
 
@@ -2754,7 +2885,7 @@ int32_t tSerializeSVCreateTbBatchRsp(void *buf, int32_t bufLen, SVCreateTbBatchR
 }
 
 int32_t tDeserializeSVCreateTbBatchRsp(void *buf, int32_t bufLen, SVCreateTbBatchRsp *pRsp) {
-  SCoder decoder = {0};
+  SCoder  decoder = {0};
   int32_t num = 0;
   tCoderInit(&decoder, TD_LITTLE_ENDIAN, buf, bufLen, TD_DECODER);
 
@@ -2777,7 +2908,6 @@ int32_t tDeserializeSVCreateTbBatchRsp(void *buf, int32_t bufLen, SVCreateTbBatc
   return 0;
 }
 
-
 int32_t tSerializeSVCreateTSmaReq(void **buf, SVCreateTSmaReq *pReq) {
   int32_t tlen = 0;
 
@@ -2826,7 +2956,8 @@ int32_t tSerializeSCMCreateStreamReq(void *buf, int32_t bufLen, const SCMCreateS
   if (tEncodeCStr(&encoder, pReq->name) < 0) return -1;
   if (tEncodeCStr(&encoder, pReq->outputTbName) < 0) return -1;
   if (tEncodeI8(&encoder, pReq->igExists) < 0) return -1;
-  if (tEncodeCStr(&encoder, pReq->sql) < 0) return -1;
+  if (tEncodeI32(&encoder, sqlLen) < 0) return -1;
+  if (tEncodeI32(&encoder, astLen) < 0) return -1;
   if (sqlLen > 0 && tEncodeCStr(&encoder, pReq->sql) < 0) return -1;
   if (astLen > 0 && tEncodeCStr(&encoder, pReq->ast) < 0) return -1;
 
@@ -2874,30 +3005,36 @@ void tFreeSCMCreateStreamReq(SCMCreateStreamReq *pReq) {
 }
 
 int32_t tEncodeSStreamTask(SCoder *pEncoder, const SStreamTask *pTask) {
-  if (tStartEncode(pEncoder) < 0) return -1;
+  /*if (tStartEncode(pEncoder) < 0) return -1;*/
   if (tEncodeI64(pEncoder, pTask->streamId) < 0) return -1;
   if (tEncodeI32(pEncoder, pTask->taskId) < 0) return -1;
   if (tEncodeI32(pEncoder, pTask->level) < 0) return -1;
   if (tEncodeI8(pEncoder, pTask->status) < 0) return -1;
+  if (tEncodeI8(pEncoder, pTask->pipeSource) < 0) return -1;
   if (tEncodeI8(pEncoder, pTask->pipeSink) < 0) return -1;
+  if (tEncodeI8(pEncoder, pTask->parallelizable) < 0) return -1;
+  if (tEncodeI8(pEncoder, pTask->nextOpDst) < 0) return -1;
   // if (tEncodeI8(pEncoder, pTask->numOfRunners) < 0) return -1;
   if (tEncodeSEpSet(pEncoder, &pTask->NextOpEp) < 0) return -1;
   if (tEncodeCStr(pEncoder, pTask->qmsg) < 0) return -1;
-  tEndEncode(pEncoder);
+  /*tEndEncode(pEncoder);*/
   return pEncoder->pos;
 }
 
 int32_t tDecodeSStreamTask(SCoder *pDecoder, SStreamTask *pTask) {
-  if (tStartDecode(pDecoder) < 0) return -1;
+  /*if (tStartDecode(pDecoder) < 0) return -1;*/
   if (tDecodeI64(pDecoder, &pTask->streamId) < 0) return -1;
   if (tDecodeI32(pDecoder, &pTask->taskId) < 0) return -1;
   if (tDecodeI32(pDecoder, &pTask->level) < 0) return -1;
   if (tDecodeI8(pDecoder, &pTask->status) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pTask->pipeSource) < 0) return -1;
   if (tDecodeI8(pDecoder, &pTask->pipeSink) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pTask->parallelizable) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pTask->nextOpDst) < 0) return -1;
   // if (tDecodeI8(pDecoder, &pTask->numOfRunners) < 0) return -1;
   if (tDecodeSEpSet(pDecoder, &pTask->NextOpEp) < 0) return -1;
   if (tDecodeCStrAlloc(pDecoder, &pTask->qmsg) < 0) return -1;
-  tEndDecode(pDecoder);
+  /*tEndDecode(pDecoder);*/
   return 0;
 }
 

source/common/src/trow.c

@@ -78,7 +78,7 @@ static FORCE_INLINE void dataColSetNoneAt(SDataCol *pCol, int index, bool setBit
     setNull(POINTER_SHIFT(pCol->pData, TYPE_BYTES[pCol->type] * index), pCol->type, pCol->bytes);
     pCol->len += TYPE_BYTES[pCol->type];
   }
-  if(setBitmap) {
+  if (setBitmap) {
     tdSetBitmapValType(pCol->pBitmap, index, TD_VTYPE_NONE);
   }
 }
@@ -118,8 +118,8 @@ int trbWriteCol(SRowBuilder *pRB, void *pData, col_id_t cid) {
 
 #endif
 
-STSRow* tdRowDup(STSRow *row) {
-  STSRow* trow = malloc(TD_ROW_LEN(row));
+STSRow *tdRowDup(STSRow *row) {
+  STSRow *trow = malloc(TD_ROW_LEN(row));
   if (trow == NULL) return NULL;
 
   tdRowCpy(trow, row);
@@ -176,7 +176,7 @@ static int32_t tdAppendTpRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols
 
   SDataCol *pDataCol = &(pCols->cols[0]);
   if (pDataCol->colId == PRIMARYKEY_TIMESTAMP_COL_ID) {
-    tdAppendValToDataCol(pDataCol, TD_VTYPE_NORM, &pRow->ts,  pCols->numOfRows, pCols->maxPoints);
+    tdAppendValToDataCol(pDataCol, TD_VTYPE_NORM, &pRow->ts, pCols->numOfRows, pCols->maxPoints);
   }
 
   while (dcol < pCols->numOfCols) {
@@ -378,9 +378,7 @@ static void tdMergeTwoDataCols(SDataCols *target, SDataCols *src1, int *iter1, i
   }
 }
 
-
-
-STSRow* mergeTwoRows(void *buffer, STSRow* row1, STSRow *row2, STSchema *pSchema1, STSchema *pSchema2) {
+STSRow *mergeTwoRows(void *buffer, STSRow *row1, STSRow *row2, STSchema *pSchema1, STSchema *pSchema2) {
 #if 0
   ASSERT(TD_ROW_KEY(row1) == TD_ROW_KEY(row2));
   ASSERT(schemaVersion(pSchema1) == TD_ROW_SVER(row1));
@@ -473,6 +471,44 @@ STSRow* mergeTwoRows(void *buffer, STSRow* row1, STSRow *row2, STSchema *pSchema
   }
   taosArrayDestroy(stashRow);
   return buffer;
-  #endif
+#endif
   return NULL;
 }
+
+SDataCols *tdDupDataCols(SDataCols *pDataCols, bool keepData) {
+  SDataCols *pRet = tdNewDataCols(pDataCols->maxCols, pDataCols->maxPoints);
+  if (pRet == NULL) return NULL;
+
+  pRet->numOfCols = pDataCols->numOfCols;
+  pRet->sversion = pDataCols->sversion;
+  if (keepData) pRet->numOfRows = pDataCols->numOfRows;
+
+  for (int i = 0; i < pDataCols->numOfCols; i++) {
+    pRet->cols[i].type = pDataCols->cols[i].type;
+    pRet->cols[i].bitmap = pDataCols->cols[i].bitmap;
+    pRet->cols[i].colId = pDataCols->cols[i].colId;
+    pRet->cols[i].bytes = pDataCols->cols[i].bytes;
+    pRet->cols[i].offset = pDataCols->cols[i].offset;
+
+    if (keepData) {
+      if (pDataCols->cols[i].len > 0) {
+        if (tdAllocMemForCol(&pRet->cols[i], pRet->maxPoints) < 0) {
+          tdFreeDataCols(pRet);
+          return NULL;
+        }
+        pRet->cols[i].len = pDataCols->cols[i].len;
+        memcpy(pRet->cols[i].pData, pDataCols->cols[i].pData, pDataCols->cols[i].len);
+        if (IS_VAR_DATA_TYPE(pRet->cols[i].type)) {
+          int dataOffSize = sizeof(VarDataOffsetT) * pDataCols->maxPoints;
+          memcpy(pRet->cols[i].dataOff, pDataCols->cols[i].dataOff, dataOffSize);
+        }
+        if (!TD_COL_ROWS_NORM(pRet->cols + i)) {
+          int32_t nBitmapBytes = (int32_t)TD_BITMAP_BYTES(pDataCols->maxPoints);
+          memcpy(pRet->cols[i].pBitmap, pDataCols->cols[i].pBitmap, nBitmapBytes);
+        }
+      }
+    }
+  }
+
+  return pRet;
+}
\ No newline at end of file

source/dnode/mgmt/mnode/src/mmMsg.c

@@ -123,6 +123,8 @@ void mmInitMsgHandles(SMgmtWrapper *pWrapper) {
   dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_MND_DROP_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_SMA, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_SMA, (NodeMsgFp)mmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_MND_TABLE_META, (NodeMsgFp)mmProcessReadMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_MND_VGROUP_LIST, (NodeMsgFp)mmProcessReadMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_MND_KILL_QUERY, (NodeMsgFp)mmProcessWriteMsg, 0);
@@ -142,6 +144,8 @@ void mmInitMsgHandles(SMgmtWrapper *pWrapper) {
   dndSetMsgHandle(pWrapper, TDMT_MND_SUBSCRIBE, (NodeMsgFp)mmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_MND_MQ_COMMIT_OFFSET, (NodeMsgFp)mmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_MND_GET_SUB_EP, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STREAM, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
 
   // Requests handled by VNODE
   dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
@@ -149,4 +153,6 @@ void mmInitMsgHandles(SMgmtWrapper *pWrapper) {
   dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_SMA_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_SMA_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
 }

source/dnode/mgmt/vnode/src/vmMsg.c

@@ -268,6 +268,9 @@ void vmInitMsgHandles(SMgmtWrapper *pWrapper) {
   dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_SMA, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_SMA, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_SMA, (NodeMsgFp)vmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)vmProcessFetchMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES_FETCH, (NodeMsgFp)vmProcessFetchMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN, (NodeMsgFp)vmProcessWriteMsg, 0);
@@ -277,6 +280,7 @@ void vmInitMsgHandles(SMgmtWrapper *pWrapper) {
   dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY, (NodeMsgFp)vmProcessWriteMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_HEARTBEAT, (NodeMsgFp)vmProcessFetchMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_VND_TASK_EXEC, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_STREAM_TRIGGER, (NodeMsgFp)vmProcessFetchMsg, 0);
 
   dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
   dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);

source/dnode/mnode/impl/inc/mndDef.h

@@ -103,6 +103,8 @@ typedef enum {
   TRN_TYPE_CREATE_STB = 4001,
   TRN_TYPE_ALTER_STB = 4002,
   TRN_TYPE_DROP_STB = 4003,
+  TRN_TYPE_CREATE_SMA = 4004,
+  TRN_TYPE_DROP_SMA = 4005,
   TRN_TYPE_STB_SCOPE_END,
 } ETrnType;
 
@@ -305,6 +307,31 @@ typedef struct {
   SVnodeGid vnodeGid[TSDB_MAX_REPLICA];
 } SVgObj;
 
+typedef struct {
+  char    name[TSDB_TABLE_FNAME_LEN];
+  char    stb[TSDB_TABLE_FNAME_LEN];
+  char    db[TSDB_DB_FNAME_LEN];
+  int64_t createdTime;
+  int64_t uid;
+  int64_t stbUid;
+  int64_t dbUid;
+  int8_t  intervalUnit;
+  int8_t  slidingUnit;
+  int8_t  timezone;
+  int32_t dstVgId;  // for stream
+  int64_t interval;
+  int64_t offset;
+  int64_t sliding;
+  int32_t exprLen;  // strlen + 1
+  int32_t tagsFilterLen;
+  int32_t sqlLen;
+  int32_t astLen;
+  char*   expr;
+  char*   tagsFilter;
+  char*   sql;
+  char*   ast;
+} SSmaObj;
+
 typedef struct {
   char     name[TSDB_TABLE_FNAME_LEN];
   char     db[TSDB_DB_FNAME_LEN];
@@ -314,12 +341,16 @@ typedef struct {
   int64_t  dbUid;
   int32_t  version;
   int32_t  nextColId;
+  float    xFilesFactor;
+  int32_t  aggregationMethod;
+  int32_t  delay;
   int32_t  numOfColumns;
   int32_t  numOfTags;
+  int32_t  commentLen;
   SSchema* pColumns;
   SSchema* pTags;
+  char*    comment;
   SRWLatch lock;
-  char     comment[TSDB_STB_COMMENT_LEN];
 } SStbObj;
 
 typedef struct {
@@ -697,12 +728,12 @@ typedef struct {
   char*   logicalPlan;
   char*   physicalPlan;
   SArray* tasks;  // SArray<SArray<SStreamTask>>
+  SArray* outputName;
 } SStreamObj;
 
 int32_t tEncodeSStreamObj(SCoder* pEncoder, const SStreamObj* pObj);
 int32_t tDecodeSStreamObj(SCoder* pDecoder, SStreamObj* pObj);
 
-
 #ifdef __cplusplus
 }
 #endif

source/dnode/mnode/impl/inc/mndInt.h

@@ -122,9 +122,9 @@ typedef struct SMnode {
   SMsgCb            msgCb;
 } SMnode;
 
-void     mndSetMsgHandle(SMnode *pMnode, tmsg_t msgType, MndMsgFp fp);
-uint64_t mndGenerateUid(char *name, int32_t len);
-void     mndGetLoad(SMnode *pMnode, SMnodeLoad *pLoad);
+void    mndSetMsgHandle(SMnode *pMnode, tmsg_t msgType, MndMsgFp fp);
+int64_t mndGenerateUid(char *name, int32_t len);
+void    mndGetLoad(SMnode *pMnode, SMnodeLoad *pLoad);
 
 #ifdef __cplusplus
 }

source/dnode/mnode/impl/inc/mndSma.h

+/*
+ * 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_MND_SMA_H_
+#define _TD_MND_SMA_H_
+
+#include "mndInt.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+int32_t  mndInitSma(SMnode *pMnode);
+void     mndCleanupSma(SMnode *pMnode);
+SSmaObj *mndAcquireSma(SMnode *pMnode, char *smaName);
+void     mndReleaseSma(SMnode *pMnode, SSmaObj *pSma);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_MND_SMA_H_*/

source/dnode/mnode/impl/inc/mndStream.h

@@ -31,6 +31,8 @@ void        mndReleaseStream(SMnode *pMnode, SStreamObj *pStream);
 SSdbRaw *mndStreamActionEncode(SStreamObj *pStream);
 SSdbRow *mndStreamActionDecode(SSdbRaw *pRaw);
 
+int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast, STrans *pTrans);
+
 #ifdef __cplusplus
 }
 #endif

source/dnode/mnode/impl/src/mndDef.c

@@ -16,6 +16,7 @@
 #include "mndDef.h"
 
 int32_t tEncodeSStreamObj(SCoder *pEncoder, const SStreamObj *pObj) {
+  int32_t outputNameSz = 0;
   if (tEncodeCStr(pEncoder, pObj->name) < 0) return -1;
   if (tEncodeCStr(pEncoder, pObj->db) < 0) return -1;
   if (tEncodeI64(pEncoder, pObj->createTime) < 0) return -1;
@@ -43,6 +44,15 @@ int32_t tEncodeSStreamObj(SCoder *pEncoder, const SStreamObj *pObj) {
   } else {
     tEncodeI32(pEncoder, 0);
   }
+
+  if (pObj->outputName != NULL) {
+    outputNameSz = taosArrayGetSize(pObj->outputName);
+  }
+  if (tEncodeI32(pEncoder, outputNameSz) < 0) return -1;
+  for (int32_t i = 0; i < outputNameSz; i++) {
+    char *name = taosArrayGetP(pObj->outputName, i);
+    if (tEncodeCStr(pEncoder, name) < 0) return -1;
+  }
   return pEncoder->pos;
 }
 
@@ -76,5 +86,16 @@ int32_t tDecodeSStreamObj(SCoder *pDecoder, SStreamObj *pObj) {
   } else {
     pObj->tasks = NULL;
   }
+  int32_t outputNameSz;
+  if (tDecodeI32(pDecoder, &outputNameSz) < 0) return -1;
+  pObj->outputName = taosArrayInit(outputNameSz, sizeof(void *));
+  if (pObj->outputName == NULL) {
+    return -1;
+  }
+  for (int32_t i = 0; i < outputNameSz; i++) {
+    char *name;
+    if (tDecodeCStrAlloc(pDecoder, &name) < 0) return -1;
+    taosArrayPush(pObj->outputName, &name);
+  }
   return 0;
 }

source/dnode/mnode/impl/src/mndScheduler.c

@@ -32,20 +32,23 @@
 #include "tname.h"
 #include "tuuid.h"
 
