Merge branch 'main' into enh/stream_buffer_param

Jenkinsfile2

@@ -430,7 +430,7 @@ pipeline {
                                         date
                                         rm -rf ${WKC}/debug
                                         cd ${WKC}/tests/parallel_test
-                                        time ./container_build.sh -w ${WKDIR} -t 10 -e
+                                        time ./container_build.sh -w ${WKDIR} -e
                                     '''
                                     def extra_param = ""
                                     def log_server_file = "/home/log_server.json"

cmake/cmake.version

@@ -2,7 +2,7 @@
 IF (DEFINED VERNUMBER)
   SET(TD_VER_NUMBER ${VERNUMBER})
 ELSE ()
-  SET(TD_VER_NUMBER "3.0.2.2")
+  SET(TD_VER_NUMBER "3.0.2.4")
 ENDIF ()
 
 IF (DEFINED VERCOMPATIBLE)

cmake/taosadapter_CMakeLists.txt.in

@@ -2,7 +2,7 @@
 # taosadapter
 ExternalProject_Add(taosadapter
         GIT_REPOSITORY https://github.com/taosdata/taosadapter.git
-        GIT_TAG a2e9920
+        GIT_TAG 213f8b3
         SOURCE_DIR "${TD_SOURCE_DIR}/tools/taosadapter"
         BINARY_DIR ""
         #BUILD_IN_SOURCE TRUE

cmake/taostools_CMakeLists.txt.in

@@ -2,7 +2,7 @@
 # taos-tools
 ExternalProject_Add(taos-tools
         GIT_REPOSITORY https://github.com/taosdata/taos-tools.git
-        GIT_TAG 4efbc10
+        GIT_TAG 7d24ed5
         SOURCE_DIR "${TD_SOURCE_DIR}/tools/taos-tools"
         BINARY_DIR ""
         #BUILD_IN_SOURCE TRUE

include/common/tdatablock.h

@@ -41,6 +41,12 @@ typedef struct SBlockOrderInfo {
     BMCharPos(bm_, r_) |= (1u << (7u - BitPos(r_))); \
   } while (0)
 
+#define colDataSetNull_f_s(c_, r_)                                        \
+  do {                                                                    \
+    colDataSetNull_f((c_)->nullbitmap, r_);                               \
+    memset(((char*)(c_)->pData) + (c_)->info.bytes * (r_), 0, (c_)->info.bytes); \
+  } while (0)
+
 #define colDataClearNull_f(bm_, r_)                             \
   do {                                                          \
     BMCharPos(bm_, r_) &= ((char)(~(1u << (7u - BitPos(r_))))); \
@@ -136,7 +142,7 @@ static FORCE_INLINE void colDataAppendNULL(SColumnInfoData* pColumnInfoData, uin
   if (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) {
     colDataSetNull_var(pColumnInfoData, currentRow);  // it is a null value of VAR type.
   } else {
-    colDataSetNull_f(pColumnInfoData->nullbitmap, currentRow);
+    colDataSetNull_f_s(pColumnInfoData, currentRow);
   }
 
   pColumnInfoData->hasNull = true;
@@ -151,6 +157,7 @@ static FORCE_INLINE void colDataAppendNNULL(SColumnInfoData* pColumnInfoData, ui
     for (int32_t i = start; i < start + nRows; ++i) {
       colDataSetNull_f(pColumnInfoData->nullbitmap, i);
     }
+    memset(pColumnInfoData->pData + start * pColumnInfoData->info.bytes, 0, pColumnInfoData->info.bytes * nRows);
   }
 
   pColumnInfoData->hasNull = true;
@@ -231,7 +238,6 @@ int32_t blockDataSort_rv(SSDataBlock* pDataBlock, SArray* pOrderInfo, bool nullF
 
 int32_t colInfoDataEnsureCapacity(SColumnInfoData* pColumn, uint32_t numOfRows, bool clearPayload);
 int32_t blockDataEnsureCapacity(SSDataBlock* pDataBlock, uint32_t numOfRows);
-int32_t blockDataEnsureCapacityNoClear(SSDataBlock* pDataBlock, uint32_t numOfRows);
 
 void colInfoDataCleanup(SColumnInfoData* pColumn, uint32_t numOfRows);
 void blockDataCleanup(SSDataBlock* pDataBlock);

include/common/tglobal.h

@@ -69,6 +69,9 @@ extern int32_t tsElectInterval;
 extern int32_t tsHeartbeatInterval;
 extern int32_t tsHeartbeatTimeout;
 
+// vnode
+extern int64_t tsVndCommitMaxIntervalMs;
+
 // monitor
 extern bool     tsEnableMonitor;
 extern int32_t  tsMonitorInterval;
@@ -82,6 +85,10 @@ extern bool     tsEnableTelem;
 extern int32_t  tsTelemInterval;
 extern char     tsTelemServer[];
 extern uint16_t tsTelemPort;
+extern bool     tsEnableCrashReport;
+extern char*    tsTelemUri;
+extern char*    tsClientCrashReportUri;
+extern char*    tsSvrCrashReportUri;
 
 // query buffer management
 extern int32_t tsQueryBufferSize;  // maximum allowed usage buffer size in MB for each data node during query processing

include/common/tmsg.h

@@ -907,6 +907,7 @@ typedef struct {
   int32_t numOfRetensions;
   SArray* pRetensions;
   int8_t  schemaless;
+  int16_t sstTrigger;
 } SDbCfgRsp;
 
 int32_t tSerializeSDbCfgRsp(void* buf, int32_t bufLen, const SDbCfgRsp* pRsp);

include/common/tmsgcb.h

@@ -39,7 +39,7 @@ typedef enum {
   QUEUE_MAX,
 } EQueueType;
 
-typedef int32_t (*UpdateDnodeInfoFp)(void* pData, int32_t* dnodeId, int64_t* clusterId, char* fqdn, uint16_t* port);
+typedef void (*UpdateDnodeInfoFp)(void* pData, int32_t* dnodeId, int64_t* clusterId, char* fqdn, uint16_t* port);
 typedef int32_t (*PutToQueueFp)(void* pMgmt, EQueueType qtype, SRpcMsg* pMsg);
 typedef int32_t (*GetQueueSizeFp)(void* pMgmt, int32_t vgId, EQueueType qtype);
 typedef int32_t (*SendReqFp)(const SEpSet* pEpSet, SRpcMsg* pMsg);
@@ -70,7 +70,8 @@ void    tmsgSendRsp(SRpcMsg* pMsg);
 void    tmsgRegisterBrokenLinkArg(SRpcMsg* pMsg);
 void    tmsgReleaseHandle(SRpcHandleInfo* pHandle, int8_t type);
 void    tmsgReportStartup(const char* name, const char* desc);
-int32_t tmsgUpdateDnodeInfo(int32_t* dnodeId, int64_t* clusterId, char* fqdn, uint16_t* port);
+void    tmsgUpdateDnodeInfo(int32_t* dnodeId, int64_t* clusterId, char* fqdn, uint16_t* port);
+void    tmsgUpdateDnodeEpSet(SEpSet* epset);
 
 #ifdef __cplusplus
 }

include/common/tname.h

@@ -78,7 +78,7 @@ typedef struct {
 
   // output
   char*    ctbShortName;  // must have size of TSDB_TABLE_NAME_LEN;
-  uint64_t uid;           // child table uid, may be useful
+//  uint64_t uid;           // child table uid, may be useful
 } RandTableName;
 
 void buildChildTableName(RandTableName* rName);

include/libs/sync/sync.h

@@ -49,10 +49,12 @@ extern "C" {
 #define SYNC_HEARTBEAT_REPLY_SLOW_MS 1500
 #define SYNC_SNAP_RESEND_MS          1000 * 60
 