-int32_t mndPersistTaskDeployReq(STrans* pTrans, SStreamTask* pTask, const SEpSet* pEpSet, tmsg_t type) {
+extern bool tsStreamSchedV;
+
+int32_t mndPersistTaskDeployReq(STrans* pTrans, SStreamTask* pTask, const SEpSet* pEpSet, tmsg_t type, int32_t nodeId) {
   SCoder encoder;
   tCoderInit(&encoder, TD_LITTLE_ENDIAN, NULL, 0, TD_ENCODER);
   tEncodeSStreamTask(&encoder, pTask);
-  int32_t tlen = sizeof(SMsgHead) + encoder.pos;
+  int32_t size = encoder.pos;
+  int32_t tlen = sizeof(SMsgHead) + size;
   tCoderClear(&encoder);
   void* buf = malloc(tlen);
   if (buf == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
-  ((SMsgHead*)buf)->streamTaskId = pTask->taskId;
+  ((SMsgHead*)buf)->streamTaskId = htonl(nodeId);
   void* abuf = POINTER_SHIFT(buf, sizeof(SMsgHead));
-  tCoderInit(&encoder, TD_LITTLE_ENDIAN, abuf, tlen, TD_ENCODER);
+  tCoderInit(&encoder, TD_LITTLE_ENDIAN, abuf, size, TD_ENCODER);
   tEncodeSStreamTask(&encoder, pTask);
   tCoderClear(&encoder);
 
@@ -70,7 +73,7 @@ int32_t mndAssignTaskToVg(SMnode* pMnode, STrans* pTrans, SStreamTask* pTask, SS
     terrno = TSDB_CODE_QRY_INVALID_INPUT;
     return -1;
   }
-  mndPersistTaskDeployReq(pTrans, pTask, &plan->execNode.epSet, TDMT_VND_TASK_DEPLOY);
+  mndPersistTaskDeployReq(pTrans, pTask, &plan->execNode.epSet, TDMT_VND_TASK_DEPLOY, pVgroup->vgId);
   return 0;
 }
 
@@ -90,7 +93,7 @@ int32_t mndAssignTaskToSnode(SMnode* pMnode, STrans* pTrans, SStreamTask* pTask,
     terrno = TSDB_CODE_QRY_INVALID_INPUT;
     return -1;
   }
-  mndPersistTaskDeployReq(pTrans, pTask, &plan->execNode.epSet, TDMT_SND_TASK_DEPLOY);
+  mndPersistTaskDeployReq(pTrans, pTask, &plan->execNode.epSet, TDMT_SND_TASK_DEPLOY, 0);
   return 0;
 }
 
@@ -106,6 +109,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
 
   int32_t totLevel = LIST_LENGTH(pPlan->pSubplans);
   pStream->tasks = taosArrayInit(totLevel, sizeof(SArray));
+  int32_t lastUsedVgId = 0;
 
   for (int32_t level = 0; level < totLevel; level++) {
     SArray*        taskOneLevel = taosArrayInit(0, sizeof(SStreamTask));
@@ -113,9 +117,9 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
     int32_t        opNum = LIST_LENGTH(inner->pNodeList);
     ASSERT(opNum == 1);
 
-    SSubplan* plan = nodesListGetNode(inner->pNodeList, level);
+    SSubplan* plan = nodesListGetNode(inner->pNodeList, 0);
     if (level == 0) {
-      ASSERT(plan->type == SUBPLAN_TYPE_SCAN);
+      ASSERT(plan->subplanType == SUBPLAN_TYPE_SCAN);
       void* pIter = NULL;
       while (1) {
         pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void**)&pVgroup);
@@ -125,11 +129,14 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
           continue;
         }
 
+        lastUsedVgId = pVgroup->vgId;
         pStream->vgNum++;
         // send to vnode
 
         SStreamTask* pTask = streamTaskNew(pStream->uid, level);
+        pTask->pipeSource = 1;
         pTask->pipeSink = level == totLevel - 1 ? 1 : 0;
+        pTask->parallelizable = 1;
         // TODO: set to
         if (mndAssignTaskToVg(pMnode, pTrans, pTask, plan, pVgroup) < 0) {
           sdbRelease(pSdb, pVgroup);
@@ -140,19 +147,35 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
       }
     } else {
       SStreamTask* pTask = streamTaskNew(pStream->uid, level);
+      pTask->pipeSource = 0;
       pTask->pipeSink = level == totLevel - 1 ? 1 : 0;
-      SSnodeObj* pSnode = mndSchedFetchSnode(pMnode);
-      if (pSnode != NULL) {
-        if (mndAssignTaskToSnode(pMnode, pTrans, pTask, plan, pSnode) < 0) {
-          sdbRelease(pSdb, pSnode);
+      pTask->parallelizable = plan->subplanType == SUBPLAN_TYPE_SCAN;
+      pTask->nextOpDst = STREAM_NEXT_OP_DST__VND;
+
+      if (tsStreamSchedV) {
+        ASSERT(lastUsedVgId != 0);
+        SVgObj* pVg = mndAcquireVgroup(pMnode, lastUsedVgId);
+        if (mndAssignTaskToVg(pMnode, pTrans, pTask, plan, pVg) < 0) {
+          sdbRelease(pSdb, pVg);
           qDestroyQueryPlan(pPlan);
           return -1;
         }
-        sdbRelease(pMnode->pSdb, pSnode);
+        sdbRelease(pSdb, pVg);
       } else {
-        // TODO: assign to one vg
-        ASSERT(0);
+        SSnodeObj* pSnode = mndSchedFetchSnode(pMnode);
+        if (pSnode != NULL) {
+          if (mndAssignTaskToSnode(pMnode, pTrans, pTask, plan, pSnode) < 0) {
+            sdbRelease(pSdb, pSnode);
+            qDestroyQueryPlan(pPlan);
+            return -1;
+          }
+          sdbRelease(pMnode->pSdb, pSnode);
+        } else {
+          // TODO: assign to one vg
+          ASSERT(0);
+        }
       }
+
       taosArrayPush(taskOneLevel, pTask);
     }
     taosArrayPush(pStream->tasks, taskOneLevel);

source/dnode/mnode/impl/src/mndShow.c

@@ -427,6 +427,8 @@ char *mndShowStr(int32_t showType) {
       return "show topics";
     case TSDB_MGMT_TABLE_FUNC:
       return "show functions";
+      case TSDB_MGMT_TABLE_INDEX:
+      return "show indexes";
     default:
       return "undefined";
   }

source/dnode/mnode/impl/src/mndStb.c

@@ -13,20 +13,19 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#define _DEFAULT_SOURCE
 #include "mndStb.h"
 #include "mndAuth.h"
 #include "mndDb.h"
 #include "mndDnode.h"
+#include "mndInfoSchema.h"
 #include "mndMnode.h"
 #include "mndShow.h"
 #include "mndTrans.h"
 #include "mndUser.h"
 #include "mndVgroup.h"
-#include "mndInfoSchema.h"
 #include "tname.h"
 
-#define TSDB_STB_VER_NUMBER 1
+#define TSDB_STB_VER_NUMBER   1
 #define TSDB_STB_RESERVE_SIZE 64
 
 static SSdbRow *mndStbActionDecode(SSdbRaw *pRaw);
@@ -86,8 +85,12 @@ SSdbRaw *mndStbActionEncode(SStbObj *pStb) {
   SDB_SET_INT64(pRaw, dataPos, pStb->dbUid, STB_ENCODE_OVER)
   SDB_SET_INT32(pRaw, dataPos, pStb->version, STB_ENCODE_OVER)
   SDB_SET_INT32(pRaw, dataPos, pStb->nextColId, STB_ENCODE_OVER)
+  SDB_SET_INT32(pRaw, dataPos, (int32_t)(pStb->xFilesFactor * 10000), STB_ENCODE_OVER)
+  SDB_SET_INT32(pRaw, dataPos, pStb->aggregationMethod, STB_ENCODE_OVER)
+  SDB_SET_INT32(pRaw, dataPos, pStb->delay, STB_ENCODE_OVER)
   SDB_SET_INT32(pRaw, dataPos, pStb->numOfColumns, STB_ENCODE_OVER)
   SDB_SET_INT32(pRaw, dataPos, pStb->numOfTags, STB_ENCODE_OVER)
+  SDB_SET_INT32(pRaw, dataPos, pStb->commentLen, STB_ENCODE_OVER)
 
   for (int32_t i = 0; i < pStb->numOfColumns; ++i) {
     SSchema *pSchema = &pStb->pColumns[i];
@@ -105,7 +108,7 @@ SSdbRaw *mndStbActionEncode(SStbObj *pStb) {
     SDB_SET_BINARY(pRaw, dataPos, pSchema->name, TSDB_COL_NAME_LEN, STB_ENCODE_OVER)
   }
 
-  SDB_SET_BINARY(pRaw, dataPos, pStb->comment, TSDB_STB_COMMENT_LEN, STB_ENCODE_OVER)
+  SDB_SET_BINARY(pRaw, dataPos, pStb->comment, pStb->commentLen, STB_ENCODE_OVER)
   SDB_SET_RESERVE(pRaw, dataPos, TSDB_STB_RESERVE_SIZE, STB_ENCODE_OVER)
   SDB_SET_DATALEN(pRaw, dataPos, STB_ENCODE_OVER)
 
@@ -148,8 +151,14 @@ static SSdbRow *mndStbActionDecode(SSdbRaw *pRaw) {
   SDB_GET_INT64(pRaw, dataPos, &pStb->dbUid, STB_DECODE_OVER)
   SDB_GET_INT32(pRaw, dataPos, &pStb->version, STB_DECODE_OVER)
   SDB_GET_INT32(pRaw, dataPos, &pStb->nextColId, STB_DECODE_OVER)
+  int32_t xFilesFactor = 0;
+  SDB_GET_INT32(pRaw, dataPos, &xFilesFactor, STB_DECODE_OVER)
+  pStb->xFilesFactor = xFilesFactor / 10000.0f;
+  SDB_GET_INT32(pRaw, dataPos, &pStb->aggregationMethod, STB_DECODE_OVER)
+  SDB_GET_INT32(pRaw, dataPos, &pStb->delay, STB_DECODE_OVER)
   SDB_GET_INT32(pRaw, dataPos, &pStb->numOfColumns, STB_DECODE_OVER)
   SDB_GET_INT32(pRaw, dataPos, &pStb->numOfTags, STB_DECODE_OVER)
+  SDB_GET_INT32(pRaw, dataPos, &pStb->commentLen, STB_DECODE_OVER)
 
   pStb->pColumns = calloc(pStb->numOfColumns, sizeof(SSchema));
   pStb->pTags = calloc(pStb->numOfTags, sizeof(SSchema));
@@ -173,7 +182,11 @@ static SSdbRow *mndStbActionDecode(SSdbRaw *pRaw) {
     SDB_GET_BINARY(pRaw, dataPos, pSchema->name, TSDB_COL_NAME_LEN, STB_DECODE_OVER)
   }
 
-  SDB_GET_BINARY(pRaw, dataPos, pStb->comment, TSDB_STB_COMMENT_LEN, STB_DECODE_OVER)
+  if (pStb->commentLen > 0) {
+    pStb->comment = calloc(pStb->commentLen, 1);
+    if (pStb->comment == NULL) goto STB_DECODE_OVER;
+    SDB_GET_BINARY(pRaw, dataPos, pStb->comment, pStb->commentLen, STB_DECODE_OVER)
+  }
   SDB_GET_RESERVE(pRaw, dataPos, TSDB_STB_RESERVE_SIZE, STB_DECODE_OVER)
 
   terrno = 0;
@@ -183,6 +196,7 @@ STB_DECODE_OVER:
     mError("stb:%s, failed to decode from raw:%p since %s", pStb->name, pRaw, terrstr());
     tfree(pStb->pColumns);
     tfree(pStb->pTags);
+    tfree(pStb->comment);
     tfree(pRow);
     return NULL;
   }
@@ -200,6 +214,7 @@ static int32_t mndStbActionDelete(SSdb *pSdb, SStbObj *pStb) {
   mTrace("stb:%s, perform delete action, row:%p", pStb->name, pStb);
   tfree(pStb->pColumns);
   tfree(pStb->pTags);
+  tfree(pStb->comment);
   return 0;
 }
 
@@ -505,6 +520,13 @@ static int32_t mndCreateStb(SMnode *pMnode, SNodeMsg *pReq, SMCreateStbReq *pCre
   stbObj.nextColId = 1;
   stbObj.numOfColumns = pCreate->numOfColumns;
   stbObj.numOfTags = pCreate->numOfTags;
+  stbObj.commentLen = pCreate->commentLen;
+  stbObj.comment = calloc(stbObj.commentLen, 1);
+  if (stbObj.comment == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return -1;
+  }
+  memcpy(stbObj.comment, pCreate->comment, stbObj.commentLen);
 
   stbObj.pColumns = malloc(stbObj.numOfColumns * sizeof(SSchema));
   stbObj.pTags = malloc(stbObj.numOfTags * sizeof(SSchema));
@@ -1165,7 +1187,7 @@ static int32_t mndSetDropStbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj *
 
 static int32_t mndDropStb(SMnode *pMnode, SNodeMsg *pReq, SDbObj *pDb, SStbObj *pStb) {
   int32_t code = -1;
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK,TRN_TYPE_DROP_STB, &pReq->rpcMsg);
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_TYPE_DROP_STB, &pReq->rpcMsg);
   if (pTrans == NULL) goto DROP_STB_OVER;
 
   mDebug("trans:%d, used to drop stb:%s", pTrans->id, pStb->name);
@@ -1571,7 +1593,11 @@ static int32_t mndRetrieveStb(SNodeMsg *pReq, SShowObj *pShow, char *data, int32
     cols++;
 
     pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
-    STR_TO_VARSTR(pWrite, pStb->comment);
+    if (pStb->commentLen != 0) {
+      STR_TO_VARSTR(pWrite, pStb->comment);
+    } else {
+      STR_TO_VARSTR(pWrite, "");
+    }
     cols++;
 
     numOfRows++;

source/dnode/mnode/impl/src/mndStream.c

@@ -21,6 +21,7 @@
 #include "mndScheduler.h"
 #include "mndShow.h"
 #include "mndStb.h"
+#include "mndTopic.h"
 #include "mndTrans.h"
 #include "mndUser.h"
 #include "mndVgroup.h"
@@ -33,6 +34,7 @@ static int32_t mndStreamActionInsert(SSdb *pSdb, SStreamObj *pStream);
 static int32_t mndStreamActionDelete(SSdb *pSdb, SStreamObj *pStream);
 static int32_t mndStreamActionUpdate(SSdb *pSdb, SStreamObj *pStream, SStreamObj *pNewStream);
 static int32_t mndProcessCreateStreamReq(SNodeMsg *pReq);
+static int32_t mndProcessTaskDeployInternalRsp(SNodeMsg *pRsp);
 /*static int32_t mndProcessDropStreamReq(SNodeMsg *pReq);*/
 /*static int32_t mndProcessDropStreamInRsp(SNodeMsg *pRsp);*/
 static int32_t mndProcessStreamMetaReq(SNodeMsg *pReq);
@@ -50,6 +52,8 @@ int32_t mndInitStream(SMnode *pMnode) {
                      .deleteFp = (SdbDeleteFp)mndStreamActionDelete};
 
   mndSetMsgHandle(pMnode, TDMT_MND_CREATE_STREAM, mndProcessCreateStreamReq);
+  mndSetMsgHandle(pMnode, TDMT_VND_TASK_DEPLOY_RSP, mndProcessTaskDeployInternalRsp);
+  mndSetMsgHandle(pMnode, TDMT_SND_TASK_DEPLOY_RSP, mndProcessTaskDeployInternalRsp);
   /*mndSetMsgHandle(pMnode, TDMT_MND_DROP_STREAM, mndProcessDropStreamReq);*/
   /*mndSetMsgHandle(pMnode, TDMT_MND_DROP_STREAM_RSP, mndProcessDropStreamInRsp);*/
 
@@ -68,7 +72,7 @@ SSdbRaw *mndStreamActionEncode(SStreamObj *pStream) {
 
   SCoder encoder;
   tCoderInit(&encoder, TD_LITTLE_ENDIAN, NULL, 0, TD_ENCODER);
-  if (tEncodeSStreamObj(NULL, pStream) < 0) {
+  if (tEncodeSStreamObj(&encoder, pStream) < 0) {
     tCoderClear(&encoder);
     goto STREAM_ENCODE_OVER;
   }
@@ -83,7 +87,7 @@ SSdbRaw *mndStreamActionEncode(SStreamObj *pStream) {
   if (buf == NULL) goto STREAM_ENCODE_OVER;
 
   tCoderInit(&encoder, TD_LITTLE_ENDIAN, buf, tlen, TD_ENCODER);
-  if (tEncodeSStreamObj(NULL, pStream) < 0) {
+  if (tEncodeSStreamObj(&encoder, pStream) < 0) {
     tCoderClear(&encoder);
     goto STREAM_ENCODE_OVER;
   }
@@ -135,7 +139,7 @@ SSdbRow *mndStreamActionDecode(SSdbRaw *pRaw) {
   SDB_GET_BINARY(pRaw, dataPos, buf, tlen, STREAM_DECODE_OVER);
 
   SCoder decoder;
-  tCoderInit(&decoder, TD_LITTLE_ENDIAN, NULL, 0, TD_DECODER);
+  tCoderInit(&decoder, TD_LITTLE_ENDIAN, buf, tlen + 1, TD_DECODER);
   if (tDecodeSStreamObj(&decoder, pStream) < 0) {
     goto STREAM_DECODE_OVER;
   }
@@ -191,6 +195,11 @@ void mndReleaseStream(SMnode *pMnode, SStreamObj *pStream) {
   sdbRelease(pSdb, pStream);
 }
 
+static int32_t mndProcessTaskDeployInternalRsp(SNodeMsg *pRsp) {
+  mndTransProcessRsp(pRsp);
+  return 0;
+}
+
 static SDbObj *mndAcquireDbByStream(SMnode *pMnode, char *streamName) {
   SName name = {0};
   tNameFromString(&name, streamName, T_NAME_ACCT | T_NAME_DB | T_NAME_TABLE);
@@ -209,6 +218,90 @@ static int32_t mndCheckCreateStreamReq(SCMCreateStreamReq *pCreate) {
   return 0;
 }
 
+static SArray *mndExtractNamesFromAst(const SNode *pAst) {
+  if (pAst->type != QUERY_NODE_SELECT_STMT) return NULL;
+
+  SArray *names = taosArrayInit(0, sizeof(void *));
+  if (names == NULL) {
+    return NULL;
+  }
+  SSelectStmt *pSelect = (SSelectStmt *)pAst;
+  SNodeList   *pNodes = pSelect->pProjectionList;
+  SListCell   *pCell = pNodes->pHead;
+  while (pCell != NULL) {
+    if (pCell->pNode->type != QUERY_NODE_FUNCTION) {
+      continue;
+    }
+    SFunctionNode *pFunction = (SFunctionNode *)pCell->pNode;
+    char          *name = strdup(pFunction->node.aliasName);
+    taosArrayPush(names, &name);
+    pCell = pCell->pNext;
+  }
+  return names;
+}
+
+static int32_t mndStreamGetPlanString(const char *ast, char **pStr) {
+  if (NULL == ast) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  SNode  *pAst = NULL;
+  int32_t code = nodesStringToNode(ast, &pAst);
+
+  SQueryPlan *pPlan = NULL;
+  if (TSDB_CODE_SUCCESS == code) {
+    SPlanContext cxt = {
+        .pAstRoot = pAst,
+        .topicQuery = false,
+        .streamQuery = true,
+    };
+    code = qCreateQueryPlan(&cxt, &pPlan, NULL);
+  }
+
+  if (TSDB_CODE_SUCCESS == code) {
+    code = nodesNodeToString(pPlan, false, pStr, NULL);
+  }
+  nodesDestroyNode(pAst);
+  nodesDestroyNode(pPlan);
+  terrno = code;
+  return code;
+}
+
+int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast, STrans *pTrans) {
+  SNode *pAst = NULL;
+  if (nodesStringToNode(ast, &pAst) < 0) {
+    return -1;
+  }
+  SArray *names = mndExtractNamesFromAst(pAst);
+  printf("|");
+  for (int i = 0; i < taosArrayGetSize(names); i++) {
+    printf(" %15s |", (char *)taosArrayGetP(names, i));
+  }
+  printf("\n=======================================================\n");
+
+  pStream->outputName = names;
+
+  if (TSDB_CODE_SUCCESS != mndStreamGetPlanString(ast, &pStream->physicalPlan)) {
+    mError("topic:%s, failed to get plan since %s", pStream->name, terrstr());
+    return -1;
+  }
+
+  if (mndScheduleStream(pMnode, pTrans, pStream) < 0) {
+    mError("stream:%ld, schedule stream since %s", pStream->uid, terrstr());
+    return -1;
+  }
+
+  SSdbRaw *pRedoRaw = mndStreamActionEncode(pStream);
+  if (pRedoRaw == NULL || mndTransAppendRedolog(pTrans, pRedoRaw) != 0) {
+    mError("trans:%d, failed to append redo log since %s", pTrans->id, terrstr());
+    mndTransDrop(pTrans);
+    return -1;
+  }
+  sdbSetRawStatus(pRedoRaw, SDB_STATUS_READY);
+
+  return 0;
+}
+
 static int32_t mndCreateStream(SMnode *pMnode, SNodeMsg *pReq, SCMCreateStreamReq *pCreate, SDbObj *pDb) {
   mDebug("stream:%s to create", pCreate->name);
   SStreamObj streamObj = {0};
@@ -220,30 +313,22 @@ static int32_t mndCreateStream(SMnode *pMnode, SNodeMsg *pReq, SCMCreateStreamRe
   streamObj.dbUid = pDb->uid;
   streamObj.version = 1;
   streamObj.sql = pCreate->sql;
-  streamObj.physicalPlan = "";
-  streamObj.logicalPlan = "";
+  /*streamObj.physicalPlan = "";*/
+  streamObj.logicalPlan = "not implemented";
 
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_TYPE_CREATE_STREAM, &pReq->rpcMsg);
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_TYPE_CREATE_STREAM, &pReq->rpcMsg);
   if (pTrans == NULL) {
     mError("stream:%s, failed to create since %s", pCreate->name, terrstr());
     return -1;
   }
   mDebug("trans:%d, used to create stream:%s", pTrans->id, pCreate->name);
 
-  if (mndScheduleStream(pMnode, pTrans, &streamObj) < 0) {
-    mError("stream:%ld, schedule stream since %s", streamObj.uid, terrstr());
+  if (mndAddStreamToTrans(pMnode, &streamObj, pCreate->ast, pTrans) != 0) {
+    mError("trans:%d, failed to add stream since %s", pTrans->id, terrstr());
     mndTransDrop(pTrans);
     return -1;
   }
 
-  SSdbRaw *pRedoRaw = mndStreamActionEncode(&streamObj);
-  if (pRedoRaw == NULL || mndTransAppendRedolog(pTrans, pRedoRaw) != 0) {
-    mError("trans:%d, failed to append redo log since %s", pTrans->id, terrstr());
-    mndTransDrop(pTrans);
-    return -1;
-  }
-  sdbSetRawStatus(pRedoRaw, SDB_STATUS_READY);
-
   if (mndTransPrepare(pMnode, pTrans) != 0) {
     mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
     mndTransDrop(pTrans);

source/dnode/mnode/impl/src/mndTopic.c

@@ -236,7 +236,7 @@ static int32_t mndCheckCreateTopicReq(SCMCreateTopicReq *pCreate) {
   return 0;
 }
 
-static int32_t mndGetPlanString(SCMCreateTopicReq *pCreate, char **pStr) {
+static int32_t mndGetPlanString(const SCMCreateTopicReq *pCreate, char **pStr) {
   if (NULL == pCreate->ast) {
     return TSDB_CODE_SUCCESS;
   }

source/dnode/mnode/impl/src/mndTrans.c

@@ -406,6 +406,10 @@ static const char *mndTransType(ETrnType type) {
       return "alter-stb";
     case TRN_TYPE_DROP_STB:
       return "drop-stb";
+    case TRN_TYPE_CREATE_SMA:
+      return "create-sma";
+    case TRN_TYPE_DROP_SMA:
+      return "drop-sma";
     default:
       return "invalid";
   }

source/dnode/mnode/impl/src/mnode.c

@@ -28,6 +28,7 @@
 #include "mndProfile.h"
 #include "mndQnode.h"
 #include "mndShow.h"
+#include "mndSma.h"
 #include "mndSnode.h"
 #include "mndStb.h"
 #include "mndStream.h"
@@ -204,6 +205,7 @@ static int32_t mndInitSteps(SMnode *pMnode) {
   if (mndAllocStep(pMnode, "mnode-offset", mndInitOffset, mndCleanupOffset) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-vgroup", mndInitVgroup, mndCleanupVgroup) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-stb", mndInitStb, mndCleanupStb) != 0) return -1;
+  if (mndAllocStep(pMnode, "mnode-stb", mndInitSma, mndCleanupSma) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-infos", mndInitInfos, mndCleanupInfos) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-db", mndInitDb, mndCleanupDb) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-func", mndInitFunc, mndCleanupFunc) != 0) return -1;
@@ -409,15 +411,15 @@ void mndSetMsgHandle(SMnode *pMnode, tmsg_t msgType, MndMsgFp fp) {
 }
 
 // Note: uid 0 is reserved
-uint64_t mndGenerateUid(char *name, int32_t len) {
+int64_t mndGenerateUid(char *name, int32_t len) {
   int32_t hashval = MurmurHash3_32(name, len);
 
   do {
     int64_t  us = taosGetTimestampUs();
-    uint64_t x = (us & 0x000000FFFFFFFFFF) << 24;
-    uint64_t uuid = x + ((hashval & ((1ul << 16) - 1ul)) << 8) + (taosRand() & ((1ul << 8) - 1ul));
+    int64_t x = (us & 0x000000FFFFFFFFFF) << 24;
+    int64_t uuid = x + ((hashval & ((1ul << 16) - 1ul)) << 8) + (taosRand() & ((1ul << 8) - 1ul));
     if (uuid) {
-      return uuid;
+      return abs(uuid);
     }
   } while (true);
 }

source/dnode/mnode/impl/test/CMakeLists.txt

@@ -12,5 +12,6 @@ add_subdirectory(dnode)
 add_subdirectory(mnode)
 add_subdirectory(db)
 add_subdirectory(stb)
+add_subdirectory(sma)
 add_subdirectory(func)
 add_subdirectory(topic)

source/dnode/mnode/impl/test/sma/CMakeLists.txt

+aux_source_directory(. SMA_SRC)
+add_executable(mnode_test_sma ${SMA_SRC})
+target_link_libraries(
+    mnode_test_sma
+    PUBLIC sut
+)
+
+add_test(
+    NAME mnode_test_sma
+    COMMAND mnode_test_sma
+)

source/dnode/mnode/impl/test/sma/sma.cpp

+/**
+ * @file sma.cpp
+ * @author slguan (slguan@taosdata.com)
+ * @brief MNODE module sma tests
+ * @version 1.0
+ * @date 2022-03-23
+ *
+ * @copyright Copyright (c) 2022
+ *
+ */
+
+#include "sut.h"
+
+class MndTestSma : public ::testing::Test {
+ protected:
+  static void SetUpTestSuite() { test.Init("/tmp/mnode_test_sma", 9035); }
+  static void TearDownTestSuite() { test.Cleanup(); }
+
+  static Testbase test;
+
+ public:
+  void SetUp() override {}
+  void TearDown() override {}
+
+  void* BuildCreateDbReq(const char* dbname, int32_t* pContLen);
+  void* BuildDropDbReq(const char* dbname, int32_t* pContLen);
+  void* BuildCreateStbReq(const char* stbname, int32_t* pContLen);
+  void* BuildDropStbReq(const char* stbname, int32_t* pContLen);
+  void* BuildCreateSmaReq(const char* smaname, const char* stbname, int8_t igExists, const char* expr,
+                          const char* tagsFilter, const char* sql, const char* ast, int32_t* pContLen);
+  void* BuildDropSmaReq(const char* smaname, int8_t igNotExists, int32_t* pContLen);
+};
+
+Testbase MndTestSma::test;
+
+void* MndTestSma::BuildCreateDbReq(const char* dbname, int32_t* pContLen) {
+  SCreateDbReq createReq = {0};
+  strcpy(createReq.db, dbname);
+  createReq.numOfVgroups = 2;
+  createReq.cacheBlockSize = 16;
+  createReq.totalBlocks = 10;
+  createReq.daysPerFile = 10;
+  createReq.daysToKeep0 = 3650;
+  createReq.daysToKeep1 = 3650;
+  createReq.daysToKeep2 = 3650;
+  createReq.minRows = 100;
+  createReq.maxRows = 4096;
+  createReq.commitTime = 3600;
+  createReq.fsyncPeriod = 3000;
+  createReq.walLevel = 1;
+  createReq.precision = 0;
+  createReq.compression = 2;
+  createReq.replications = 1;
+  createReq.quorum = 1;
+  createReq.update = 0;
+  createReq.cacheLastRow = 0;
+  createReq.ignoreExist = 1;
+
+  int32_t contLen = tSerializeSCreateDbReq(NULL, 0, &createReq);
+  void*   pReq = rpcMallocCont(contLen);
+  tSerializeSCreateDbReq(pReq, contLen, &createReq);
+
+  *pContLen = contLen;
+  return pReq;
+}
+
+void* MndTestSma::BuildDropDbReq(const char* dbname, int32_t* pContLen) {
+  SDropDbReq dropdbReq = {0};
+  strcpy(dropdbReq.db, dbname);
+
+  int32_t contLen = tSerializeSDropDbReq(NULL, 0, &dropdbReq);
+  void*   pReq = rpcMallocCont(contLen);
+  tSerializeSDropDbReq(pReq, contLen, &dropdbReq);
+
+  *pContLen = contLen;
+  return pReq;
+}
+
+void* MndTestSma::BuildCreateStbReq(const char* stbname, int32_t* pContLen) {
+  SMCreateStbReq createReq = {0};
+  createReq.numOfColumns = 3;
+  createReq.numOfTags = 1;
+  createReq.igExists = 0;
+  createReq.pColumns = taosArrayInit(createReq.numOfColumns, sizeof(SField));
+  createReq.pTags = taosArrayInit(createReq.numOfTags, sizeof(SField));
+  strcpy(createReq.name, stbname);
+
+  {
+    SField field = {0};
+    field.bytes = 8;
+    field.type = TSDB_DATA_TYPE_TIMESTAMP;
+    strcpy(field.name, "ts");
+    taosArrayPush(createReq.pColumns, &field);
+  }
+
+  {
+    SField field = {0};
+    field.bytes = 2;
+    field.type = TSDB_DATA_TYPE_TINYINT;
+    strcpy(field.name, "col1");
+    taosArrayPush(createReq.pColumns, &field);
+  }
+
+  {
+    SField field = {0};
+    field.bytes = 8;
+    field.type = TSDB_DATA_TYPE_BIGINT;
+    strcpy(field.name, "col2");
+    taosArrayPush(createReq.pColumns, &field);
+  }
+
+  {
+    SField field = {0};
+    field.bytes = 2;
+    field.type = TSDB_DATA_TYPE_TINYINT;
+    strcpy(field.name, "tag1");
+    taosArrayPush(createReq.pTags, &field);
+  }
+
+  int32_t tlen = tSerializeSMCreateStbReq(NULL, 0, &createReq);
+  void*   pHead = rpcMallocCont(tlen);
+  tSerializeSMCreateStbReq(pHead, tlen, &createReq);
+  tFreeSMCreateStbReq(&createReq);
+  *pContLen = tlen;
+  return pHead;
+}
+
+void* MndTestSma::BuildDropStbReq(const char* stbname, int32_t* pContLen) {
+  SMDropStbReq dropstbReq = {0};
+  strcpy(dropstbReq.name, stbname);
+
+  int32_t contLen = tSerializeSMDropStbReq(NULL, 0, &dropstbReq);
+  void*   pReq = rpcMallocCont(contLen);
+  tSerializeSMDropStbReq(pReq, contLen, &dropstbReq);
+
+  *pContLen = contLen;
+  return pReq;
+}
+
+void* MndTestSma::BuildCreateSmaReq(const char* smaname, const char* stbname, int8_t igExists, const char* expr,
+                                    const char* tagsFilter, const char* sql, const char* ast, int32_t* pContLen) {
+  SMCreateSmaReq createReq = {0};
+  strcpy(createReq.name, smaname);
+  strcpy(createReq.stb, stbname);
+  createReq.igExists = igExists;
+  createReq.intervalUnit = 1;
+  createReq.slidingUnit = 2;
+  createReq.timezone = 3;
+  createReq.dstVgId = 4;
+  createReq.interval = 10;
+  createReq.offset = 5;
+  createReq.sliding = 6;
+  createReq.expr = (char*)expr;
+  createReq.exprLen = strlen(createReq.expr) + 1;
+  createReq.tagsFilter = (char*)tagsFilter;
+  createReq.tagsFilterLen = strlen(createReq.tagsFilter) + 1;
+  createReq.sql = (char*)sql;
+  createReq.sqlLen = strlen(createReq.sql) + 1;
+  createReq.ast = (char*)ast;
+  createReq.astLen = strlen(createReq.ast) + 1;
+
+  int32_t tlen = tSerializeSMCreateSmaReq(NULL, 0, &createReq);
+  void*   pHead = rpcMallocCont(tlen);
+  tSerializeSMCreateSmaReq(pHead, tlen, &createReq);
+  *pContLen = tlen;
+  return pHead;
+}
+
+void* MndTestSma::BuildDropSmaReq(const char* smaname, int8_t igNotExists, int32_t* pContLen) {
+  SMDropSmaReq dropsmaReq = {0};
+  dropsmaReq.igNotExists = igNotExists;
+  strcpy(dropsmaReq.name, smaname);
+
+  int32_t contLen = tSerializeSMDropSmaReq(NULL, 0, &dropsmaReq);
+  void*   pReq = rpcMallocCont(contLen);
+  tSerializeSMDropSmaReq(pReq, contLen, &dropsmaReq);
+
+  *pContLen = contLen;
+  return pReq;
+}
+
+TEST_F(MndTestSma, 01_Create_Show_Meta_Drop_Restart_Stb) {
+  const char* dbname = "1.d1";
+  const char* stbname = "1.d1.stb";
+  const char* smaname = "1.d1.sma";
+  int32_t     contLen = 0;
+  void*       pReq;
+  SRpcMsg*    pRsp;
+
+  {
+    pReq = BuildCreateDbReq(dbname, &contLen);
+    pRsp = test.SendReq(TDMT_MND_CREATE_DB, pReq, contLen);
+    ASSERT_EQ(pRsp->code, 0);
+  }
+
+  {
+    pReq = BuildCreateStbReq(stbname, &contLen);
+    pRsp = test.SendReq(TDMT_MND_CREATE_STB, pReq, contLen);
+    ASSERT_EQ(pRsp->code, 0);
+    test.SendShowMetaReq(TSDB_MGMT_TABLE_STB, dbname);
+    test.SendShowRetrieveReq();
+    EXPECT_EQ(test.GetShowRows(), 1);
+  }
+#if 0
+  {
+    pReq = BuildCreateSmaReq(smaname, stbname, 0, "expr", "tagsFilter", "sql", "ast", &contLen);
+    pRsp = test.SendReq(TDMT_MND_CREATE_SMA, pReq, contLen);
+    ASSERT_EQ(pRsp->code, 0);
+    test.SendShowMetaReq(TSDB_MGMT_TABLE_INDEX, dbname);
+    test.SendShowRetrieveReq();
+    EXPECT_EQ(test.GetShowRows(), 1);
+  }
+
+  // restart
+  test.Restart();
+
+  {
+    test.SendShowMetaReq(TSDB_MGMT_TABLE_INDEX, dbname);
+    CHECK_META("show indexes", 3);
+    test.SendShowRetrieveReq();
+    EXPECT_EQ(test.GetShowRows(), 1);
+
+    CheckBinary("sma", TSDB_INDEX_NAME_LEN);
+    CheckTimestamp();
+    CheckBinary("stb", TSDB_TABLE_NAME_LEN);
+  }
+
+  {
+     pReq = BuildDropSmaReq(smaname, 0, &contLen);
+    pRsp = test.SendReq(TDMT_MND_DROP_SMA, pReq, contLen);
+    ASSERT_EQ(pRsp->code, 0);
+    test.SendShowMetaReq(TSDB_MGMT_TABLE_INDEX, dbname);
+    test.SendShowRetrieveReq();
+    EXPECT_EQ(test.GetShowRows(), 0);
+  }
+#endif  
+}

source/dnode/vnode/inc/meta.h

@@ -51,7 +51,7 @@ int     metaCreateTable(SMeta *pMeta, STbCfg *pTbCfg);
 int     metaDropTable(SMeta *pMeta, tb_uid_t uid);
 int     metaCommit(SMeta *pMeta);
 int32_t metaCreateTSma(SMeta *pMeta, SSmaCfg *pCfg);
-int32_t metaDropTSma(SMeta *pMeta, char *indexName);
+int32_t metaDropTSma(SMeta *pMeta, int64_t indexUid);
 