+#define SYNC_VND_COMMIT_MIN_MS 1000
+
 #define SYNC_MAX_BATCH_SIZE 1
 #define SYNC_INDEX_BEGIN    0
 #define SYNC_INDEX_INVALID  -1
-#define SYNC_TERM_INVALID   -1  // 0xFFFFFFFFFFFFFFFF
+#define SYNC_TERM_INVALID   -1
 
 typedef enum {
   SYNC_STRATEGY_NO_SNAPSHOT = 0,
@@ -191,7 +193,7 @@ typedef struct SSyncLogStore {
   SyncIndex (*syncLogLastIndex)(struct SSyncLogStore* pLogStore);
   SyncTerm (*syncLogLastTerm)(struct SSyncLogStore* pLogStore);
 
-  int32_t (*syncLogAppendEntry)(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry);
+  int32_t (*syncLogAppendEntry)(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry, bool forcSync);
   int32_t (*syncLogGetEntry)(struct SSyncLogStore* pLogStore, SyncIndex index, SSyncRaftEntry** ppEntry);
   int32_t (*syncLogTruncate)(struct SSyncLogStore* pLogStore, SyncIndex fromIndex);
 

include/libs/transport/thttp.h

@@ -24,7 +24,7 @@ extern "C" {
 
 typedef enum { HTTP_GZIP, HTTP_FLAT } EHttpCompFlag;
 
-int32_t taosSendHttpReport(const char* server, uint16_t port, char* pCont, int32_t contLen, EHttpCompFlag flag);
+int32_t taosSendHttpReport(const char* server, const char* uri, uint16_t port, char* pCont, int32_t contLen, EHttpCompFlag flag);
 
 #ifdef __cplusplus
 }

include/os/osSystem.h

@@ -46,27 +46,73 @@ void    taosSetTerminalMode();
 int32_t taosGetOldTerminalMode();
 void    taosResetTerminalMode();
 
+#define STACKSIZE 100
+
 #if !defined(WINDOWS)
-#define taosPrintTrace(flags, level, dflag)                                \
+#define taosLogTraceToBuf(buf, bufSize, ignoreNum) {                                                                      \
-  {                                                                        \
+  void*   array[STACKSIZE];                                                                                                     \
-    void*   array[100];                                                    \
+  int32_t size = backtrace(array, STACKSIZE);                                                                                   \
-    int32_t size = backtrace(array, 100);                                  \
+  char**  strings = backtrace_symbols(array, size);                                                                       \
-    char**  strings = backtrace_symbols(array, size);                      \
+  int32_t offset = 0;                                                                                                     \
-    if (strings != NULL) {                                                 \
+  if (strings != NULL) {                                                                                                  \
-      taosPrintLog(flags, level, dflag, "obtained %d stack frames", size); \
+    offset = snprintf(buf, bufSize - 1, "obtained %d stack frames\n", (ignoreNum > 0) ? size - ignoreNum : size);           \
-      for (int32_t i = 0; i < size; i++) {                                 \
+    for (int32_t i = (ignoreNum > 0) ? ignoreNum : 0; i < size; i++) {                                                    \
-        taosPrintLog(flags, level, dflag, "frame:%d, %s", i, strings[i]);  \
+      offset += snprintf(buf + offset, bufSize - 1 - offset, "frame:%d, %s\n", (ignoreNum > 0) ? i - ignoreNum : i, strings[i]);     \
-      }                                                                    \
+    }                                                                                                                     \
-    }                                                                      \
+  }                                                                                                                       \
-                                                                           \
+                                                                                                                          \
-    taosMemoryFree(strings);                                               \
+  taosMemoryFree(strings);                                                                                                \
+}
+
+#define taosPrintTrace(flags, level, dflag, ignoreNum)                                                          \
+  {                                                                                                             \
+    void*   array[STACKSIZE];                                                                                         \
+    int32_t size = backtrace(array, STACKSIZE);                                                                       \
+    char**  strings = backtrace_symbols(array, size);                                                           \
+    if (strings != NULL) {                                                                                      \
+      taosPrintLog(flags, level, dflag, "obtained %d stack frames", (ignoreNum > 0) ? size - ignoreNum : size); \
+      for (int32_t i = (ignoreNum > 0) ? ignoreNum : 0; i < size; i++) {                                        \
+        taosPrintLog(flags, level, dflag, "frame:%d, %s", (ignoreNum > 0) ? i - ignoreNum : i, strings[i]);     \
+      }                                                                                                         \
+    }                                                                                                           \
+                                                                                                                \
+    taosMemoryFree(strings);                                                                                    \
   }
 #else
+
 #include <windows.h>
 #include <dbghelp.h>
 
-#define STACKSIZE 64
+#define taosLogTraceToBuf(buf, bufSize, ignoreNum) {                                                                                                        \
-#define taosPrintTrace(flags, level, dflag)                                                                \
+    unsigned int   i;                                                                                                 \
+    void*          stack[STACKSIZE];                                                                                  \
+    unsigned short frames;                                                                                            \
+    SYMBOL_INFO*   symbol;                                                                                            \
+    HANDLE         process;                                                                                           \
+    int32_t        offset = 0;                                                                                        \
+                                                                                                                      \
+    process = GetCurrentProcess();                                                                                    \
+                                                                                                                      \
+    SymInitialize(process, NULL, TRUE);                                                                               \
+                                                                                                                      \
+    frames = CaptureStackBackTrace(0, STACKSIZE, stack, NULL);                                                        \
+    symbol = (SYMBOL_INFO*)calloc(sizeof(SYMBOL_INFO) + 256 * sizeof(char), 1);                                       \
+    if (symbol != NULL) {                                                                                             \
+        symbol->MaxNameLen = 255;                                                                                     \
+        symbol->SizeOfStruct = sizeof(SYMBOL_INFO);                                                                   \
+                                                                                                                      \
+        if (frames > 0) {                                                                                             \
+          offset = snprintf(buf, bufSize - 1, "obtained %d stack frames\n", (ignoreNum > 0) ? frames - ignoreNum : frames);                                    \
+          for (i = (ignoreNum > 0) ? ignoreNum : 0; i < frames; i++) {                                                                              \
+            SymFromAddr(process, (DWORD64)(stack[i]), 0, symbol);                                                     \
+            offset += snprintf(buf + offset, bufSize - 1 - offset, "frame:%i, %s - 0x%0X\n", (ignoreNum > 0) ? i - ignoreNum : i, symbol->Name, symbol->Address); \
+          }                                                                                                           \
+        }                                                                                                             \
+        free(symbol);                                                                                                 \
+    }                                                                                                                 \
+  }
+
+#define taosPrintTrace(flags, level, dflag, ignoreNum)                                                                \
   {                                                                                                        \
     unsigned int   i;                                                                                                 \
     void*          stack[STACKSIZE];                                                                                  \
@@ -85,10 +131,10 @@ void    taosResetTerminalMode();
         symbol->SizeOfStruct = sizeof(SYMBOL_INFO);                                                                   \
                                                                                                                       \
         if (frames > 0) {                                                                                             \
-          taosPrintLog(flags, level, dflag, "obtained %d stack frames", frames);                                      \
+          taosPrintLog(flags, level, dflag, "obtained %d stack frames\n", (ignoreNum > 0) ? frames - ignoreNum : frames);                                      \
-          for (i = 0; i < frames; i++) {                                                                              \
+          for (i = (ignoreNum > 0) ? ignoreNum : 0; i < frames; i++) {                                                                              \
             SymFromAddr(process, (DWORD64)(stack[i]), 0, symbol);                                                     \
-            taosPrintLog(flags, level, dflag, "frame:%i: %s - 0x%0X", frames - i - 1, symbol->Name, symbol->Address); \
+            taosPrintLog(flags, level, dflag, "frame:%i, %s - 0x%0X\n", (ignoreNum > 0) ? i - ignoreNum : i, symbol->Name, symbol->Address); \
           }                                                                                                           \
         }                                                                                                             \
         free(symbol);                                                                                                 \

include/util/tdef.h

@@ -499,7 +499,7 @@ enum {
 #define DEFAULT_PAGESIZE 4096
 
 #define VNODE_TIMEOUT_SEC 60
-#define MNODE_TIMEOUT_SEC 10
+#define MNODE_TIMEOUT_SEC 60
 