 // For Query
 STbCfg *        metaGetTbInfoByUid(SMeta *pMeta, tb_uid_t uid);

source/dnode/vnode/inc/tsdb.h

@@ -96,6 +96,7 @@ int    tsdbCommit(STsdb *pTsdb);
  */
 int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg);
 int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, int8_t smaType, char *msg);
+int32_t tsdbDropTSmaData(STsdb *pTsdb, int64_t indexUid);
 
 /**
  * @brief Insert RSma(Time-range-wise Rollup SMA) data.

source/dnode/vnode/src/inc/metaDef.h

@@ -34,7 +34,7 @@ void metaCloseDB(SMeta* pMeta);
 int  metaSaveTableToDB(SMeta* pMeta, STbCfg* pTbCfg);
 int  metaRemoveTableFromDb(SMeta* pMeta, tb_uid_t uid);
 int  metaSaveSmaToDB(SMeta* pMeta, STSma* pTbCfg);
-int  metaRemoveSmaFromDb(SMeta* pMeta, const char* indexName);
+int  metaRemoveSmaFromDb(SMeta* pMeta, int64_t indexUid);
 
 // SMetaCache
 int  metaOpenCache(SMeta* pMeta);

source/dnode/vnode/src/inc/tqInt.h

@@ -167,6 +167,7 @@ struct STQ {
   STqMetaStore* tqMeta;
   STqPushMgr*   tqPushMgr;
   SHashObj*     pStreamTasks;
+  SVnode*       pVnode;
   SWal*         pWal;
   SMeta*        pVnodeMeta;
 };

source/dnode/vnode/src/inc/tsdbSma.h

@@ -16,15 +16,15 @@
 #ifndef _TD_TSDB_SMA_H_
 #define _TD_TSDB_SMA_H_
 
-typedef struct SSmaStat SSmaStat;
-typedef struct SSmaEnv  SSmaEnv;
+typedef struct SSmaStat     SSmaStat;
+typedef struct SSmaEnv      SSmaEnv;
 
 struct SSmaEnv {
   TdThreadRwlock lock;
-  SDiskID          did;
-  TDBEnv           dbEnv;  // TODO: If it's better to put it in smaIndex level?
-  char *           path;   // relative path
-  SSmaStat *       pStat;
+  SDiskID        did;
+  TDBEnv         dbEnv;  // TODO: If it's better to put it in smaIndex level?
+  char          *path;   // relative path
+  SSmaStat      *pStat;
 };
 
 #define SMA_ENV_LOCK(env)       ((env)->lock)

source/dnode/vnode/src/inc/vnd.h

@@ -82,12 +82,6 @@ struct SVnode {
 
 int vnodeScheduleTask(SVnodeTask* task);
 
-int32_t vnodePutToVQueryQ(SVnode* pVnode, struct SRpcMsg* pReq);
-int32_t vnodePutToVFetchQ(SVnode* pVnode, struct SRpcMsg* pReq);
-int32_t vnodeSendReq(SVnode* pVnode, struct SEpSet* epSet, struct SRpcMsg* pReq);
-int32_t vnodeSendMnodeReq(SVnode* pVnode, struct SRpcMsg* pReq);
-void    vnodeSendRsp(SVnode* pVnode, struct SEpSet* epSet, struct SRpcMsg* pRsp);
-
 #define vFatal(...)                                              \
   do {                                                           \
     if (vDebugFlag & DEBUG_FATAL) {                              \
@@ -177,19 +171,20 @@ int  tqInit();
 void tqCleanUp();
 
 // open in each vnode
-STQ* tqOpen(const char* path, SWal* pWal, SMeta* pMeta, STqCfg* tqConfig, SMemAllocatorFactory* allocFac);
+STQ* tqOpen(const char* path, SVnode* pVnode, SWal* pWal, SMeta* pMeta, STqCfg* tqConfig,
+            SMemAllocatorFactory* allocFac);
 void tqClose(STQ*);
 
 // required by vnode
-int tqPushMsg(STQ*, void* msg, tmsg_t msgType, int64_t version);
+int tqPushMsg(STQ*, void* msg, int32_t msgLen, tmsg_t msgType, int64_t version);
 int tqCommit(STQ*);
 
 int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg);
 int32_t tqProcessSetConnReq(STQ* pTq, char* msg);
 int32_t tqProcessRebReq(STQ* pTq, char* msg);
 int32_t tqProcessTaskExec(STQ* pTq, SRpcMsg* msg);
-
 int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen);
+int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen);
 
 #ifdef __cplusplus
 }

source/dnode/vnode/src/meta/metaBDBImpl.c

@@ -259,7 +259,7 @@ int metaSaveSmaToDB(SMeta *pMeta, STSma *pSmaCfg) {
   return 0;
 }
 
-int metaRemoveSmaFromDb(SMeta *pMeta, const char *indexName) {
+int metaRemoveSmaFromDb(SMeta *pMeta,  int64_t indexUid) {
   // TODO
 #if 0
   DBT key = {0};

source/dnode/vnode/src/meta/metaIdx.c

@@ -121,11 +121,11 @@ int32_t metaCreateTSma(SMeta *pMeta, SSmaCfg *pCfg) {
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t metaDropTSma(SMeta *pMeta, char* indexName) {
+int32_t metaDropTSma(SMeta *pMeta, int64_t indexUid) {
   // TODO: Validate the cfg
   // TODO: add atomicity
 
-  if (metaRemoveSmaFromDb(pMeta, indexName) < 0) {
+  if (metaRemoveSmaFromDb(pMeta, indexUid) < 0) {
     // TODO: handle error
     return -1;
   }

source/dnode/vnode/src/tq/tq.c

@@ -17,11 +17,14 @@
 #include "tqInt.h"
 #include "tqMetaStore.h"
 
+void tqDebugShowSSData(SArray* dataBlocks);
+
 int32_t tqInit() { return tqPushMgrInit(); }
 
 void tqCleanUp() { tqPushMgrCleanUp(); }
 
-STQ* tqOpen(const char* path, SWal* pWal, SMeta* pVnodeMeta, STqCfg* tqConfig, SMemAllocatorFactory* allocFac) {
+STQ* tqOpen(const char* path, SVnode* pVnode, SWal* pWal, SMeta* pVnodeMeta, STqCfg* tqConfig,
+            SMemAllocatorFactory* allocFac) {
   STQ* pTq = malloc(sizeof(STQ));
   if (pTq == NULL) {
     terrno = TSDB_CODE_TQ_OUT_OF_MEMORY;
@@ -29,6 +32,7 @@ STQ* tqOpen(const char* path, SWal* pWal, SMeta* pVnodeMeta, STqCfg* tqConfig, S
   }
   pTq->path = strdup(path);
   pTq->tqConfig = tqConfig;
+  pTq->pVnode = pVnode;
   pTq->pWal = pWal;
   pTq->pVnodeMeta = pVnodeMeta;
 #if 0
@@ -68,46 +72,19 @@ void tqClose(STQ* pTq) {
   // TODO
 }
 
-int tqPushMsg(STQ* pTq, void* msg, tmsg_t msgType, int64_t version) {
+int tqPushMsg(STQ* pTq, void* msg, int32_t msgLen, tmsg_t msgType, int64_t version) {
   if (msgType != TDMT_VND_SUBMIT) return 0;
-
-  void* pIter = NULL;
-
-  while (1) {
-    pIter = taosHashIterate(pTq->pStreamTasks, pIter);
-    if (pIter == NULL) break;
-    SStreamTask* pTask = (SStreamTask*)pIter;
-    if (!pTask->pipeSource) continue;
-
-    int32_t workerId = 0;
-    void*   exec = pTask->runner[workerId].executor;
-    qSetStreamInput(exec, msg, STREAM_DATA_TYPE_SUBMIT_BLOCK);
-    SArray* pRes = taosArrayInit(0, sizeof(SSDataBlock));
-    while (1) {
-      SSDataBlock* output;
-      uint64_t     ts;
-      if (qExecTask(exec, &output, &ts) < 0) {
-        ASSERT(false);
-      }
-      if (output == NULL) {
-        break;
-      }
-      taosArrayPush(pRes, output);
-    }
-    if (pTask->pipeSink) {
-      // write back
-    } else {
-      int32_t tlen = sizeof(SStreamExecMsgHead) + tEncodeDataBlocks(NULL, pRes);
-      void*   buf = rpcMallocCont(tlen);
-      if (buf == NULL) {
-        return -1;
-      }
-      void* abuf = POINTER_SHIFT(buf, sizeof(SStreamExecMsgHead));
-      tEncodeDataBlocks(abuf, pRes);
-      // serialize
-      // to next level
-    }
+  void* data = malloc(msgLen);
+  if (data == NULL) {
+    return -1;
   }
+  memcpy(data, msg, msgLen);
+  SRpcMsg req = {
+      .msgType = TDMT_VND_STREAM_TRIGGER,
+      .pCont = data,
+      .contLen = msgLen,
+  };
+  tmsgPutToQueue(&pTq->pVnode->msgCb, FETCH_QUEUE, &req);
 
 #if 0
   void* pIter = taosHashIterate(pTq->tqPushMgr->pHash, NULL);
@@ -483,7 +460,9 @@ int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
   }
   SCoder decoder;
   tCoderInit(&decoder, TD_LITTLE_ENDIAN, (uint8_t*)msg, msgLen, TD_DECODER);
-  tDecodeSStreamTask(&decoder, pTask);
+  if (tDecodeSStreamTask(&decoder, pTask) < 0) {
+    ASSERT(0);
+  }
   tCoderClear(&decoder);
 
   tqExpandTask(pTq, pTask, 8);
@@ -560,7 +539,11 @@ void tqDebugShowSSData(SArray* dataBlocks) {
             break;
           case TSDB_DATA_TYPE_INT:
           case TSDB_DATA_TYPE_UINT:
-            printf(" %15u |", *(uint32_t*)var);
+            printf(" %15d |", *(int32_t*)var);
+            break;
+          case TSDB_DATA_TYPE_BIGINT:
+          case TSDB_DATA_TYPE_UBIGINT:
+            printf(" %15ld |", *(int64_t*)var);
             break;
         }
       }
@@ -569,6 +552,62 @@ void tqDebugShowSSData(SArray* dataBlocks) {
   }
 }
 
+int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen) {
+  void* pIter = NULL;
+
+  while (1) {
+    pIter = taosHashIterate(pTq->pStreamTasks, pIter);
+    if (pIter == NULL) break;
+    SStreamTask* pTask = (SStreamTask*)pIter;
+    if (!pTask->pipeSource) continue;
+
+    int32_t workerId = 0;
+    void*   exec = pTask->runner[workerId].executor;
+    qSetStreamInput(exec, data, STREAM_DATA_TYPE_SUBMIT_BLOCK);
+    SArray* pRes = taosArrayInit(0, sizeof(SSDataBlock));
+    while (1) {
+      SSDataBlock* output;
+      uint64_t     ts;
+      if (qExecTask(exec, &output, &ts) < 0) {
+        ASSERT(false);
+      }
+      if (output == NULL) {
+        break;
+      }
+      taosArrayPush(pRes, output);
+    }
+    if (pTask->pipeSink) {
+      // write back
+      /*printf("reach end\n");*/
+      tqDebugShowSSData(pRes);
+    } else {
+      int32_t tlen = sizeof(SStreamExecMsgHead) + tEncodeDataBlocks(NULL, pRes);
+      void*   buf = rpcMallocCont(tlen);
+      if (buf == NULL) {
+        return -1;
+      }
+      void* abuf = POINTER_SHIFT(buf, sizeof(SStreamExecMsgHead));
+      tEncodeDataBlocks(abuf, pRes);
+      tmsg_t type;
+
+      if (pTask->nextOpDst == STREAM_NEXT_OP_DST__VND) {
+        type = TDMT_VND_TASK_EXEC;
+      } else {
+        type = TDMT_SND_TASK_EXEC;
+      }
+
+      SRpcMsg reqMsg = {
+          .pCont = buf,
+          .contLen = tlen,
+          .code = 0,
+          .msgType = type,
+      };
+      tmsgSendReq(&pTq->pVnode->msgCb, &pTask->NextOpEp, &reqMsg);
+    }
+  }
+  return 0;
+}
+
 int32_t tqProcessTaskExec(STQ* pTq, SRpcMsg* msg) {
   SStreamTaskExecReq* pReq = msg->pCont;
 

source/dnode/vnode/src/tq/tqRead.c

@@ -126,6 +126,36 @@ SArray* tqRetrieveDataBlock(STqReadHandle* pHandle) {
   if (pArray == NULL) {
     return NULL;
   }
+  int32_t colMeta = 0;
+  int32_t colNeed = 0;
+  while (colMeta < pSchemaWrapper->nCols && colNeed < colNumNeed) {
+    SSchema* pColSchema = &pSchemaWrapper->pSchema[colMeta];
+    int16_t  colIdSchema = pColSchema->colId;
+    int16_t  colIdNeed = *(int16_t*)taosArrayGet(pHandle->pColIdList, colNeed);
+    if (colIdSchema < colIdNeed) {
+      colMeta++;
+    } else if (colIdSchema > colIdNeed) {
+      colNeed++;
+    } else {
+      SColumnInfoData colInfo = {0};
+      int             sz = numOfRows * pColSchema->bytes;
+      colInfo.info.bytes = pColSchema->bytes;
+      colInfo.info.colId = pColSchema->colId;
+      colInfo.info.type = pColSchema->type;
+
+      colInfo.pData = calloc(1, sz);
+      if (colInfo.pData == NULL) {
+        // TODO free
+        taosArrayDestroy(pArray);
+        return NULL;
+      }
+
+      blockDataEnsureColumnCapacity(&colInfo, numOfRows);
+      taosArrayPush(pArray, &colInfo);
+      colMeta++;
+      colNeed++;
+    }
+  }
 
   int j = 0;
   for (int32_t i = 0; i < colNumNeed; i++) {
@@ -163,11 +193,23 @@ SArray* tqRetrieveDataBlock(STqReadHandle* pHandle) {
   while ((row = tGetSubmitBlkNext(&pHandle->blkIter)) != NULL) {
     tdSTSRowIterReset(&iter, row);
     // get all wanted col of that block
+    int32_t colTot = taosArrayGetSize(pArray);
+    for (int32_t i = 0; i < colTot; i++) {
+      SColumnInfoData* pColData = taosArrayGet(pArray, i);
+      SCellVal         sVal = {0};
+      if (!tdSTSRowIterNext(&iter, pColData->info.colId, pColData->info.type, &sVal)) {
+        break;
+      }
+      memcpy(POINTER_SHIFT(pColData->pData, curRow * pColData->info.bytes), sVal.val, pColData->info.bytes);
+    }
+#if 0
     for (int32_t i = 0; i < colNumNeed; i++) {
       SColumnInfoData* pColData = taosArrayGet(pArray, i);
       STColumn*        pCol = schemaColAt(pTschema, i);
       // TODO
-      ASSERT(pCol->colId == pColData->info.colId);
+      if(pCol->colId != pColData->info.colId) {
+        continue;
+      }
       // void* val = tdGetMemRowDataOfColEx(row, pCol->colId, pCol->type, TD_DATA_ROW_HEAD_SIZE + pCol->offset, &kvIdx);
       SCellVal sVal = {0};
       if (!tdSTSRowIterNext(&iter, pCol->colId, pCol->type, &sVal)) {
@@ -176,6 +218,7 @@ SArray* tqRetrieveDataBlock(STqReadHandle* pHandle) {
       }
       memcpy(POINTER_SHIFT(pColData->pData, curRow * pCol->bytes), sVal.val, pCol->bytes);
     }
+#endif
     curRow++;
   }
   return pArray;

source/dnode/vnode/src/tsdb/tsdbRead.c

@@ -403,18 +403,12 @@ static STsdbReadHandle* tsdbQueryTablesImpl(STsdb* tsdb, STsdbQueryCond* pCond,
       SColumnInfoData colInfo = {{0}, 0};
       colInfo.info = pCond->colList[i];
 