 #ifdef __cplusplus
 }

include/util/tlog.h

@@ -99,6 +99,11 @@ bool taosAssertRelease(bool condition);
 #endif
 #endif
 
+void taosLogCrashInfo(char* nodeType, char* pMsg, int64_t msgLen, int signum, void *sigInfo);
+void taosReadCrashInfo(char* filepath, char** pMsg, int64_t* pMsgLen, TdFilePtr* pFd);
+void taosReleaseCrashLogFile(TdFilePtr pFile, bool truncateFile);
+int32_t taosGenCrashJsonMsg(int signum, char** pMsg, int64_t clusterId, int64_t startTime);
+
 // clang-format off
 #define uFatal(...) { if (uDebugFlag & DEBUG_FATAL) { taosPrintLog("UTL FATAL", DEBUG_FATAL, tsLogEmbedded ? 255 : uDebugFlag, __VA_ARGS__); }}
 #define uError(...) { if (uDebugFlag & DEBUG_ERROR) { taosPrintLog("UTL ERROR ", DEBUG_ERROR, tsLogEmbedded ? 255 : uDebugFlag, __VA_ARGS__); }}

packaging/cfg/taos.cfg

@@ -43,6 +43,9 @@
 # Switch for allowing TDengine to collect and report service usage information
 # telemetryReporting        1
 
+# Switch for allowing TDengine to collect and report crash information
+# crashReporting            1
+
 # The maximum number of vnodes supported by this dnode
 # supportVnodes             0
 

packaging/release.sh

@@ -273,7 +273,7 @@ if [ "$osType" != "Darwin" ]; then
           [ -z "$taos_tools_ver" ] && taos_tools_ver="0.1.0"
 
           ${csudo}./make-taos-tools-deb.sh ${top_dir} \
-            ${compile_dir} ${output_dir} ${taos_tools_ver} ${cpuType} ${osType} ${verMode} ${verType}
+            ${compile_dir} ${output_dir} ${taos_tools_ver} ${cpuType} ${osType} ${verMode} ${verType} ${verNumberComp}
         fi
       fi
     else
@@ -298,7 +298,7 @@ if [ "$osType" != "Darwin" ]; then
           [ -z "$taos_tools_ver" ] && taos_tools_ver="0.1.0"
 
           ${csudo}./make-taos-tools-rpm.sh ${top_dir} \
-            ${compile_dir} ${output_dir} ${taos_tools_ver} ${cpuType} ${osType} ${verMode} ${verType}
+            ${compile_dir} ${output_dir} ${taos_tools_ver} ${cpuType} ${osType} ${verMode} ${verType} ${verNumberComp}
         fi
       fi
     else

packaging/tools/install.sh

@@ -481,11 +481,11 @@ function install_adapter_config() {
     ${csudo}mkdir -p ${cfg_install_dir}
     [ -f ${script_dir}/cfg/${adapterName}.toml ] && ${csudo}cp ${script_dir}/cfg/${adapterName}.toml ${cfg_install_dir}
     [ -f ${cfg_install_dir}/${adapterName}.toml ] && ${csudo}chmod 644 ${cfg_install_dir}/${adapterName}.toml
+  else
+    [ -f ${script_dir}/cfg/${adapterName}.toml ] &&
+      ${csudo}cp -f ${script_dir}/cfg/${adapterName}.toml ${cfg_install_dir}/${adapterName}.toml.new
   fi
 
-  [ -f ${script_dir}/cfg/${adapterName}.toml ] &&
-    ${csudo}cp -f ${script_dir}/cfg/${adapterName}.toml ${cfg_install_dir}/${adapterName}.toml.new
-
   [ -f ${cfg_install_dir}/${adapterName}.toml ] &&
     ${csudo}ln -s ${cfg_install_dir}/${adapterName}.toml ${install_main_dir}/cfg/${adapterName}.toml
 
@@ -499,9 +499,10 @@ function install_config() {
     ${csudo}mkdir -p ${cfg_install_dir}
     [ -f ${script_dir}/cfg/${configFile} ] && ${csudo}cp ${script_dir}/cfg/${configFile} ${cfg_install_dir}
     ${csudo}chmod 644 ${cfg_install_dir}/*
+  else
+    ${csudo}cp -f ${script_dir}/cfg/${configFile} ${cfg_install_dir}/${configFile}.new
   fi
 
-  ${csudo}cp -f ${script_dir}/cfg/${configFile} ${cfg_install_dir}/${configFile}.new
   ${csudo}ln -s ${cfg_install_dir}/${configFile} ${install_main_dir}/cfg
 
   [ ! -z $1 ] && return 0 || : # only install client
@@ -742,6 +743,34 @@ function is_version_compatible() {
   esac
 }
 
+deb_erase() {
+  confirm=""
+  while [ "" == "${confirm}" ]; do
+    echo -e -n "${RED}Exist tdengine deb detected, do you want to remove it? [yes|no] ${NC}:"
+    read confirm
+    if [ "yes" == "$confirm" ]; then
+      ${csudo}dpkg --remove tdengine ||:
+      break
+    elif [ "no" == "$confirm" ]; then
+      break
+    fi
+  done
+}
+
+rpm_erase() {
+  confirm=""
+  while [ "" == "${confirm}" ]; do
+    echo -e -n "${RED}Exist tdengine rpm detected, do you want to remove it? [yes|no] ${NC}:"
+    read confirm
+    if [ "yes" == "$confirm" ]; then
+      ${csudo}rpm -e tdengine ||:
+      break
+    elif [ "no" == "$confirm" ]; then
+      break
+    fi
+  done
+}
+
 function updateProduct() {
   # Check if version compatible
   if ! is_version_compatible; then
@@ -754,6 +783,13 @@ function updateProduct() {
     echo "File ${tarName} does not exist"
     exit 1
   fi
+
+  if echo $osinfo | grep -qwi "centos"; then
+    rpm -q tdengine 2>&1 > /dev/null && rpm_erase tdengine ||:
+  elif echo $osinfo | grep -qwi "ubuntu"; then
+    dpkg -l tdengine 2>&1 > /dev/null && deb_erase tdengine ||:
+  fi
+
   tar -zxf ${tarName}
   install_jemalloc
 

source/client/inc/clientInt.h

@@ -313,6 +313,8 @@ extern SAppInfo appInfo;
 extern int32_t  clientReqRefPool;
 extern int32_t  clientConnRefPool;
 extern int32_t  timestampDeltaLimit;
+extern int64_t  lastClusterId;
+
 
 __async_send_cb_fn_t getMsgRspHandle(int32_t msgType);
 
@@ -340,6 +342,7 @@ void  resetConnectDB(STscObj* pTscObj);
 int taos_options_imp(TSDB_OPTION option, const char* str);
 
 void* openTransporter(const char* user, const char* auth, int32_t numOfThreads);
+void tscStopCrashReport();
 
 typedef struct AsyncArg {
   SRpcMsg msg;

source/client/src/clientEnv.c

@@ -28,13 +28,16 @@
 #include "trpc.h"
 #include "tsched.h"
 #include "ttime.h"
+#include "thttp.h"
 