-      colInfo.pData = calloc(1, EXTRA_BYTES + pReadHandle->outputCapacity * pCond->colList[i].bytes);
-      if (!IS_VAR_DATA_TYPE(colInfo.info.type)) {
-        colInfo.nullbitmap = calloc(1, BitmapLen(pReadHandle->outputCapacity));
-      }
-
-      if (colInfo.pData == NULL || (colInfo.nullbitmap == NULL && (!IS_VAR_DATA_TYPE(colInfo.info.type)))) {
+      int32_t code = blockDataEnsureColumnCapacity(&colInfo, pReadHandle->outputCapacity);
+      if (code != TSDB_CODE_SUCCESS) {
         goto _end;
       }
 
       taosArrayPush(pReadHandle->pColumns, &colInfo);
-
-
       pReadHandle->statis[i].colId = colInfo.info.colId;
     }
 
@@ -1383,7 +1377,6 @@ static int doBinarySearchKey(char* pValue, int num, TSKEY key, int order) {
 }
 
 static int32_t doCopyRowsFromFileBlock(STsdbReadHandle* pTsdbReadHandle, int32_t capacity, int32_t numOfRows, int32_t start, int32_t end) {
-  char* pData = NULL;
   int32_t step = ASCENDING_TRAVERSE(pTsdbReadHandle->order)? 1 : -1;
 
   SDataCols* pCols = pTsdbReadHandle->rhelper.pDCols[0];
@@ -1409,43 +1402,38 @@ static int32_t doCopyRowsFromFileBlock(STsdbReadHandle* pTsdbReadHandle, int32_t
       continue;
     }
 
-    int32_t bytes = pColInfo->info.bytes;
-
-    if (ASCENDING_TRAVERSE(pTsdbReadHandle->order)) {
-      pData = (char*)pColInfo->pData + numOfRows * pColInfo->info.bytes;
-    } else {
-      pData = (char*)pColInfo->pData + (capacity - numOfRows - num) * pColInfo->info.bytes;
-    }
-
     if (!isAllRowsNull(src) && pColInfo->info.colId == src->colId) {
-      if (pColInfo->info.type != TSDB_DATA_TYPE_BINARY && pColInfo->info.type != TSDB_DATA_TYPE_NCHAR) {
-        memmove(pData, (char*)src->pData + bytes * start, bytes * num);
-      } else {  // handle the var-string
-        char* dst = pData;
+      if (!IS_VAR_DATA_TYPE(pColInfo->info.type)) {  // todo opt performance
+//        memmove(pData, (char*)src->pData + bytes * start, bytes * num);
+        for(int32_t k = start; k < num + start; ++k) {
+          SCellVal sVal = {0};
+          if (tdGetColDataOfRow(&sVal, src, k) < 0) {
+            TASSERT(0);
+          }
 
+          if (sVal.valType == TD_VTYPE_NULL) {
+            colDataAppend(pColInfo, k, NULL, true);
+          } else {
+            colDataAppend(pColInfo, k, sVal.val, false);
+          }
+        }
+      } else {  // handle the var-string
         // todo refactor, only copy one-by-one
         for (int32_t k = start; k < num + start; ++k) {
-          SCellVal    sVal = {0};
+          SCellVal sVal = {0};
           if(tdGetColDataOfRow(&sVal, src, k) < 0){
             TASSERT(0);
           }
-          memcpy(dst, sVal.val, varDataTLen(sVal.val));
-          dst += bytes;
+
+          colDataAppend(pColInfo, k, sVal.val, false);
         }
       }
 
       j++;
       i++;
     } else { // pColInfo->info.colId < src->colId, it is a NULL data
-      if (pColInfo->info.type == TSDB_DATA_TYPE_BINARY || pColInfo->info.type == TSDB_DATA_TYPE_NCHAR) {
-        char* dst = pData;
-
-        for(int32_t k = start; k < num + start; ++k) {
-          setVardataNull(dst, pColInfo->info.type);
-          dst += bytes;
-        }
-      } else {
-        setNullN(pData, pColInfo->info.type, pColInfo->info.bytes, num);
+      for(int32_t k = start; k < num + start; ++k) {  // TODO opt performance
+        colDataAppend(pColInfo, k, NULL, true);
       }
       i++;
     }
@@ -1453,23 +1441,9 @@ static int32_t doCopyRowsFromFileBlock(STsdbReadHandle* pTsdbReadHandle, int32_t
 
   while (i < requiredNumOfCols) { // the remain columns are all null data
     SColumnInfoData* pColInfo = taosArrayGet(pTsdbReadHandle->pColumns, i);
-    if (ASCENDING_TRAVERSE(pTsdbReadHandle->order)) {
-      pData = (char*)pColInfo->pData + numOfRows * pColInfo->info.bytes;
-    } else {
-      pData = (char*)pColInfo->pData + (capacity - numOfRows - num) * pColInfo->info.bytes;
-    }
-
-    if (pColInfo->info.type == TSDB_DATA_TYPE_BINARY || pColInfo->info.type == TSDB_DATA_TYPE_NCHAR) {
-      char* dst = pData;
-
-      for(int32_t k = start; k < num + start; ++k) {
-        setVardataNull(dst, pColInfo->info.type);
-        dst += pColInfo->info.bytes;
-      }
-    } else {
-      setNullN(pData, pColInfo->info.type, pColInfo->info.bytes, num);
+    for(int32_t k = start; k < num + start; ++k) {
+      colDataAppend(pColInfo, k, NULL, true); // TODO add a fast version to set a number of consecutive NULL value.
     }
-
     i++;
   }
 
@@ -1479,14 +1453,12 @@ static int32_t doCopyRowsFromFileBlock(STsdbReadHandle* pTsdbReadHandle, int32_t
   return numOfRows + num;
 }
 
-// TODO fix bug for reverse copy data
-// TODO handle the null data
+// TODO fix bug for reverse copy data problem
 // Note: row1 always has high priority
 static void mergeTwoRowFromMem(STsdbReadHandle* pTsdbReadHandle, int32_t capacity, int32_t numOfRows, STSRow* row1,
                                STSRow* row2, int32_t numOfCols, uint64_t uid, STSchema* pSchema1, STSchema* pSchema2,
                                bool forceSetNull) {
 #if 1
-  char*       pData = NULL;
   STSchema*   pSchema;
   STSRow*     row;
   int16_t     colId;
@@ -1528,12 +1500,6 @@ static void mergeTwoRowFromMem(STsdbReadHandle* pTsdbReadHandle, int32_t capacit
   while(i < numOfCols && (j < numOfColsOfRow1 || k < numOfColsOfRow2)) {
     SColumnInfoData* pColInfo = taosArrayGet(pTsdbReadHandle->pColumns, i);
 
-    if (ASCENDING_TRAVERSE(pTsdbReadHandle->order)) {
-      pData = (char*)pColInfo->pData + numOfRows * pColInfo->info.bytes;
-    } else {
-      pData = (char*)pColInfo->pData + (capacity - numOfRows - 1) * pColInfo->info.bytes;
-    }
-
     int32_t colIdOfRow1;
     if(j >= numOfColsOfRow1) {
       colIdOfRow1 = INT32_MAX;
@@ -1596,43 +1562,11 @@ static void mergeTwoRowFromMem(STsdbReadHandle* pTsdbReadHandle, int32_t capacit
 
     if (colId == pColInfo->info.colId) {
       if (tdValTypeIsNorm(sVal.valType)) {
-        switch (pColInfo->info.type) {
-          case TSDB_DATA_TYPE_BINARY:
-          case TSDB_DATA_TYPE_NCHAR:
-            memcpy(pData, sVal.val, varDataTLen(sVal.val));
-            break;
-          case TSDB_DATA_TYPE_BOOL:
-          case TSDB_DATA_TYPE_TINYINT:
-          case TSDB_DATA_TYPE_UTINYINT:
-            *(uint8_t *)pData = *(uint8_t *)sVal.val;
-            break;
-          case TSDB_DATA_TYPE_SMALLINT:
-          case TSDB_DATA_TYPE_USMALLINT:
-            *(uint16_t *)pData = *(uint16_t *)sVal.val;
-            break;
-          case TSDB_DATA_TYPE_INT:
-          case TSDB_DATA_TYPE_UINT:
-            *(uint32_t *)pData = *(uint32_t *)sVal.val;
-            break;
-          case TSDB_DATA_TYPE_BIGINT:
-          case TSDB_DATA_TYPE_UBIGINT:
-            *(uint64_t *)pData = *(uint64_t *)sVal.val;
-            break;
-          case TSDB_DATA_TYPE_FLOAT:
-            SET_FLOAT_PTR(pData, sVal.val);
-            break;
-          case TSDB_DATA_TYPE_DOUBLE:
-            SET_DOUBLE_PTR(pData, sVal.val);
-            break;
-          case TSDB_DATA_TYPE_TIMESTAMP:
-            *(TSKEY*)pData = *(TSKEY*)sVal.val;
-            break;
-          default:
-            memcpy(pData, sVal.val, pColInfo->info.bytes);
-        }
+        colDataAppend(pColInfo, numOfRows, sVal.val, false);
       } else if (forceSetNull) {
         colDataAppend(pColInfo, numOfRows, NULL, true);
       }
+
       i++;
 
       if(row == row1) {

source/dnode/vnode/src/tsdb/tsdbSma.c

@@ -25,6 +25,7 @@ static const char *TSDB_SMA_DNAME[] = {
 #define SMA_STORAGE_TSDB_TIMES  10
 #define SMA_STORAGE_SPLIT_HOURS 24
 #define SMA_KEY_LEN             18  // tableUid_colId_TSKEY 8+2+8
+#define SMA_DROP_EXPIRED_TIME   10  // default is 10 seconds
 
 #define SMA_STATE_HASH_SLOT      4
 #define SMA_STATE_ITEM_HASH_SLOT 32
@@ -60,10 +61,11 @@ typedef struct {
 typedef struct {
   /**
    * @brief The field 'state' is here to demonstrate if one smaIndex is ready to provide service.
-   *    - TSDB_SMA_STAT_EXPIRED: 1) If sma calculation of history TS data is not finished; 2) Or if the TSDB is open,
-   * without information about its previous state.
    *    - TSDB_SMA_STAT_OK: 1) The sma calculation of history data is finished; 2) Or recevied information from
    * Streaming Module or TSDB local persistence.
+   *    - TSDB_SMA_STAT_EXPIRED: 1) If sma calculation of history TS data is not finished; 2) Or if the TSDB is open,
+   * without information about its previous state.
+   *    - TSDB_SMA_STAT_DROPPED: 1)sma dropped
    */
   int8_t    state;           // ETsdbSmaStat
   SHashObj *expiredWindows;  // key: skey of time window, value: N/A
@@ -80,6 +82,7 @@ struct SSmaStat {
 // expired window
 static int32_t  tsdbUpdateExpiredWindow(STsdb *pTsdb, ETsdbSmaType smaType, char *msg);
 static int32_t  tsdbInitSmaStat(SSmaStat **pSmaStat);
+static void *   tsdbFreeSmaStatItem(SSmaStatItem *pSmaStatItem);
 static int32_t  tsdbDestroySmaState(SSmaStat *pSmaStat);
 static SSmaEnv *tsdbNewSmaEnv(const STsdb *pTsdb, const char *path, SDiskID did);
 static int32_t  tsdbInitSmaEnv(STsdb *pTsdb, const char *path, SDiskID did, SSmaEnv **pEnv);
@@ -109,7 +112,55 @@ static void    tsdbGetSmaDir(int32_t vgId, ETsdbSmaType smaType, char dirName[])
 static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg);
 static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg);
 
+// mgmt interface
+static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid);
+
 // implementation
+static FORCE_INLINE int8_t tsdbSmaStat(SSmaStatItem *pStatItem) {
+  if (pStatItem) {
+    return atomic_load_8(&pStatItem->state);
+  }
+  return TSDB_SMA_STAT_UNKNOWN;
+}
+
+static FORCE_INLINE bool tsdbSmaStatIsOK(SSmaStatItem *pStatItem, int8_t *state) {
+  if(!pStatItem) {
+    return false;
+  }
+
+  if (state) {
+    *state = atomic_load_8(&pStatItem->state);
+    return *state == TSDB_SMA_STAT_OK;
+  }
+  return atomic_load_8(&pStatItem->state) == TSDB_SMA_STAT_OK;
+}
+
+static FORCE_INLINE bool tsdbSmaStatIsExpired(SSmaStatItem *pStatItem) {
+  return pStatItem ? (atomic_load_8(&pStatItem->state) & TSDB_SMA_STAT_EXPIRED) : true;
+}
+
+static FORCE_INLINE bool tsdbSmaStatIsDropped(SSmaStatItem *pStatItem) {
+  return pStatItem ? (atomic_load_8(&pStatItem->state) & TSDB_SMA_STAT_DROPPED) : true;
+}
+
+static FORCE_INLINE void tsdbSmaStatSetOK(SSmaStatItem *pStatItem) {
+  if (pStatItem) {
+    atomic_store_8(&pStatItem->state, TSDB_SMA_STAT_OK);
+  }
+}
+
+static FORCE_INLINE void tsdbSmaStatSetExpired(SSmaStatItem *pStatItem) {
+  if (pStatItem) {
+    atomic_or_fetch_8(&pStatItem->state, TSDB_SMA_STAT_EXPIRED);
+  }
+}
+
+static FORCE_INLINE void tsdbSmaStatSetDropped(SSmaStatItem *pStatItem) {
+  if (pStatItem) {
+    atomic_or_fetch_8(&pStatItem->state, TSDB_SMA_STAT_DROPPED);
+  }
+}
+
 static void tsdbGetSmaDir(int32_t vgId, ETsdbSmaType smaType, char dirName[]) {
   snprintf(dirName, TSDB_FILENAME_LEN, "vnode%svnode%d%stsdb%s%s", TD_DIRSEP, vgId, TD_DIRSEP, TD_DIRSEP,
            TSDB_SMA_DNAME[smaType]);
@@ -252,6 +303,16 @@ static SSmaStatItem *tsdbNewSmaStatItem(int8_t state) {
   return pItem;
 }
 
+static void *tsdbFreeSmaStatItem(SSmaStatItem *pSmaStatItem) {
+  if (pSmaStatItem != NULL) {
+    tdDestroyTSma(pSmaStatItem->pSma);
+    tfree(pSmaStatItem->pSma);
+    taosHashCleanup(pSmaStatItem->expiredWindows);
+    tfree(pSmaStatItem);
+  }
+  return NULL;
+}
+
 /**
  * @brief Release resources allocated for its member fields, not including itself.
  *
@@ -264,12 +325,7 @@ int32_t tsdbDestroySmaState(SSmaStat *pSmaStat) {
     void *item = taosHashIterate(pSmaStat->smaStatItems, NULL);
     while (item != NULL) {
       SSmaStatItem *pItem = *(SSmaStatItem **)item;
-      if (pItem != NULL) {
-        tdDestroyTSma(pItem->pSma);
-        tfree(pItem->pSma);
-        taosHashCleanup(pItem->expiredWindows);
-        tfree(pItem);
-      }
+      tsdbFreeSmaStatItem(pItem);
       item = taosHashIterate(pSmaStat->smaStatItems, item);
     }
     taosHashCleanup(pSmaStat->smaStatItems);
@@ -437,6 +493,15 @@ static int32_t tsdbResetExpiredWindow(STsdb *pTsdb, SSmaStat *pStat, int64_t ind
                skey, indexUid);
       return TSDB_CODE_FAILED;
     }
+    // TODO: use a standalone interface to received state upate notification from stream computing module.
+    /**
+     * @brief state
+     *  - When SMA env init in TSDB, its status is TSDB_SMA_STAT_OK.
+     *  - In startup phase of stream computing module, it should notify the SMA env in TSDB to expired if needed(e.g.
+     * when batch data caculation not finised)
+     *  - When TSDB_SMA_STAT_OK, the stream computing module should also notify that to the SMA env in TSDB.
+     */
+    pItem->state = TSDB_SMA_STAT_OK;
   } else {
     // error handling
     tsdbUnRefSmaStat(pTsdb, pStat);
@@ -711,6 +776,8 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
   STsdbCfg *        pCfg = REPO_CFG(pTsdb);
   STSmaDataWrapper *pData = (STSmaDataWrapper *)msg;
   SSmaEnv *         pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
+  int64_t indexUid = SMA_TEST_INDEX_UID;
+
 
   if (pEnv == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
@@ -730,13 +797,28 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
     return TSDB_CODE_FAILED;
   }
 