 #define TSC_VAR_NOT_RELEASE 1
 #define TSC_VAR_RELEASED    0
 
 SAppInfo appInfo;
+int64_t  lastClusterId = 0;
 int32_t  clientReqRefPool = -1;
 int32_t  clientConnRefPool = -1;
+int32_t  clientStop = 0;
 
 int32_t timestampDeltaLimit = 900;  // s
 
@@ -354,6 +357,7 @@ void doDestroyRequest(void *p) {
   taosMemoryFreeClear(pRequest->pDb);
 
   doFreeReqResultInfo(&pRequest->body.resInfo);
+  tsem_destroy(&pRequest->body.rspSem);
 
   taosArrayDestroy(pRequest->tableList);
   taosArrayDestroy(pRequest->dbList);
@@ -368,6 +372,9 @@ void doDestroyRequest(void *p) {
   }
 
   if (pRequest->syncQuery) {
+      if (pRequest->body.param){
+        tsem_destroy(&((SSyncQueryParam*)pRequest->body.param)->sem);
+      }
     taosMemoryFree(pRequest->body.param);
   }
 
@@ -385,6 +392,122 @@ void destroyRequest(SRequestObj *pRequest) {
   removeRequest(pRequest->self);
 }
 
+void crashReportThreadFuncUnexpectedStopped(void) { atomic_store_32(&clientStop, -1); }
+
+static void *tscCrashReportThreadFp(void *param) {
+  setThreadName("client-crashReport");
+  char filepath[PATH_MAX] = {0};
+  snprintf(filepath, sizeof(filepath), "%s%s.taosCrashLog", tsLogDir, TD_DIRSEP);
+  char *pMsg = NULL;
+  int64_t msgLen = 0;
+  TdFilePtr pFile = NULL;
+  bool truncateFile = false;
+  int32_t sleepTime = 200;
+  int32_t reportPeriodNum = 3600 * 1000 / sleepTime;
+  int32_t loopTimes = reportPeriodNum;
+
+#ifdef WINDOWS
+  if (taosCheckCurrentInDll()) {
+    atexit(crashReportThreadFuncUnexpectedStopped);
+  }
+#endif
+  
+  while (1) {
+    if (clientStop) break;
+    if (loopTimes++ < reportPeriodNum) {
+      taosMsleep(sleepTime);
+      continue;
+    }
+
+    taosReadCrashInfo(filepath, &pMsg, &msgLen, &pFile);
+    if (pMsg && msgLen > 0) {
+      if (taosSendHttpReport(tsTelemServer, tsClientCrashReportUri, tsTelemPort, pMsg, msgLen, HTTP_FLAT) != 0) {
+        tscError("failed to send crash report");
+        if (pFile) {
+          taosReleaseCrashLogFile(pFile, false);
+          continue;
+        }
+      } else {
+        tscInfo("succeed to send crash report");
+        truncateFile = true;
+      }
+    } else {
+      tscDebug("no crash info");
+    }
+
+    taosMemoryFree(pMsg);
+
+    if (pMsg && msgLen > 0) {
+      pMsg = NULL;
+      continue;
+    }
+    
+    if (pFile) {
+      taosReleaseCrashLogFile(pFile, truncateFile);
+      truncateFile = false;
+    }
+    
+    taosMsleep(sleepTime);
+    loopTimes = 0;
+  }
+
+  clientStop = -1;
+  return NULL;
+}
+
+int32_t tscCrashReportInit() {
+  if (!tsEnableCrashReport) {
+    return 0;
+  }
+  
+  TdThreadAttr thAttr;
+  taosThreadAttrInit(&thAttr);
+  taosThreadAttrSetDetachState(&thAttr, PTHREAD_CREATE_JOINABLE);
+  TdThread            crashReportThread;
+  if (taosThreadCreate(&crashReportThread, &thAttr, tscCrashReportThreadFp, NULL) != 0) {
+    tscError("failed to create crashReport thread since %s", strerror(errno));
+    return -1;
+  }
+
+  taosThreadAttrDestroy(&thAttr);
+  return 0;
+}
+
+void tscStopCrashReport() {
+  if (!tsEnableCrashReport) {
+    return;
+  }
+
+  if (atomic_val_compare_exchange_32(&clientStop, 0, 1)) {
+    tscDebug("hb thread already stopped");
+    return;
+  }
+
+  while (atomic_load_32(&clientStop) > 0) {
+    taosMsleep(100);
+  }
+}
+
+
+void tscWriteCrashInfo(int signum, void *sigInfo, void *context) {
+  char *pMsg = NULL;
+  const char *flags = "UTL FATAL ";
+  ELogLevel   level = DEBUG_FATAL;
+  int32_t     dflag = 255;
+  int64_t     msgLen= -1;
+
+  if (tsEnableCrashReport) {
+    if (taosGenCrashJsonMsg(signum, &pMsg, lastClusterId, appInfo.startTime)) {
+      taosPrintLog(flags, level, dflag, "failed to generate crash json msg");
+    } else {
+      msgLen = strlen(pMsg);  
+    }
+  }
+
+  taosLogCrashInfo("taos", pMsg, msgLen, signum, sigInfo);
+}
+
+
 void taos_init_imp(void) {
   // In the APIs of other program language, taos_cleanup is not available yet.
   // So, to make sure taos_cleanup will be invoked to clean up the allocated resource to suppress the valgrind warning.
@@ -392,6 +515,10 @@ void taos_init_imp(void) {
   errno = TSDB_CODE_SUCCESS;
   taosSeedRand(taosGetTimestampSec());
 
+  appInfo.pid = taosGetPId();
+  appInfo.startTime = taosGetTimestampMs();
+  appInfo.pInstMap = taosHashInit(4, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK);
+
   deltaToUtcInitOnce();
 
   if (taosCreateLog("taoslog", 10, configDir, NULL, NULL, NULL, NULL, 1) != 0) {
@@ -433,9 +560,8 @@ void taos_init_imp(void) {
   taosGetAppName(appInfo.appName, NULL);
   taosThreadMutexInit(&appInfo.mutex, NULL);
 
-  appInfo.pid = taosGetPId();
+  tscCrashReportInit();
-  appInfo.startTime = taosGetTimestampMs();
+  
-  appInfo.pInstMap = taosHashInit(4, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK);
   tscDebug("client is initialized successfully");
 }
 

source/client/src/clientImpl.c

@@ -159,6 +159,12 @@ STscObj* taos_connect_internal(const char* ip, const char* user, const char* pas
   return taosConnectImpl(user, &secretEncrypt[0], localDb, NULL, NULL, *pInst, connType);
 }
 
+void freeQueryParam(SSyncQueryParam* param) {
+  if (param == NULL) return;
+  tsem_destroy(&param->sem);
+  taosMemoryFree(param);
+}
+
 int32_t buildRequest(uint64_t connId, const char* sql, int sqlLen, void* param, bool validateSql,
                      SRequestObj** pRequest, int64_t reqid) {
   *pRequest = createRequest(connId, TSDB_SQL_SELECT, reqid);
@@ -180,17 +186,18 @@ int32_t buildRequest(uint64_t connId, const char* sql, int sqlLen, void* param,
   (*pRequest)->sqlLen = sqlLen;
   (*pRequest)->validateOnly = validateSql;
 
+  SSyncQueryParam* newpParam;
   if (param == NULL) {
-    SSyncQueryParam* pParam = taosMemoryCalloc(1, sizeof(SSyncQueryParam));
+    newpParam = taosMemoryCalloc(1, sizeof(SSyncQueryParam));
-    if (pParam == NULL) {
+    if (newpParam == NULL) {
       destroyRequest(*pRequest);
       *pRequest = NULL;
       return TSDB_CODE_OUT_OF_MEMORY;
     }
 