-  int64_t indexUid = SMA_TEST_INDEX_UID;
+  SSmaStat *pStat = SMA_ENV_STAT(pTsdb->pTSmaEnv);
+  SSmaStatItem *pItem = NULL;
+
+  tsdbRefSmaStat(pTsdb, pStat);
+
+  if (pStat && pStat->smaStatItems) {
+    pItem = taosHashGet(pStat->smaStatItems, &indexUid, sizeof(indexUid));
+  }
+
+  if ((pItem == NULL) || ((pItem = *(SSmaStatItem **)pItem) == NULL) || tsdbSmaStatIsDropped(pItem)) {
+    terrno = TSDB_CODE_TDB_INVALID_SMA_STAT;
+    tsdbUnRefSmaStat(pTsdb, pStat);
+    return TSDB_CODE_FAILED;
+  }
+
   char    rPath[TSDB_FILENAME_LEN] = {0};
   char    aPath[TSDB_FILENAME_LEN] = {0};
   snprintf(rPath, TSDB_FILENAME_LEN, "%s%s%" PRIi64, SMA_ENV_PATH(pEnv), TD_DIRSEP, indexUid);
   tfsAbsoluteName(REPO_TFS(pTsdb), SMA_ENV_DID(pEnv), rPath, aPath);
   if (!taosCheckExistFile(aPath)) {
     if (tfsMkdirRecurAt(REPO_TFS(pTsdb), rPath, SMA_ENV_DID(pEnv)) != TSDB_CODE_SUCCESS) {
+      tsdbUnRefSmaStat(pTsdb, pStat);
       return TSDB_CODE_FAILED;
     }
   }
@@ -754,12 +836,14 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
     tsdbWarn("vgId:%d open DB file %s failed since %s", REPO_ID(pTsdb),
              tSmaH.dFile.path ? tSmaH.dFile.path : "path is NULL", tstrerror(terrno));
     tsdbDestroyTSmaWriteH(&tSmaH);
+    tsdbUnRefSmaStat(pTsdb, pStat);
     return TSDB_CODE_FAILED;
   }
 
   if (tsdbInsertTSmaDataSection(&tSmaH, pData) != 0) {
     tsdbWarn("vgId:%d insert tSma data section failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
     tsdbDestroyTSmaWriteH(&tSmaH);
+    tsdbUnRefSmaStat(pTsdb, pStat);
     return TSDB_CODE_FAILED;
   }
   // TODO:tsdbEndTSmaCommit();
@@ -768,9 +852,60 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
   tsdbResetExpiredWindow(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv), pData->indexUid, pData->skey);
 
   tsdbDestroyTSmaWriteH(&tSmaH);
+  tsdbUnRefSmaStat(pTsdb, pStat);
   return TSDB_CODE_SUCCESS;
 }
 
+/**
+ * @brief Drop tSma data and local cache
+ *        - insert/query reference
+ * @param pTsdb 
+ * @param msg 
+ * @return int32_t 
+ */
+static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid) {
+  SSmaEnv *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
+
+  // clear local cache
+  if (pEnv) {
+    tsdbDebug("vgId:%d drop tSma local cache for %" PRIi64, REPO_ID(pTsdb), indexUid);
+
+    SSmaStatItem *pItem = taosHashGet(SMA_ENV_STAT_ITEMS(pEnv), &indexUid, sizeof(indexUid));
+    if ((pItem != NULL) || ((pItem = *(SSmaStatItem **)pItem) != NULL)) {
+      if (tsdbSmaStatIsDropped(pItem)) {
+        tsdbDebug("vgId:%d tSma stat is already dropped for %" PRIi64, REPO_ID(pTsdb), indexUid);
+        return TSDB_CODE_TDB_INVALID_ACTION;  // TODO: duplicate drop msg would be intercepted by mnode
+      }
+
+      tsdbWLockSma(pEnv);
+      if (tsdbSmaStatIsDropped(pItem)) {
+        tsdbUnLockSma(pEnv);
+        tsdbDebug("vgId:%d tSma stat is already dropped for %" PRIi64, REPO_ID(pTsdb), indexUid);
+        return TSDB_CODE_TDB_INVALID_ACTION;  // TODO: duplicate drop msg would be intercepted by mnode
+      }
+      tsdbSmaStatSetDropped(pItem);
+      tsdbUnLockSma(pEnv);
+
+      int32_t nSleep = 0;
+      while (true) {
+        if (T_REF_VAL_GET(SMA_ENV_STAT(pEnv)) <= 0) {
+          break;
+        }
+        taosSsleep(1);
+        if (++nSleep > SMA_DROP_EXPIRED_TIME) {
+          break;
+        };
+      }
+
+      tsdbFreeSmaStatItem(pItem);
+      tsdbDebug("vgId:%d getTSmaDataImpl failed since no index %" PRIi64 " in local cache", REPO_ID(pTsdb), indexUid);
+    }
+  }
+  // clear sma data files
+  // TODO: 
+
+}
+
 static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int32_t fid) {
   STsdb *pTsdb = pSmaH->pTsdb;
 
@@ -784,35 +919,64 @@ static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData,
 static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
   STsdbCfg *        pCfg = REPO_CFG(pTsdb);
   STSmaDataWrapper *pData = (STSmaDataWrapper *)msg;
-  STSmaWriteH       tSmaH = {0};
+  SSmaEnv *         pEnv = atomic_load_ptr(&pTsdb->pRSmaEnv);
 
-  tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData);
+  if (pEnv == NULL) {
+    terrno = TSDB_CODE_INVALID_PTR;
+    tsdbWarn("vgId:%d insert tSma data failed since pTSmaEnv is NULL", REPO_ID(pTsdb));
+    return terrno;
+  }
 
   if (pData->dataLen <= 0) {
     TASSERT(0);
     terrno = TSDB_CODE_INVALID_PARA;
-    return terrno;
+    return TSDB_CODE_FAILED;
   }
 
-  // Step 1: Judge the storage level
-  int32_t storageLevel = tsdbGetSmaStorageLevel(pData->interval, pData->intervalUnit);
-  int32_t daysPerFile = storageLevel == SMA_STORAGE_LEVEL_TSDB ? SMA_STORAGE_TSDB_DAYS : pCfg->daysPerFile;
+  STSmaWriteH tSmaH = {0};
 
-  // Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index file
-  //         - Set and open the DFile or the B+Tree file
+  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData) != 0) {
+    return TSDB_CODE_FAILED;
+  }
+
+  int64_t indexUid = SMA_TEST_INDEX_UID;
+  char    rPath[TSDB_FILENAME_LEN] = {0};
+  char    aPath[TSDB_FILENAME_LEN] = {0};
+  snprintf(rPath, TSDB_FILENAME_LEN, "%s%s%" PRIi64, SMA_ENV_PATH(pEnv), TD_DIRSEP, indexUid);
+  tfsAbsoluteName(REPO_TFS(pTsdb), SMA_ENV_DID(pEnv), rPath, aPath);
+  if (!taosCheckExistFile(aPath)) {
+    if (tfsMkdirRecurAt(REPO_TFS(pTsdb), rPath, SMA_ENV_DID(pEnv)) != TSDB_CODE_SUCCESS) {
+      return TSDB_CODE_FAILED;
+    }
+  }
 
+  // Step 1: Judge the storage level and days
+  int32_t storageLevel = tsdbGetSmaStorageLevel(pData->interval, pData->intervalUnit);
+  int32_t daysPerFile = tsdbGetTSmaDays(pTsdb, tSmaH.interval, storageLevel);
   int32_t fid = (int32_t)(TSDB_KEY_FID(pData->skey, daysPerFile, pCfg->precision));
 
-  // Save all the TSma data to one file
+  // Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index file
+  //         - Set and open the DFile or the B+Tree file
   // TODO: tsdbStartTSmaCommit();
-  tsdbSetTSmaDataFile(&tSmaH, pData, storageLevel, fid);
+  tsdbSetTSmaDataFile(&tSmaH, pData, indexUid, fid);
+  if (tsdbOpenDBF(pTsdb->pTSmaEnv->dbEnv, &tSmaH.dFile) != 0) {
+    tsdbWarn("vgId:%d open DB file %s failed since %s", REPO_ID(pTsdb),
+             tSmaH.dFile.path ? tSmaH.dFile.path : "path is NULL", tstrerror(terrno));
+    tsdbDestroyTSmaWriteH(&tSmaH);
+    return TSDB_CODE_FAILED;
+  }
 
-  tsdbInsertTSmaDataSection(&tSmaH, pData);
+  if (tsdbInsertTSmaDataSection(&tSmaH, pData) != 0) {
+    tsdbWarn("vgId:%d insert tSma data section failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
+    tsdbDestroyTSmaWriteH(&tSmaH);
+    return TSDB_CODE_FAILED;
+  }
   // TODO:tsdbEndTSmaCommit();
 
-  // reset the SSmaStat
-  tsdbResetExpiredWindow(pTsdb, SMA_ENV_STAT(pTsdb->pRSmaEnv), pData->indexUid, pData->skey);
+  // Step 3: reset the SSmaStat
+  tsdbResetExpiredWindow(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv), pData->indexUid, pData->skey);
 
+  tsdbDestroyTSmaWriteH(&tSmaH);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -934,6 +1098,15 @@ static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_
 #endif
 
 #if 1
+  int8_t smaStat = 0;
+  if (!tsdbSmaStatIsOK(pItem, &smaStat)) {  // TODO: multiple check for large scale sma query
+    tsdbUnRefSmaStat(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv));
+    terrno = TSDB_CODE_TDB_INVALID_SMA_STAT;
+    tsdbWarn("vgId:%d getTSmaDataImpl failed from index %" PRIi64 " since %s %" PRIi8, REPO_ID(pTsdb), indexUid,
+             tstrerror(terrno), smaStat);
+    return TSDB_CODE_FAILED;
+  }
+
   if (taosHashGet(pItem->expiredWindows, &querySKey, sizeof(TSKEY)) != NULL) {
     // TODO: mark this window as expired.
     tsdbDebug("vgId:%d skey %" PRIi64 " of window exists in expired window for index %" PRIi64, REPO_ID(pTsdb),
@@ -1086,6 +1259,20 @@ int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
   return code;
 }
 
+/**
+ * @brief Get tSma data
+ * 
+ * @param pTsdb 
+ * @param pData 
+ * @param indexUid 
+ * @param interval 
+ * @param intervalUnit 
+ * @param tableUid 
+ * @param colId 
+ * @param querySKey 
+ * @param nMaxResult 
+ * @return int32_t 
+ */
 int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
                         tb_uid_t tableUid, col_id_t colId, TSKEY querySKey, int32_t nMaxResult) {
   int32_t code = TSDB_CODE_SUCCESS;
@@ -1094,4 +1281,19 @@ int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid,
     tsdbWarn("vgId:%d get tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
   }
   return code;
+}
+
+/**
+ * @brief Drop tSma Data and caches
+ * 
+ * @param pTsdb 
+ * @param msg 
+ * @return int32_t 
+ */
+int32_t tsdbDropTSmaData(STsdb *pTsdb, int64_t indexUid) {
+  int32_t code = TSDB_CODE_SUCCESS;
+  if ((code = tsdbDropTSmaDataImpl(pTsdb, indexUid)) < 0) {
+    tsdbWarn("vgId:%d drop tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
+  }
+  return code;
 }
\ No newline at end of file

source/dnode/vnode/src/vnd/vnodeMain.c

@@ -115,7 +115,8 @@ static int vnodeOpenImpl(SVnode *pVnode) {
 
   // Open tsdb
   sprintf(dir, "%s/tsdb", pVnode->path);
-  pVnode->pTsdb = tsdbOpen(dir, pVnode->vgId, &(pVnode->config.tsdbCfg), vBufPoolGetMAF(pVnode), pVnode->pMeta, pVnode->pTfs);
+  pVnode->pTsdb =
+      tsdbOpen(dir, pVnode->vgId, &(pVnode->config.tsdbCfg), vBufPoolGetMAF(pVnode), pVnode->pMeta, pVnode->pTfs);
   if (pVnode->pTsdb == NULL) {
     // TODO: handle error
     return -1;
@@ -131,7 +132,7 @@ static int vnodeOpenImpl(SVnode *pVnode) {
 
   // Open TQ
   sprintf(dir, "%s/tq", pVnode->path);
-  pVnode->pTq = tqOpen(dir, pVnode->pWal, pVnode->pMeta, &(pVnode->config.tqCfg), vBufPoolGetMAF(pVnode));
+  pVnode->pTq = tqOpen(dir, pVnode, pVnode->pWal, pVnode->pMeta, &(pVnode->config.tqCfg), vBufPoolGetMAF(pVnode));
   if (pVnode->pTq == NULL) {
     // TODO: handle error
     return -1;

source/dnode/vnode/src/vnd/vnodeQuery.c

@@ -68,6 +68,8 @@ int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg) {
       return tqProcessPollReq(pVnode->pTq, pMsg);
     case TDMT_VND_TASK_EXEC:
       return tqProcessTaskExec(pVnode->pTq, pMsg);
+    case TDMT_VND_STREAM_TRIGGER:
+      return tqProcessStreamTrigger(pVnode->pTq, pMsg->pCont, pMsg->contLen);
     case TDMT_VND_QUERY_HEARTBEAT:
       return qWorkerProcessHbMsg(pVnode, pVnode->pQuery, pMsg);
     default:
@@ -163,7 +165,6 @@ static int vnodeGetTableMeta(SVnode *pVnode, SRpcMsg *pMsg) {
     memcpy(POINTER_SHIFT(metaRsp.pSchemas, sizeof(SSchema) * pSW->nCols), pTagSchema, sizeof(SSchema) * nTagCols);
   }
 
-
 _exit:
 
   rspLen = tSerializeSTableMetaRsp(NULL, 0, &metaRsp);
@@ -179,7 +180,6 @@ _exit:
   }
   tSerializeSTableMetaRsp(pRsp, rspLen, &metaRsp);
 
-
   tFreeSTableMetaRsp(&metaRsp);
   if (pSW != NULL) {
     tfree(pSW->pSchema);

source/dnode/vnode/src/vnd/vnodeWrite.c

@@ -59,7 +59,7 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
   // todo: change the interface here
   int64_t ver;
   taosDecodeFixedI64(POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), &ver);
-  if (tqPushMsg(pVnode->pTq, ptr, pMsg->msgType, ver) < 0) {
+  if (tqPushMsg(pVnode->pTq, pMsg->pCont, pMsg->contLen, pMsg->msgType, ver) < 0) {
     // TODO: handle error
   }
 
@@ -85,10 +85,10 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       for (int i = 0; i < reqNum; i++) {
         SVCreateTbReq *pCreateTbReq = taosArrayGet(vCreateTbBatchReq.pArray, i);
 
-        char tableFName[TSDB_TABLE_FNAME_LEN];
+        char      tableFName[TSDB_TABLE_FNAME_LEN];
         SMsgHead *pHead = (SMsgHead *)pMsg->pCont;
         sprintf(tableFName, "%s.%s", pCreateTbReq->dbFName, pCreateTbReq->name);
-        
+
         int32_t code = vnodeValidateTableHash(&pVnode->config, tableFName);
         if (code) {
           SVCreateTbRsp rsp;
@@ -96,7 +96,7 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
 
           taosArrayPush(vCreateTbBatchRsp.rspList, &rsp);
         }
-        
+
         if (metaCreateTable(pVnode->pMeta, pCreateTbReq) < 0) {
           // TODO: handle error
           vError("vgId:%d, failed to create table: %s", pVnode->vgId, pCreateTbReq->name);
@@ -116,10 +116,10 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       taosArrayDestroy(vCreateTbBatchReq.pArray);
       if (vCreateTbBatchRsp.rspList) {
         int32_t contLen = tSerializeSVCreateTbBatchRsp(NULL, 0, &vCreateTbBatchRsp);
-        void *msg = rpcMallocCont(contLen);
+        void   *msg = rpcMallocCont(contLen);
         tSerializeSVCreateTbBatchRsp(msg, contLen, &vCreateTbBatchRsp);
         taosArrayDestroy(vCreateTbBatchRsp.rspList);
-        
+
         *pRsp = calloc(1, sizeof(SRpcMsg));
         (*pRsp)->msgType = TDMT_VND_CREATE_TABLE_RSP;
         (*pRsp)->pCont = msg;
@@ -168,6 +168,7 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       }
     } break;
     case TDMT_VND_CREATE_SMA: {  // timeRangeSMA
+#if 0
       SSmaCfg vCreateSmaReq = {0};
       if (tDeserializeSVCreateTSmaReq(POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), &vCreateSmaReq) == NULL) {
         terrno = TSDB_CODE_OUT_OF_MEMORY;
@@ -189,26 +190,37 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       // }
       tdDestroyTSma(&vCreateSmaReq.tSma);
       // TODO: return directly or go on follow steps?
+#endif
     } break;
     case TDMT_VND_CANCEL_SMA: {  // timeRangeSMA
     } break;
     case TDMT_VND_DROP_SMA: {  // timeRangeSMA
+#if 0    
       SVDropTSmaReq vDropSmaReq = {0};
       if (tDeserializeSVDropTSmaReq(POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), &vDropSmaReq) == NULL) {
         terrno = TSDB_CODE_OUT_OF_MEMORY;
         return -1;
       }
 