-    tsem_init(&pParam->sem, 0, 0);
+    tsem_init(&newpParam->sem, 0, 0);
-    pParam->pRequest = (*pRequest);
+    newpParam->pRequest = (*pRequest);
-    param = pParam;
+    param = newpParam;
   }
 
   (*pRequest)->body.param = param;
@@ -201,8 +208,7 @@ int32_t buildRequest(uint64_t connId, const char* sql, int sqlLen, void* param,
   if (err) {
     tscError("%" PRId64 " failed to add to request container, reqId:0x%" PRIx64 ", conn:%" PRId64 ", %s",
              (*pRequest)->self, (*pRequest)->requestId, pTscObj->id, sql);
-
+    freeQueryParam(newpParam);
-    taosMemoryFree(param);
     destroyRequest(*pRequest);
     *pRequest = NULL;
     return TSDB_CODE_OUT_OF_MEMORY;
@@ -214,6 +220,7 @@ int32_t buildRequest(uint64_t connId, const char* sql, int sqlLen, void* param,
         nodesCreateAllocator((*pRequest)->requestId, tsQueryNodeChunkSize, &((*pRequest)->allocatorRefId))) {
       tscError("%" PRId64 " failed to create node allocator, reqId:0x%" PRIx64 ", conn:%" PRId64 ", %s",
                (*pRequest)->self, (*pRequest)->requestId, pTscObj->id, sql);
+      freeQueryParam(newpParam);
       destroyRequest(*pRequest);
       *pRequest = NULL;
       return TSDB_CODE_OUT_OF_MEMORY;

source/client/src/clientMain.c

@@ -55,6 +55,8 @@ void taos_cleanup(void) {
     return;
   }
 
+  tscStopCrashReport();
+
   int32_t id = clientReqRefPool;
   clientReqRefPool = -1;
   taosCloseRef(id);
@@ -106,7 +108,7 @@ TAOS *taos_connect(const char *ip, const char *user, const char *pass, const cha
   if (pass == NULL) {
     pass = TSDB_DEFAULT_PASS;
   }
-
+  
   STscObj *pObj = taos_connect_internal(ip, user, pass, NULL, db, port, CONN_TYPE__QUERY);
   if (pObj) {
     int64_t *rid = taosMemoryCalloc(1, sizeof(int64_t));
@@ -507,9 +509,8 @@ void taos_stop_query(TAOS_RES *res) {
   SRequestObj *pRequest = (SRequestObj *)res;
   pRequest->killed = true;
 
-  int32_t numOfFields = taos_num_fields(pRequest);
   // It is not a query, no need to stop.
-  if (numOfFields == 0) {
+  if (NULL == pRequest->pQuery || QUERY_EXEC_MODE_SCHEDULE != pRequest->pQuery->execMode) {
     tscDebug("request 0x%" PRIx64 " no need to be killed since not query", pRequest->requestId);
     return;
   }

source/client/src/clientMsgHandler.c

@@ -119,6 +119,7 @@ int32_t processConnectRsp(void* param, SDataBuf* pMsg, int32_t code) {
 
   // update the appInstInfo
   pTscObj->pAppInfo->clusterId = connectRsp.clusterId;
+  lastClusterId = connectRsp.clusterId;
 
   pTscObj->connType = connectRsp.connType;
 

source/client/src/clientRawBlockWrite.c

@@ -1448,6 +1448,7 @@ int taos_write_raw_block_with_fields(TAOS* taos, int rows, char* pData, const ch
   end:
   taosMemoryFreeClear(pTableMeta);
   qDestroyQuery(pQuery);
+  destroyRequest(pRequest);
   taosMemoryFree(subReq);
   return code;
 }
@@ -1639,6 +1640,7 @@ int taos_write_raw_block(TAOS* taos, int rows, char* pData, const char* tbname)
   end:
   taosMemoryFreeClear(pTableMeta);
   qDestroyQuery(pQuery);
+  destroyRequest(pRequest);
   taosMemoryFree(subReq);
   return code;
 }

source/client/src/clientSml.c

@@ -182,6 +182,7 @@ typedef struct {
   SSmlMsgBuf   msgBuf;
   SHashObj    *dumplicateKey;  // for dumplicate key
   SArray      *colsContainer;  // for cols parse, if dataFormat == false
+  int32_t      uid;            // used for automatic create child table
 
   cJSON       *root;  // for parse json
 } SSmlHandle;
@@ -2155,13 +2156,11 @@ static int32_t smlParseInfluxLine(SSmlHandle *info, const char *sql, const int l
     (*oneTable)->sTableNameLen = elements.measureLen;
     if (strlen((*oneTable)->childTableName) == 0) {
       RandTableName rName = {(*oneTable)->tags, (*oneTable)->sTableName, (uint8_t)(*oneTable)->sTableNameLen,
-                             (*oneTable)->childTableName, 0};
+                             (*oneTable)->childTableName};
 
       buildChildTableName(&rName);
-      (*oneTable)->uid = rName.uid;
-    } else {
-      (*oneTable)->uid = *(uint64_t *)((*oneTable)->childTableName);
     }
+    (*oneTable)->uid = info->uid++;
   }
 
   SSmlSTableMeta **tableMeta = (SSmlSTableMeta **)taosHashGet(info->superTables, elements.measure, elements.measureLen);
@@ -2226,11 +2225,8 @@ static int32_t smlParseTelnetLine(SSmlHandle *info, void *data, const int len) {
   taosHashClear(info->dumplicateKey);
 
   if (strlen(tinfo->childTableName) == 0) {
-    RandTableName rName = {tinfo->tags, tinfo->sTableName, (uint8_t)tinfo->sTableNameLen, tinfo->childTableName, 0};
+    RandTableName rName = {tinfo->tags, tinfo->sTableName, (uint8_t)tinfo->sTableNameLen, tinfo->childTableName};
     buildChildTableName(&rName);
-    tinfo->uid = rName.uid;
-  } else {
-    tinfo->uid = *(uint64_t *)(tinfo->childTableName);  // generate uid by name simple
   }
 
   bool            hasTable = true;
@@ -2239,6 +2235,7 @@ static int32_t smlParseTelnetLine(SSmlHandle *info, void *data, const int len) {
   if (!oneTable) {
     taosHashPut(info->childTables, tinfo->childTableName, strlen(tinfo->childTableName), &tinfo, POINTER_BYTES);
     oneTable = &tinfo;
+    tinfo->uid = info->uid++;
     hasTable = false;
   } else {
     smlDestroyTableInfo(info, tinfo);

source/client/src/clientStmt.c

@@ -300,11 +300,7 @@ int32_t stmtCleanExecInfo(STscStmt* pStmt, bool keepTable, bool deepClean) {
       continue;
     }
 