-      if (metaDropTSma(pVnode->pMeta, vDropSmaReq.indexName) < 0) {
-        // TODO: handle error
-        return -1;
-      }
       // TODO: send msg to stream computing to drop tSma
       // if ((send msg to stream computing) < 0) {
       //   tdDestroyTSma(&vCreateSmaReq);
       //   return -1;
       // }
+      // 
+
+      if (metaDropTSma(pVnode->pMeta, vDropSmaReq.indexUid) < 0) {
+        // TODO: handle error
+        return -1;
+      }
+
+      if(tsdbDropTSmaData(pVnode->pTsdb, vDropSmaReq.indexUid) < 0) {
+        // TODO: handle error
+        return -1;
+      }
+
       // TODO: return directly or go on follow steps?
+#endif
     } break;
     default:
       ASSERT(0);

source/dnode/vnode/test/tsdbSmaTest.cpp

@@ -272,8 +272,8 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   taosArrayDestroy(pUids);
 
   // resource release
-  metaRemoveSmaFromDb(pMeta, smaIndexName1);
-  metaRemoveSmaFromDb(pMeta, smaIndexName2);
+  metaRemoveSmaFromDb(pMeta, indexUid1);
+  metaRemoveSmaFromDb(pMeta, indexUid2);
 
   tdDestroyTSma(&tSma);
   metaClose(pMeta);

source/libs/executor/inc/executorimpl.h

@@ -76,11 +76,12 @@ typedef struct SResultRowCell {
  * If the number of generated results is greater than this value,
  * query query will be halt and return results to client immediate.
  */
-typedef struct SRspResultInfo {
-  int64_t total;      // total generated result size in rows
-  int32_t capacity;   // capacity of current result output buffer
-  int32_t threshold;  // result size threshold in rows.
-} SRspResultInfo;
+typedef struct SResultInfo { // TODO refactor
+  int64_t totalRows;      // total generated result size in rows
+  int64_t totalBytes;     // total results in bytes.
+  int32_t capacity;       // capacity of current result output buffer
+  int32_t threshold;      // result size threshold in rows.
+} SResultInfo;
 
 typedef struct SColumnFilterElem {
   int16_t           bytes;  // column length
@@ -160,8 +161,8 @@ typedef struct STaskCostInfo {
 typedef struct SOperatorCostInfo {
   uint64_t openCost;
   uint64_t execCost;
-  uint64_t totalRows;
-  uint64_t totalBytes;
+//  uint64_t totalRows;
+//  uint64_t totalBytes;
 } SOperatorCostInfo;
 
 typedef struct {
@@ -301,7 +302,7 @@ typedef struct STaskRuntimeEnv {
   int64_t         currentOffset;  // dynamic offset value
 
   STableQueryInfo* current;
-  SRspResultInfo   resultInfo;
+  SResultInfo   resultInfo;
   SHashObj*        pTableRetrieveTsMap;
   struct SUdfInfo* pUdfInfo;
 } STaskRuntimeEnv;
@@ -324,7 +325,7 @@ typedef struct SOperatorInfo {
   STaskRuntimeEnv*       pRuntimeEnv;   // todo remove it
   SExecTaskInfo*         pTaskInfo;
   SOperatorCostInfo      cost;
-
+  SResultInfo            resultInfo;
   struct SOperatorInfo** pDownstream;      // downstram pointer list
   int32_t                numOfDownstream;  // number of downstream. The value is always ONE expect for join operator
   __optr_fn_t            getNextFn;
@@ -539,12 +540,24 @@ typedef struct SFillOperatorInfo {
   void**            p;
   SSDataBlock*      existNewGroupBlock;
   bool              multigroupResult;
+  SInterval         intervalInfo;
+  int32_t           capacity;
 } SFillOperatorInfo;
 
+typedef struct SGroupKeys {
+  char   *pData;
+  bool    isNull;
+  int16_t type;
+  int32_t bytes;
+}SGroupKeys;
+
 typedef struct SGroupbyOperatorInfo {
   SOptrBasicInfo binfo;
   SArray*        pGroupCols;
-  char*          prevData;  // previous group by value
+  SArray*        pGroupColVals; // current group column values, SArray<SGroupKeys>
+  bool           isInit;       // denote if current val is initialized or not
+  char*          keyBuf;       // group by keys for hash
+  int32_t        groupKeyLen;  // total group by column width
   SGroupResInfo  groupResInfo;
   SAggSupporter  aggSup;
 } SGroupbyOperatorInfo;
@@ -639,21 +652,19 @@ SOperatorInfo* createOrderOperatorInfo(SOperatorInfo* downstream, SExprInfo* pEx
 SOperatorInfo* createSortedMergeOperatorInfo(SOperatorInfo** downstream, int32_t numOfDownstream, SExprInfo* pExprInfo, int32_t num, SArray* pOrderVal, SArray* pGroupInfo, SExecTaskInfo* pTaskInfo);
 SOperatorInfo* createSysTableScanOperatorInfo(void* pSysTableReadHandle, SSDataBlock* pResBlock, const SName* pName,
                                               SNode* pCondition, SEpSet epset, SArray* colList, SExecTaskInfo* pTaskInfo);
-SOperatorInfo* createLimitOperatorInfo(SOperatorInfo* downstream, int32_t numOfDownstream, SLimit* pLimit, SExecTaskInfo* pTaskInfo);
+SOperatorInfo* createLimitOperatorInfo(SOperatorInfo* downstream, SLimit* pLimit, SExecTaskInfo* pTaskInfo);
 
 SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResBlock, SInterval* pInterval,
                                           const STableGroupInfo* pTableGroupInfo, SExecTaskInfo* pTaskInfo);
 SOperatorInfo* createSessionAggOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResBlock, SExecTaskInfo* pTaskInfo);
 SOperatorInfo* createGroupOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResultBlock,
                                        SArray* pGroupColList, SExecTaskInfo* pTaskInfo, const STableGroupInfo* pTableGroupInfo);
+SOperatorInfo* createFillOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExpr, int32_t numOfCols, SInterval* pInterval, SSDataBlock* pResBlock, bool multigroupResult, SExecTaskInfo* pTaskInfo);
 
 SOperatorInfo* createTableSeqScanOperatorInfo(void* pTsdbReadHandle, STaskRuntimeEnv* pRuntimeEnv);
 SOperatorInfo* createAllTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,
                                                  SExprInfo* pExpr, int32_t numOfOutput);
 
-SOperatorInfo* createFillOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream, SExprInfo* pExpr,
-                                      int32_t numOfOutput, bool multigroupResult);
-
 SOperatorInfo* createMultiTableTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,
                                                         SExprInfo* pExpr, int32_t numOfOutput);
 SOperatorInfo* createAllMultiTableTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,

source/libs/executor/src/executor.c

@@ -30,7 +30,7 @@ static int32_t doSetStreamBlock(SOperatorInfo* pOperator, void* input, int32_t t
       qError("join not supported for stream block scan, %s" PRIx64, id);
       return TSDB_CODE_QRY_APP_ERROR;
     }
-
+    pOperator->status = OP_NOT_OPENED;
     return doSetStreamBlock(pOperator->pDownstream[0], input, type, id);
   } else {
     SStreamBlockScanInfo* pInfo = pOperator->info;

source/libs/sync/inc/syncRaftLog.h

@@ -47,6 +47,8 @@ SyncIndex       logStoreGetCommitIndex(SSyncLogStore* pLogStore);
 SSyncRaftEntry* logStoreGetLastEntry(SSyncLogStore* pLogStore);
 cJSON*          logStore2Json(SSyncLogStore* pLogStore);
 char*           logStore2Str(SSyncLogStore* pLogStore);
+cJSON*          logStoreSimple2Json(SSyncLogStore* pLogStore);
+char*           logStoreSimple2Str(SSyncLogStore* pLogStore);
 
 // for debug
 void logStorePrint(SSyncLogStore* pLogStore);
@@ -54,6 +56,11 @@ void logStorePrint2(char* s, SSyncLogStore* pLogStore);
 void logStoreLog(SSyncLogStore* pLogStore);
 void logStoreLog2(char* s, SSyncLogStore* pLogStore);
 
+void logStoreSimplePrint(SSyncLogStore* pLogStore);
+void logStoreSimplePrint2(char* s, SSyncLogStore* pLogStore);
+void logStoreSimpleLog(SSyncLogStore* pLogStore);
+void logStoreSimpleLog2(char* s, SSyncLogStore* pLogStore);
+
 #ifdef __cplusplus
 }
 #endif

source/libs/sync/src/syncAppendEntries.c

@@ -120,6 +120,11 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
   // reject request
   if ((pMsg->term < ths->pRaftStore->currentTerm) ||
       ((pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) && !logOK)) {
+    sTrace(
+        "syncNodeOnAppendEntriesCb --> reject, pMsg->term:%lu, ths->pRaftStore->currentTerm:%lu, ths->state:%d, "
+        "logOK:%d",
+        pMsg->term, ths->pRaftStore->currentTerm, ths->state, logOK);
+
     SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild();
     pReply->srcId = ths->myRaftId;
     pReply->destId = pMsg->srcId;
@@ -137,6 +142,11 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
 
   // return to follower state
   if (pMsg->term == ths->pRaftStore->currentTerm && ths->state == TAOS_SYNC_STATE_CANDIDATE) {
+    sTrace(
+        "syncNodeOnAppendEntriesCb --> return to follower, pMsg->term:%lu, ths->pRaftStore->currentTerm:%lu, "
+        "ths->state:%d, logOK:%d",
+        pMsg->term, ths->pRaftStore->currentTerm, ths->state, logOK);
+
     syncNodeBecomeFollower(ths);
 
     // ret or reply?
@@ -145,89 +155,60 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
 
   // accept request
   if (pMsg->term == ths->pRaftStore->currentTerm && ths->state == TAOS_SYNC_STATE_FOLLOWER && logOK) {
-    bool preMatch = false;
-    if (pMsg->prevLogIndex == SYNC_INDEX_INVALID &&
-        ths->pLogStore->getLastIndex(ths->pLogStore) == SYNC_INDEX_INVALID) {
-      preMatch = true;
-    }
-    if (pMsg->prevLogIndex >= SYNC_INDEX_BEGIN && pMsg->prevLogIndex <= ths->pLogStore->getLastIndex(ths->pLogStore)) {
-      SSyncRaftEntry* pPreEntry = logStoreGetEntry(ths->pLogStore, pMsg->prevLogIndex);
-      assert(pPreEntry != NULL);
-      if (pMsg->prevLogTerm == pPreEntry->term) {
-        preMatch = true;
-      }
-      syncEntryDestory(pPreEntry);
-    }
+    // preIndex = -1, or has preIndex entry in local log
+    assert(pMsg->prevLogIndex <= ths->pLogStore->getLastIndex(ths->pLogStore));
 
-    if (preMatch) {
-      // must has preIndex in local log
-      assert(pMsg->prevLogIndex <= ths->pLogStore->getLastIndex(ths->pLogStore));
+    // has extra entries (> preIndex) in local log
+    bool hasExtraEntries = pMsg->prevLogIndex < ths->pLogStore->getLastIndex(ths->pLogStore);
 
-      bool hasExtraEntries = pMsg->prevLogIndex < ths->pLogStore->getLastIndex(ths->pLogStore);
-      bool hasAppendEntries = pMsg->dataLen > 0;
+    // has entries in SyncAppendEntries msg
+    bool hasAppendEntries = pMsg->dataLen > 0;
 
-      if (hasExtraEntries && hasAppendEntries) {
-        // conflict
-        bool conflict = false;
+    sTrace(
+        "syncNodeOnAppendEntriesCb --> accept, pMsg->term:%lu, ths->pRaftStore->currentTerm:%lu, ths->state:%d, "
+        "logOK:%d, hasExtraEntries:%d, hasAppendEntries:%d",
+        pMsg->term, ths->pRaftStore->currentTerm, ths->state, logOK, hasExtraEntries, hasAppendEntries);
 
-        SyncIndex       extraIndex = pMsg->prevLogIndex + 1;
-        SSyncRaftEntry* pExtraEntry = logStoreGetEntry(ths->pLogStore, extraIndex);
-        assert(pExtraEntry != NULL);
+    if (hasExtraEntries && hasAppendEntries) {
+      // not conflict by default
+      bool conflict = false;
 
-        SSyncRaftEntry* pAppendEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
-        assert(pAppendEntry != NULL);
+      SyncIndex       extraIndex = pMsg->prevLogIndex + 1;
+      SSyncRaftEntry* pExtraEntry = logStoreGetEntry(ths->pLogStore, extraIndex);
+      assert(pExtraEntry != NULL);
 
-        assert(extraIndex == pAppendEntry->index);
-        if (pExtraEntry->term == pAppendEntry->term) {
-          conflict = true;
-        }
+      SSyncRaftEntry* pAppendEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
+      assert(pAppendEntry != NULL);
 
-        if (conflict) {
-          // roll back
-          SyncIndex delBegin = ths->pLogStore->getLastIndex(ths->pLogStore);
-          SyncIndex delEnd = extraIndex;
+      // log not match, conflict
+      assert(extraIndex == pAppendEntry->index);
+      if (pExtraEntry->term != pAppendEntry->term) {
+        conflict = true;
+      }
 
-          // notice! reverse roll back!
-          for (SyncIndex index = delEnd; index >= delBegin; --index) {
-            if (ths->pFsm->FpRollBackCb != NULL) {
-              SSyncRaftEntry* pRollBackEntry = logStoreGetEntry(ths->pLogStore, index);
-              assert(pRollBackEntry != NULL);
+      if (conflict) {
+        // roll back
+        SyncIndex delBegin = ths->pLogStore->getLastIndex(ths->pLogStore);
+        SyncIndex delEnd = extraIndex;
 
-              SRpcMsg rpcMsg;
-              syncEntry2OriginalRpc(pRollBackEntry, &rpcMsg);
-              ths->pFsm->FpRollBackCb(ths->pFsm, &rpcMsg, pRollBackEntry->index, pRollBackEntry->isWeak, 0);
-              rpcFreeCont(rpcMsg.pCont);
-              syncEntryDestory(pRollBackEntry);
-            }
-          }
+        sTrace("syncNodeOnAppendEntriesCb --> conflict:%d, delBegin:%ld, delEnd:%ld", conflict, delBegin, delEnd);
 
-          // delete confict entries
-          ths->pLogStore->truncate(ths->pLogStore, extraIndex);
+        // notice! reverse roll back!
+        for (SyncIndex index = delEnd; index >= delBegin; --index) {
+          if (ths->pFsm->FpRollBackCb != NULL) {
+            SSyncRaftEntry* pRollBackEntry = logStoreGetEntry(ths->pLogStore, index);
+            assert(pRollBackEntry != NULL);
 
-          // append new entries
-          ths->pLogStore->appendEntry(ths->pLogStore, pAppendEntry);
-
-          // pre commit
-          SRpcMsg rpcMsg;
-          syncEntry2OriginalRpc(pAppendEntry, &rpcMsg);
-          if (ths->pFsm != NULL) {
-            if (ths->pFsm->FpPreCommitCb != NULL) {
-              ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pAppendEntry->index, pAppendEntry->isWeak, 0);
-            }
+            SRpcMsg rpcMsg;
+            syncEntry2OriginalRpc(pRollBackEntry, &rpcMsg);
+            ths->pFsm->FpRollBackCb(ths->pFsm, &rpcMsg, pRollBackEntry->index, pRollBackEntry->isWeak, 0, ths->state);
+            rpcFreeCont(rpcMsg.pCont);
+            syncEntryDestory(pRollBackEntry);
           }
-          rpcFreeCont(rpcMsg.pCont);
         }
 
-        // free memory
-        syncEntryDestory(pExtraEntry);
-        syncEntryDestory(pAppendEntry);
-
-      } else if (hasExtraEntries && !hasAppendEntries) {
-        // do nothing
-
-      } else if (!hasExtraEntries && hasAppendEntries) {
-        SSyncRaftEntry* pAppendEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
-        assert(pAppendEntry != NULL);
+        // delete confict entries
+        ths->pLogStore->truncate(ths->pLogStore, extraIndex);
 
         // append new entries
         ths->pLogStore->appendEntry(ths->pLogStore, pAppendEntry);
@@ -237,53 +218,63 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
         syncEntry2OriginalRpc(pAppendEntry, &rpcMsg);
         if (ths->pFsm != NULL) {
           if (ths->pFsm->FpPreCommitCb != NULL) {
-            ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pAppendEntry->index, pAppendEntry->isWeak, 0);
+            ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pAppendEntry->index, pAppendEntry->isWeak, 2, ths->state);
           }
         }
         rpcFreeCont(rpcMsg.pCont);
+      }
 
-        // free memory
-        syncEntryDestory(pAppendEntry);
+      // free memory
+      syncEntryDestory(pExtraEntry);
+      syncEntryDestory(pAppendEntry);
 
-      } else if (!hasExtraEntries && !hasAppendEntries) {
-        // do nothing
+    } else if (hasExtraEntries && !hasAppendEntries) {
+      // do nothing
 
-      } else {
-        assert(0);
-      }
+    } else if (!hasExtraEntries && hasAppendEntries) {
+      SSyncRaftEntry* pAppendEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
+      assert(pAppendEntry != NULL);
 
-      SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild();
-      pReply->srcId = ths->myRaftId;
-      pReply->destId = pMsg->srcId;
-      pReply->term = ths->pRaftStore->currentTerm;
-      pReply->success = true;
+      // append new entries
+      ths->pLogStore->appendEntry(ths->pLogStore, pAppendEntry);
 
-      if (hasAppendEntries) {
-        pReply->matchIndex = pMsg->prevLogIndex + 1;
-      } else {
-        pReply->matchIndex = pMsg->prevLogIndex;
+      // pre commit
+      SRpcMsg rpcMsg;
+      syncEntry2OriginalRpc(pAppendEntry, &rpcMsg);
+      if (ths->pFsm != NULL) {
+        if (ths->pFsm->FpPreCommitCb != NULL) {
+          ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pAppendEntry->index, pAppendEntry->isWeak, 3, ths->state);
+        }
       }
+      rpcFreeCont(rpcMsg.pCont);
 