-    if (STMT_TYPE_MULTI_INSERT == pStmt->sql.type) {
+    qDestroyStmtDataBlock(pBlocks);
-      qFreeStmtDataBlock(pBlocks);
-    } else {
-      qDestroyStmtDataBlock(pBlocks);
-    }
     taosHashRemove(pStmt->exec.pBlockHash, key, keyLen);
 
     pIter = taosHashIterate(pStmt->exec.pBlockHash, pIter);

source/common/src/systable.c

@@ -67,6 +67,8 @@ static const SSysDbTableSchema clusterSchema[] = {
     {.name = "name", .bytes = TSDB_CLUSTER_ID_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true},
     {.name = "uptime", .bytes = 4, .type = TSDB_DATA_TYPE_INT, .sysInfo = true},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP, .sysInfo = true},
+    {.name = "version", .bytes = 10 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true},  
+    {.name = "expire_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP, .sysInfo = true},
 };
 
 static const SSysDbTableSchema userDBSchema[] = {

source/common/src/tdatablock.c

@@ -69,7 +69,7 @@ int32_t colDataAppend(SColumnInfoData* pColumnInfoData, uint32_t currentRow, con
     if (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) {
       pColumnInfoData->varmeta.offset[currentRow] = -1;  // it is a null value of VAR type.
     } else {
-      colDataSetNull_f(pColumnInfoData->nullbitmap, currentRow);
+      colDataSetNull_f_s(pColumnInfoData, currentRow);
     }
 
     pColumnInfoData->hasNull = true;
@@ -825,7 +825,7 @@ static int32_t blockDataAssign(SColumnInfoData* pCols, const SSDataBlock* pDataB
     } else {
       for (int32_t j = 0; j < pDataBlock->info.rows; ++j) {
         if (colDataIsNull_f(pSrc->nullbitmap, index[j])) {
-          colDataSetNull_f(pDst->nullbitmap, j);
+          colDataSetNull_f_s(pDst, j);
           continue;
         }
         memcpy(pDst->pData + j * pDst->info.bytes, pSrc->pData + index[j] * pDst->info.bytes, pDst->info.bytes);
@@ -1161,15 +1161,16 @@ void blockDataEmpty(SSDataBlock* pDataBlock) {
   pInfo->window.skey = 0;
 }
 
-// todo temporarily disable it
+/*
+ * NOTE: the type of the input column may be TSDB_DATA_TYPE_NULL, which is used to denote
+ * the all NULL value in this column. It is an internal representation of all NULL value column, and no visible to
+ * any users. The length of TSDB_DATA_TYPE_NULL is 0, and it is an special case.
+ */
 static int32_t doEnsureCapacity(SColumnInfoData* pColumn, const SDataBlockInfo* pBlockInfo, uint32_t numOfRows, bool clearPayload) {
   if (numOfRows <= 0 || numOfRows <= pBlockInfo->capacity) {
     return TSDB_CODE_SUCCESS;
   }
 
-  // todo temp disable it
-  //  ASSERT(pColumn->info.bytes != 0);
-
   int32_t existedRows = pBlockInfo->rows;
 
   if (IS_VAR_DATA_TYPE(pColumn->info.type)) {
@@ -1194,7 +1195,8 @@ static int32_t doEnsureCapacity(SColumnInfoData* pColumn, const SDataBlockInfo*
       return TSDB_CODE_FAILED;
     }
 
-    // make sure the allocated memory is MALLOC_ALIGN_BYTES aligned
+    // here we employ the aligned malloc function, to make sure that the address of allocated memory is aligned
+    // to MALLOC_ALIGN_BYTES
     tmp = taosMemoryMallocAlign(MALLOC_ALIGN_BYTES, numOfRows * pColumn->info.bytes);
     if (tmp == NULL) {
       return TSDB_CODE_OUT_OF_MEMORY;
@@ -1208,7 +1210,7 @@ static int32_t doEnsureCapacity(SColumnInfoData* pColumn, const SDataBlockInfo*
 
     pColumn->pData = tmp;
 
-    // todo remove it soon
+    // check if the allocated memory is aligned to the requried bytes.
 #if defined LINUX
     if ((((uint64_t)pColumn->pData) & (MALLOC_ALIGN_BYTES - 1)) != 0x0) {
       return TSDB_CODE_FAILED;
@@ -1249,25 +1251,6 @@ int32_t blockDataEnsureCapacity(SSDataBlock* pDataBlock, uint32_t numOfRows) {
     return TSDB_CODE_SUCCESS;
   }
 
-  size_t numOfCols = taosArrayGetSize(pDataBlock->pDataBlock);
-  for (int32_t i = 0; i < numOfCols; ++i) {
-    SColumnInfoData* p = taosArrayGet(pDataBlock->pDataBlock, i);
-    code = doEnsureCapacity(p, &pDataBlock->info, numOfRows, true);
-    if (code) {
-      return code;
-    }
-  }
-
-  pDataBlock->info.capacity = numOfRows;
-  return TSDB_CODE_SUCCESS;
-}
-
-int32_t blockDataEnsureCapacityNoClear(SSDataBlock* pDataBlock, uint32_t numOfRows) {
-  int32_t code = 0;
-  if (numOfRows == 0 || numOfRows <= pDataBlock->info.capacity) {
-    return TSDB_CODE_SUCCESS;
-  }
-
   size_t numOfCols = taosArrayGetSize(pDataBlock->pDataBlock);
   for (int32_t i = 0; i < numOfCols; ++i) {
     SColumnInfoData* p = taosArrayGet(pDataBlock->pDataBlock, i);

source/common/src/tglobal.c

@@ -60,6 +60,9 @@ int32_t tsElectInterval = 25 * 1000;
 int32_t tsHeartbeatInterval = 1000;
 int32_t tsHeartbeatTimeout = 20 * 1000;
 
+// vnode
+int64_t tsVndCommitMaxIntervalMs = 60 * 1000;
+
 // monitor
 bool     tsEnableMonitor = true;
 int32_t  tsMonitorInterval = 30;
@@ -73,6 +76,11 @@ bool     tsEnableTelem = true;
 int32_t  tsTelemInterval = 43200;
 char     tsTelemServer[TSDB_FQDN_LEN] = "telemetry.taosdata.com";
 uint16_t tsTelemPort = 80;
+char*    tsTelemUri = "/report";
+
+bool     tsEnableCrashReport = true;
+char*    tsClientCrashReportUri = "/ccrashreport";
+char*    tsSvrCrashReportUri = "/dcrashreport";
 
 // schemaless
 char tsSmlTagName[TSDB_COL_NAME_LEN] = "_tag_null";
@@ -202,7 +210,9 @@ int32_t taosSetTfsCfg(SConfig *pCfg) {
 int32_t taosSetTfsCfg(SConfig *pCfg);
 #endif
 
-struct SConfig *taosGetCfg() { return tsCfg; }
+struct SConfig *taosGetCfg() {
+  return tsCfg;
+}
 
 static int32_t taosLoadCfg(SConfig *pCfg, const char **envCmd, const char *inputCfgDir, const char *envFile,
                            char *apolloUrl) {
@@ -314,6 +324,7 @@ static int32_t taosAddClientCfg(SConfig *pCfg) {
   if (cfgAddInt32(pCfg, "maxMemUsedByInsert", tsMaxMemUsedByInsert, 1, INT32_MAX, true) != 0) return -1;
   if (cfgAddInt32(pCfg, "maxRetryWaitTime", tsMaxRetryWaitTime, 0, 86400000, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "useAdapter", tsUseAdapter, true) != 0) return -1;
+  if (cfgAddBool(pCfg, "crashReporting", tsEnableCrashReport, true) != 0) return -1;
 
   tsNumOfTaskQueueThreads = tsNumOfCores / 2;
   tsNumOfTaskQueueThreads = TMAX(tsNumOfTaskQueueThreads, 4);
@@ -377,7 +388,7 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddInt32(pCfg, "queryRspPolicy", tsQueryRspPolicy, 0, 1, 0) != 0) return -1;
 
   tsNumOfRpcThreads = tsNumOfCores / 2;
-  tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 1, 4);
+  tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 1, TSDB_MAX_RPC_THREADS);
   if (cfgAddInt32(pCfg, "numOfRpcThreads", tsNumOfRpcThreads, 1, 1024, 0) != 0) return -1;
 
   tsNumOfCommitThreads = tsNumOfCores / 2;
@@ -427,6 +438,8 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddInt32(pCfg, "syncHeartbeatInterval", tsHeartbeatInterval, 10, 1000 * 60 * 24 * 2, 0) != 0) return -1;
   if (cfgAddInt32(pCfg, "syncHeartbeatTimeout", tsHeartbeatTimeout, 10, 1000 * 60 * 24 * 2, 0) != 0) return -1;
 
+  if (cfgAddInt64(pCfg, "vndCommitMaxInterval", tsVndCommitMaxIntervalMs, 1000, 1000 * 60 * 60, 0) != 0) return -1;
+
   if (cfgAddBool(pCfg, "monitor", tsEnableMonitor, 0) != 0) return -1;
   if (cfgAddInt32(pCfg, "monitorInterval", tsMonitorInterval, 1, 200000, 0) != 0) return -1;
   if (cfgAddString(pCfg, "monitorFqdn", tsMonitorFqdn, 0) != 0) return -1;
@@ -434,6 +447,7 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddInt32(pCfg, "monitorMaxLogs", tsMonitorMaxLogs, 1, 1000000, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "monitorComp", tsMonitorComp, 0) != 0) return -1;
 
+  if (cfgAddBool(pCfg, "crashReporting", tsEnableCrashReport, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "telemetryReporting", tsEnableTelem, 0) != 0) return -1;
   if (cfgAddInt32(pCfg, "telemetryInterval", tsTelemInterval, 1, 200000, 0) != 0) return -1;
   if (cfgAddString(pCfg, "telemetryServer", tsTelemServer, 0) != 0) return -1;
@@ -665,6 +679,7 @@ static int32_t taosSetClientCfg(SConfig *pCfg) {
   tsQueryUseNodeAllocator = cfgGetItem(pCfg, "queryUseNodeAllocator")->bval;
   tsKeepColumnName = cfgGetItem(pCfg, "keepColumnName")->bval;
   tsUseAdapter = cfgGetItem(pCfg, "useAdapter")->bval;
+  tsEnableCrashReport = cfgGetItem(pCfg, "crashReporting")->bval;
 
   tsMaxRetryWaitTime = cfgGetItem(pCfg, "maxRetryWaitTime")->i32;
   return 0;
@@ -715,7 +730,7 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsNumOfSnodeWriteThreads = cfgGetItem(pCfg, "numOfSnodeUniqueThreads")->i32;
   tsRpcQueueMemoryAllowed = cfgGetItem(pCfg, "rpcQueueMemoryAllowed")->i64;
 
-  tsSIMDBuiltins = (bool) cfgGetItem(pCfg, "SIMD-builtins")->bval;
+  tsSIMDBuiltins = (bool)cfgGetItem(pCfg, "SIMD-builtins")->bval;
 
   tsEnableMonitor = cfgGetItem(pCfg, "monitor")->bval;
   tsMonitorInterval = cfgGetItem(pCfg, "monitorInterval")->i32;
@@ -726,6 +741,7 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsQueryRspPolicy = cfgGetItem(pCfg, "queryRspPolicy")->i32;
 
   tsEnableTelem = cfgGetItem(pCfg, "telemetryReporting")->bval;
+  tsEnableCrashReport = cfgGetItem(pCfg, "crashReporting")->bval;
   tsTelemInterval = cfgGetItem(pCfg, "telemetryInterval")->i32;
   tstrncpy(tsTelemServer, cfgGetItem(pCfg, "telemetryServer")->str, TSDB_FQDN_LEN);
   tsTelemPort = (uint16_t)cfgGetItem(pCfg, "telemetryPort")->i32;
@@ -743,6 +759,8 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsHeartbeatInterval = cfgGetItem(pCfg, "syncHeartbeatInterval")->i32;
   tsHeartbeatTimeout = cfgGetItem(pCfg, "syncHeartbeatTimeout")->i32;
 
+  tsVndCommitMaxIntervalMs = cfgGetItem(pCfg, "vndCommitMaxInterval")->i64;
+
   tsStartUdfd = cfgGetItem(pCfg, "udf")->bval;
   tstrncpy(tsUdfdResFuncs, cfgGetItem(pCfg, "udfdResFuncs")->str, sizeof(tsUdfdResFuncs));
   tstrncpy(tsUdfdLdLibPath, cfgGetItem(pCfg, "udfdLdLibPath")->str, sizeof(tsUdfdLdLibPath));
@@ -795,6 +813,8 @@ int32_t taosSetCfg(SConfig *pCfg, char *name) {
         tsCountAlwaysReturnValue = cfgGetItem(pCfg, "countAlwaysReturnValue")->i32;
       } else if (strcasecmp("cDebugFlag", name) == 0) {
         cDebugFlag = cfgGetItem(pCfg, "cDebugFlag")->i32;
+      } else if (strcasecmp("crashReporting", name) == 0) {
+        tsEnableCrashReport = cfgGetItem(pCfg, "crashReporting")->bval;
       }
       break;
     }

source/common/src/tmsg.c

@@ -2821,8 +2821,8 @@ int32_t tSerializeSDbCfgRsp(void *buf, int32_t bufLen, const SDbCfgRsp *pRsp) {
     if (tEncodeI8(&encoder, pRetension->keepUnit) < 0) return -1;
   }
   if (tEncodeI8(&encoder, pRsp->schemaless) < 0) return -1;
+  if (tEncodeI16(&encoder, pRsp->sstTrigger) < 0) return -1;
   tEndEncode(&encoder);
-
   int32_t tlen = encoder.pos;
   tEncoderClear(&encoder);
   return tlen;
@@ -2873,6 +2873,7 @@ int32_t tDeserializeSDbCfgRsp(void *buf, int32_t bufLen, SDbCfgRsp *pRsp) {
     }
   }
   if (tDecodeI8(&decoder, &pRsp->schemaless) < 0) return -1;
+  if (tDecodeI16(&decoder, &pRsp->sstTrigger) < 0) return -1;
   tEndDecode(&decoder);
 
   tDecoderClear(&decoder);

source/common/src/tname.c

@@ -330,5 +330,4 @@ void buildChildTableName(RandTableName* rName) {
     strcat(rName->ctbShortName, temp);
   }
   taosStringBuilderDestroy(&sb);
-  rName->uid = *(uint64_t*)(context.digest);
 }

source/dnode/mgmt/exe/dmMain.c

@@ -44,6 +44,7 @@ static struct {
   char         apolloUrl[PATH_MAX];
   const char **envCmd;
   SArray      *pArgs;  // SConfigPair
+  int64_t      startTime;
 } global = {0};
 
 static void dmSetDebugFlag(int32_t signum, void *sigInfo, void *context) { taosSetAllDebugFlag(143, true); }
@@ -67,23 +68,71 @@ static void dmStopDnode(int signum, void *sigInfo, void *context) {
   dmStop();
 }
 
+void dmLogCrash(int signum, void *sigInfo, void *context) {
+  taosIgnSignal(SIGTERM);
+  taosIgnSignal(SIGHUP);
+  taosIgnSignal(SIGINT);
+  taosIgnSignal(SIGBREAK);
+
+#ifndef WINDOWS  
+  taosIgnSignal(SIGBUS);
+#endif
+  taosIgnSignal(SIGABRT);
+  taosIgnSignal(SIGFPE);
+  taosIgnSignal(SIGSEGV);
+
+  char *pMsg = NULL;
+  const char *flags = "UTL FATAL ";
+  ELogLevel   level = DEBUG_FATAL;
+  int32_t     dflag = 255;
+  int64_t     msgLen= -1;
+  
+  if (tsEnableCrashReport) {
+    if (taosGenCrashJsonMsg(signum, &pMsg, dmGetClusterId(), global.startTime)) {
+      taosPrintLog(flags, level, dflag, "failed to generate crash json msg");
+      goto _return;
+    } else {
+      msgLen = strlen(pMsg);  
+    }
+  }
+  
+_return:
+
+  taosLogCrashInfo("taosd", pMsg, msgLen, signum, sigInfo);
+
+#ifdef _TD_DARWIN_64
+  exit(signum);
+#elif defined(WINDOWS)
+  exit(signum);
+#endif
+}
+
 static void dmSetSignalHandle() {
   taosSetSignal(SIGUSR1, dmSetDebugFlag);
   taosSetSignal(SIGUSR2, dmSetAssert);
   taosSetSignal(SIGTERM, dmStopDnode);
   taosSetSignal(SIGHUP, dmStopDnode);
   taosSetSignal(SIGINT, dmStopDnode);
-  taosSetSignal(SIGABRT, dmStopDnode);
   taosSetSignal(SIGBREAK, dmStopDnode);
 #ifndef WINDOWS
   taosSetSignal(SIGTSTP, dmStopDnode);
   taosSetSignal(SIGQUIT, dmStopDnode);
 #endif
+
+#ifndef WINDOWS
+  taosSetSignal(SIGBUS, dmLogCrash);
+#endif  
+  taosSetSignal(SIGABRT, dmLogCrash);
+  taosSetSignal(SIGFPE, dmLogCrash);
+  taosSetSignal(SIGSEGV, dmLogCrash);
 }
 