-      SRpcMsg rpcMsg;
-      syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
-      syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
+      // free memory
+      syncEntryDestory(pAppendEntry);
 
-      syncAppendEntriesReplyDestroy(pReply);
+    } else if (!hasExtraEntries && !hasAppendEntries) {
+      // do nothing
 
     } else {
-      SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild();
-      pReply->srcId = ths->myRaftId;
-      pReply->destId = pMsg->srcId;
-      pReply->term = ths->pRaftStore->currentTerm;
-      pReply->success = false;
-      pReply->matchIndex = SYNC_INDEX_INVALID;
+      assert(0);
+    }
 
-      SRpcMsg rpcMsg;
-      syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
-      syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
-      syncAppendEntriesReplyDestroy(pReply);
+    SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild();
+    pReply->srcId = ths->myRaftId;
+    pReply->destId = pMsg->srcId;
+    pReply->term = ths->pRaftStore->currentTerm;
+    pReply->success = true;
+
+    if (hasAppendEntries) {
+      pReply->matchIndex = pMsg->prevLogIndex + 1;
+    } else {
+      pReply->matchIndex = pMsg->prevLogIndex;
     }
 
+    SRpcMsg rpcMsg;
+    syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
+    syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
+    syncAppendEntriesReplyDestroy(pReply);
+
     // maybe update commit index from leader
     if (pMsg->commitIndex > ths->commitIndex) {
       // has commit entry in local
@@ -308,7 +299,7 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
               syncEntry2OriginalRpc(pEntry, &rpcMsg);
 
               if (ths->pFsm->FpCommitCb != NULL) {
-                ths->pFsm->FpCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0);
+                ths->pFsm->FpCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0, ths->state);
               }
 
               rpcFreeCont(rpcMsg.pCont);

source/libs/sync/src/syncAppendEntriesReply.c

@@ -48,6 +48,9 @@ int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* p
     return ret;
   }
 
+  syncIndexMgrLog2("==syncNodeOnAppendEntriesReplyCb== before pNextIndex", ths->pNextIndex);
+  syncIndexMgrLog2("==syncNodeOnAppendEntriesReplyCb== before pMatchIndex", ths->pMatchIndex);
+
   // no need this code, because if I receive reply.term, then I must have sent for that term.
   //  if (pMsg->term > ths->pRaftStore->currentTerm) {
   //    syncNodeUpdateTerm(ths, pMsg->term);
@@ -77,5 +80,8 @@ int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* p
     syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), nextIndex);
   }
 
+  syncIndexMgrLog2("==syncNodeOnAppendEntriesReplyCb== after pNextIndex", ths->pNextIndex);
+  syncIndexMgrLog2("==syncNodeOnAppendEntriesReplyCb== after pMatchIndex", ths->pMatchIndex);
+
   return ret;
 }

source/libs/sync/src/syncCommit.c

@@ -63,7 +63,14 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
       if (pEntry->term == pSyncNode->pRaftStore->currentTerm) {
         // update commit index
         newCommitIndex = index;
+        sTrace("syncMaybeAdvanceCommitIndex maybe to update, newCommitIndex:%ld commit, pSyncNode->commitIndex:%ld",
+               newCommitIndex, pSyncNode->commitIndex);
         break;
+      } else {
+        sTrace(
+            "syncMaybeAdvanceCommitIndex can not commit due to term not equal, pEntry->term:%lu, "
+            "pSyncNode->pRaftStore->currentTerm:%lu",
+            pEntry->term, pSyncNode->pRaftStore->currentTerm);
       }
     }
   }
@@ -91,7 +98,7 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
           syncEntry2OriginalRpc(pEntry, &rpcMsg);
 
           if (pSyncNode->pFsm->FpCommitCb != NULL) {
-            pSyncNode->pFsm->FpCommitCb(pSyncNode->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0);
+            pSyncNode->pFsm->FpCommitCb(pSyncNode->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0, pSyncNode->state);
           }
 
           rpcFreeCont(rpcMsg.pCont);

source/libs/sync/src/syncIO.c

@@ -257,13 +257,6 @@ static void *syncIOConsumerFunc(void *param) {
           assert(pSyncMsg != NULL);
           io->FpOnSyncPing(io->pSyncNode, pSyncMsg);
           syncPingDestroy(pSyncMsg);
-          /*
-          pSyncMsg = syncPingBuild(pRpcMsg->contLen);
-          syncPingFromRpcMsg(pRpcMsg, pSyncMsg);
-          // memcpy(pSyncMsg, tmpRpcMsg.pCont, tmpRpcMsg.contLen);
-          io->FpOnSyncPing(io->pSyncNode, pSyncMsg);
-          syncPingDestroy(pSyncMsg);
-          */
         }
 
       } else if (pRpcMsg->msgType == SYNC_PING_REPLY) {

source/libs/sync/src/syncMain.c

@@ -104,9 +104,26 @@ int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) {
 }
 
 int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
-  int32_t ret = 0;
+  int32_t ret = syncPropose2(rid, pMsg, isWeak, 0);
+  return ret;
+}
+
+int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
+  int32_t ret = syncPropose(rid, pMsg, isWeak);
+  return ret;
+}
+
+ESyncState syncGetMyRole(int64_t rid) {
+  SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
+  if (pSyncNode == NULL) {
+    return TAOS_SYNC_STATE_ERROR;
+  }
+  assert(rid == pSyncNode->rid);
+  return pSyncNode->state;
+}
 
-  // todo : get pointer from rid
+int32_t syncPropose2(int64_t rid, const SRpcMsg* pMsg, bool isWeak, uint64_t seqNum) {
+  int32_t    ret = 0;
   SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
   if (pSyncNode == NULL) {
     return -1;
@@ -114,7 +131,7 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
   assert(rid == pSyncNode->rid);
 
   if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
-    SyncClientRequest* pSyncMsg = syncClientRequestBuild2(pMsg, 0, isWeak);
+    SyncClientRequest* pSyncMsg = syncClientRequestBuild2(pMsg, seqNum, isWeak);
     SRpcMsg            rpcMsg;
     syncClientRequest2RpcMsg(pSyncMsg, &rpcMsg);
     pSyncNode->FpEqMsg(pSyncNode->queue, &rpcMsg);
@@ -122,7 +139,7 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
     ret = 0;
 
   } else {
-    sTrace("syncForwardToPeer not leader, %s", syncUtilState2String(pSyncNode->state));
+    sTrace("syncPropose not leader, %s", syncUtilState2String(pSyncNode->state));
     ret = -1;  // todo : need define err code !!
   }
 
@@ -130,19 +147,6 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
   return ret;
 }
 
-int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
-  int32_t ret = syncPropose(rid, pMsg, isWeak);
-  return ret;
-}
-
-ESyncState syncGetMyRole(int64_t rid) {
-  // todo : get pointer from rid
-  SSyncNode* pSyncNode = NULL;
-  return pSyncNode->state;
-}
-
-void syncGetNodesRole(int64_t rid, SNodesRole* pNodeRole) {}
-
 // open/close --------------
 SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
   SSyncNode* pSyncNode = (SSyncNode*)malloc(sizeof(SSyncNode));
@@ -848,7 +852,7 @@ static int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg
 
     if (ths->pFsm != NULL) {
       if (ths->pFsm->FpPreCommitCb != NULL) {
-        ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0);
+        ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0, ths->state);
       }
     }
     rpcFreeCont(rpcMsg.pCont);
@@ -863,7 +867,7 @@ static int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg
 
     if (ths->pFsm != NULL) {
       if (ths->pFsm->FpPreCommitCb != NULL) {
-        ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, -2);
+        ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 1, ths->state);
       }
     }
     rpcFreeCont(rpcMsg.pCont);

source/libs/sync/src/syncMessage.c

@@ -834,7 +834,7 @@ cJSON* syncRequestVote2Json(const SyncRequestVote* pMsg) {
 
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
     cJSON_AddStringToObject(pRoot, "term", u64buf);
-    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastLogIndex);
+    snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->lastLogIndex);
     cJSON_AddStringToObject(pRoot, "lastLogIndex", u64buf);
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastLogTerm);
     cJSON_AddStringToObject(pRoot, "lastLogTerm", u64buf);
@@ -1127,14 +1127,14 @@ cJSON* syncAppendEntries2Json(const SyncAppendEntries* pMsg) {
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
     cJSON_AddStringToObject(pRoot, "term", u64buf);
 
-    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->prevLogIndex);
-    cJSON_AddStringToObject(pRoot, "pre_log_index", u64buf);
+    snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->prevLogIndex);
+    cJSON_AddStringToObject(pRoot, "prevLogIndex", u64buf);
 
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->prevLogTerm);
     cJSON_AddStringToObject(pRoot, "pre_log_term", u64buf);
 
-    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->commitIndex);
-    cJSON_AddStringToObject(pRoot, "commit_index", u64buf);
+    snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->commitIndex);
+    cJSON_AddStringToObject(pRoot, "commitIndex", u64buf);
 
     cJSON_AddNumberToObject(pRoot, "dataLen", pMsg->dataLen);
     char* s;
@@ -1288,7 +1288,7 @@ cJSON* syncAppendEntriesReply2Json(const SyncAppendEntriesReply* pMsg) {
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
     cJSON_AddStringToObject(pRoot, "term", u64buf);
     cJSON_AddNumberToObject(pRoot, "success", pMsg->success);
-    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->matchIndex);
+    snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->matchIndex);
     cJSON_AddStringToObject(pRoot, "matchIndex", u64buf);
   }
 

source/libs/sync/src/syncRaftLog.c

@@ -165,6 +165,34 @@ char* logStore2Str(SSyncLogStore* pLogStore) {
   return serialized;
 }
 
+cJSON* logStoreSimple2Json(SSyncLogStore* pLogStore) {
+  char               u64buf[128];
+  SSyncLogStoreData* pData = (SSyncLogStoreData*)pLogStore->data;
+  cJSON*             pRoot = cJSON_CreateObject();
+
+  if (pData != NULL && pData->pWal != NULL) {
+    snprintf(u64buf, sizeof(u64buf), "%p", pData->pSyncNode);
+    cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
+    snprintf(u64buf, sizeof(u64buf), "%p", pData->pWal);
+    cJSON_AddStringToObject(pRoot, "pWal", u64buf);
+    snprintf(u64buf, sizeof(u64buf), "%ld", logStoreLastIndex(pLogStore));
+    cJSON_AddStringToObject(pRoot, "LastIndex", u64buf);
+    snprintf(u64buf, sizeof(u64buf), "%lu", logStoreLastTerm(pLogStore));
+    cJSON_AddStringToObject(pRoot, "LastTerm", u64buf);
+  }
+
+  cJSON* pJson = cJSON_CreateObject();
+  cJSON_AddItemToObject(pJson, "SSyncLogStoreSimple", pRoot);
+  return pJson;
+}
+
+char* logStoreSimple2Str(SSyncLogStore* pLogStore) {
+  cJSON* pJson = logStoreSimple2Json(pLogStore);
+  char*  serialized = cJSON_Print(pJson);
+  cJSON_Delete(pJson);
+  return serialized;
+}
+
 // for debug -----------------
 void logStorePrint(SSyncLogStore* pLogStore) {
   char* serialized = logStore2Str(pLogStore);
@@ -191,3 +219,30 @@ void logStoreLog2(char* s, SSyncLogStore* pLogStore) {
   sTrace("logStorePrint | len:%lu | %s | %s", strlen(serialized), s, serialized);
   free(serialized);
 }
+
+// for debug -----------------
+void logStoreSimplePrint(SSyncLogStore* pLogStore) {
+  char* serialized = logStoreSimple2Str(pLogStore);
+  printf("logStoreSimplePrint | len:%lu | %s \n", strlen(serialized), serialized);
+  fflush(NULL);
+  free(serialized);
+}
+
+void logStoreSimplePrint2(char* s, SSyncLogStore* pLogStore) {
+  char* serialized = logStoreSimple2Str(pLogStore);
+  printf("logStoreSimplePrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
+  fflush(NULL);
+  free(serialized);
+}
+
+void logStoreSimpleLog(SSyncLogStore* pLogStore) {
+  char* serialized = logStoreSimple2Str(pLogStore);
+  sTrace("logStoreSimpleLog | len:%lu | %s", strlen(serialized), serialized);
+  free(serialized);
+}
+
+void logStoreSimpleLog2(char* s, SSyncLogStore* pLogStore) {
+  char* serialized = logStoreSimple2Str(pLogStore);
+  sTrace("logStoreSimpleLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+  free(serialized);
+}
\ No newline at end of file

source/libs/sync/src/syncReplication.c

@@ -49,6 +49,10 @@
 int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode) {
   assert(pSyncNode->state == TAOS_SYNC_STATE_LEADER);
 
+  syncIndexMgrLog2("==syncNodeAppendEntriesPeers== pNextIndex", pSyncNode->pNextIndex);
+  syncIndexMgrLog2("==syncNodeAppendEntriesPeers== pMatchIndex", pSyncNode->pMatchIndex);
+  logStoreSimpleLog2("==syncNodeAppendEntriesPeers==", pSyncNode->pLogStore);
+
   int32_t ret = 0;
   for (int i = 0; i < pSyncNode->peersNum; ++i) {
     SRaftId* pDestId = &(pSyncNode->peersId[i]);
@@ -99,6 +103,8 @@ int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode) {
     pMsg->prevLogTerm = preLogTerm;
     pMsg->commitIndex = pSyncNode->commitIndex;
 
+    syncAppendEntriesLog2("==syncNodeAppendEntriesPeers==", pMsg);
+
     // send AppendEntries
     syncNodeAppendEntries(pSyncNode, pDestId, pMsg);
     syncAppendEntriesDestroy(pMsg);

source/libs/sync/test/CMakeLists.txt

@@ -29,10 +29,14 @@ add_executable(syncPingTimerTest2 "")
 add_executable(syncPingSelfTest "")
 add_executable(syncElectTest "")
 add_executable(syncElectTest2 "")
+add_executable(syncElectTest3 "")
 add_executable(syncEncodeTest "")
 add_executable(syncWriteTest "")
 add_executable(syncReplicateTest "")
+add_executable(syncReplicateTest2 "")
+add_executable(syncReplicateLoadTest "")
 add_executable(syncRefTest "")
+add_executable(syncLogStoreCheck "")
 
 
 target_sources(syncTest
@@ -159,6 +163,10 @@ target_sources(syncElectTest2
     PRIVATE
     "syncElectTest2.cpp"
 )
+target_sources(syncElectTest3
+    PRIVATE
+    "syncElectTest3.cpp"
+)
 target_sources(syncEncodeTest
     PRIVATE
     "syncEncodeTest.cpp"
@@ -171,10 +179,22 @@ target_sources(syncReplicateTest
     PRIVATE
     "syncReplicateTest.cpp"
 )
+target_sources(syncReplicateTest2
+    PRIVATE
+    "syncReplicateTest2.cpp"
+)
+target_sources(syncReplicateLoadTest
+    PRIVATE
+    "syncReplicateLoadTest.cpp"
+)
 target_sources(syncRefTest
     PRIVATE
     "syncRefTest.cpp"
 )
+target_sources(syncLogStoreCheck
+    PRIVATE
+    "syncLogStoreCheck.cpp"
+)
 
 
 target_include_directories(syncTest
@@ -332,6 +352,11 @@ target_include_directories(syncElectTest2
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
 )
+target_include_directories(syncElectTest3
+    PUBLIC
+    "${CMAKE_SOURCE_DIR}/include/libs/sync"
+    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+)
 target_include_directories(syncEncodeTest
     PUBLIC
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
@@ -347,11 +372,26 @@ target_include_directories(syncReplicateTest
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
 )
+target_include_directories(syncReplicateTest2
+    PUBLIC
+    "${CMAKE_SOURCE_DIR}/include/libs/sync"
+    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+)
+target_include_directories(syncReplicateLoadTest
+    PUBLIC
+    "${CMAKE_SOURCE_DIR}/include/libs/sync"
+    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+)
 target_include_directories(syncRefTest
     PUBLIC
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
 )
+target_include_directories(syncLogStoreCheck
+    PUBLIC
+    "${CMAKE_SOURCE_DIR}/include/libs/sync"
+    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+)
 
 
 target_link_libraries(syncTest
@@ -478,6 +518,10 @@ target_link_libraries(syncElectTest2
     sync
     gtest_main
 )
+target_link_libraries(syncElectTest3
+    sync
+    gtest_main
+)
 target_link_libraries(syncEncodeTest
     sync
     gtest_main
@@ -490,10 +534,22 @@ target_link_libraries(syncReplicateTest
     sync
     gtest_main
 )
+target_link_libraries(syncReplicateTest2
+    sync
+    gtest_main
+)
+target_link_libraries(syncReplicateLoadTest
+    sync
+    gtest_main
+)
 target_link_libraries(syncRefTest
     sync
     gtest_main
 )
+target_link_libraries(syncLogStoreCheck
+    sync
+    gtest_main
+)
 
 
 enable_testing()

tests/script/tsim/query/interval-offset.sim

@@ -9,7 +9,7 @@ sql drop database d0 -x step1
 step1:
 sql create database d0
 sql show databases
-if $rows != 1 then 
+if $rows != 2 then
   return -1
 endi