 static int32_t dmParseArgs(int32_t argc, char const *argv[]) {
+  global.startTime = taosGetTimestampMs();
+
   int32_t cmdEnvIndex = 0;
   if (argc < 2) return 0;
+  
   global.envCmd = taosMemoryMalloc((argc - 1) * sizeof(char *));
   memset(global.envCmd, 0, (argc - 1) * sizeof(char *));
   for (int32_t i = 1; i < argc; ++i) {

source/dnode/mgmt/mgmt_dnode/inc/dmInt.h

@@ -29,6 +29,7 @@ typedef struct SDnodeMgmt {
   const char         *name;
   TdThread            statusThread;
   TdThread            monitorThread;
+  TdThread            crashReportThread;
   SSingleWorker       mgmtWorker;
   ProcessCreateNodeFp processCreateNodeFp;
   ProcessDropNodeFp   processDropNodeFp;
@@ -55,6 +56,8 @@ int32_t dmStartStatusThread(SDnodeMgmt *pMgmt);
 void    dmStopStatusThread(SDnodeMgmt *pMgmt);
 int32_t dmStartMonitorThread(SDnodeMgmt *pMgmt);
 void    dmStopMonitorThread(SDnodeMgmt *pMgmt);
+int32_t dmStartCrashReportThread(SDnodeMgmt *pMgmt);
+void    dmStopCrashReportThread(SDnodeMgmt *pMgmt);
 int32_t dmStartWorker(SDnodeMgmt *pMgmt);
 void    dmStopWorker(SDnodeMgmt *pMgmt);
 

source/dnode/mgmt/mgmt_dnode/src/dmInt.c

@@ -23,6 +23,9 @@ static int32_t dmStartMgmt(SDnodeMgmt *pMgmt) {
   if (dmStartMonitorThread(pMgmt) != 0) {
     return -1;
   }
+  if (dmStartCrashReportThread(pMgmt) != 0) {
+    return -1;
+  }
   return 0;
 }
 
@@ -30,6 +33,7 @@ static void dmStopMgmt(SDnodeMgmt *pMgmt) {
   pMgmt->pData->stopped = true;
   dmStopMonitorThread(pMgmt);
   dmStopStatusThread(pMgmt);
+  dmStopCrashReportThread(pMgmt);
 }
 
 static int32_t dmOpenMgmt(SMgmtInputOpt *pInput, SMgmtOutputOpt *pOutput) {

source/dnode/mgmt/mgmt_dnode/src/dmWorker.c

@@ -15,6 +15,7 @@
 
 #define _DEFAULT_SOURCE
 #include "dmInt.h"
+#include "thttp.h"
 
 static void *dmStatusThreadFp(void *param) {
   SDnodeMgmt *pMgmt = param;
@@ -63,6 +64,63 @@ static void *dmMonitorThreadFp(void *param) {
   return NULL;
 }
 
+static void *dmCrashReportThreadFp(void *param) {
+  SDnodeMgmt *pMgmt = param;
+  int64_t     lastTime = taosGetTimestampMs();
+  setThreadName("dnode-crashReport");
+  char filepath[PATH_MAX] = {0};
+  snprintf(filepath, sizeof(filepath), "%s%s.taosdCrashLog", tsLogDir, TD_DIRSEP);
+  char *pMsg = NULL;
+  int64_t msgLen = 0;
+  TdFilePtr pFile = NULL;
+  bool truncateFile = false;
+  int32_t sleepTime = 200;
+  int32_t reportPeriodNum = 3600 * 1000 / sleepTime;;
+  int32_t loopTimes = reportPeriodNum;
+  
+  while (1) {
+    if (pMgmt->pData->dropped || pMgmt->pData->stopped) break;
+    if (loopTimes++ < reportPeriodNum) {
+      taosMsleep(sleepTime);
+      continue;
+    }
+
+    taosReadCrashInfo(filepath, &pMsg, &msgLen, &pFile);
+    if (pMsg && msgLen > 0) {
+      if (taosSendHttpReport(tsTelemServer, tsSvrCrashReportUri, tsTelemPort, pMsg, msgLen, HTTP_FLAT) != 0) {
+        dError("failed to send crash report");
+        if (pFile) {
+          taosReleaseCrashLogFile(pFile, false);
+          continue;
+        }
+      } else {
+        dInfo("succeed to send crash report");
+        truncateFile = true;
+      }
+    } else {
+      dDebug("no crash info");
+    }
+
+    taosMemoryFree(pMsg);
+
+    if (pMsg && msgLen > 0) {
+      pMsg = NULL;
+      continue;
+    }
+    
+    if (pFile) {
+      taosReleaseCrashLogFile(pFile, truncateFile);
+      truncateFile = false;
+    }
+    
+    taosMsleep(sleepTime);
+    loopTimes = 0;
+  }
+
+  return NULL;
+}
+
+
 int32_t dmStartStatusThread(SDnodeMgmt *pMgmt) {
   TdThreadAttr thAttr;
   taosThreadAttrInit(&thAttr);
@@ -105,6 +163,36 @@ void dmStopMonitorThread(SDnodeMgmt *pMgmt) {
   }
 }
 
+int32_t dmStartCrashReportThread(SDnodeMgmt *pMgmt) {
+  if (!tsEnableCrashReport) {
+    return 0;
+  }
+
+  TdThreadAttr thAttr;
+  taosThreadAttrInit(&thAttr);
+  taosThreadAttrSetDetachState(&thAttr, PTHREAD_CREATE_JOINABLE);
+  if (taosThreadCreate(&pMgmt->crashReportThread, &thAttr, dmCrashReportThreadFp, pMgmt) != 0) {
+    dError("failed to create crashReport thread since %s", strerror(errno));
+    return -1;
+  }
+
+  taosThreadAttrDestroy(&thAttr);
+  tmsgReportStartup("dnode-crashReport", "initialized");
+  return 0;
+}
+
+void dmStopCrashReportThread(SDnodeMgmt *pMgmt) {
+  if (!tsEnableCrashReport) {
+    return;
+  }
+
+  if (taosCheckPthreadValid(pMgmt->crashReportThread)) {
+    taosThreadJoin(pMgmt->crashReportThread, NULL);
+    taosThreadClear(&pMgmt->crashReportThread);
+  }
+}
+
+
 static void dmProcessMgmtQueue(SQueueInfo *pInfo, SRpcMsg *pMsg) {
   SDnodeMgmt *pMgmt = pInfo->ahandle;
   int32_t     code = -1;

source/dnode/mgmt/mgmt_mnode/src/mmFile.c

@@ -15,171 +15,166 @@
 
 #define _DEFAULT_SOURCE
 #include "mmInt.h"
+#include "tjson.h"
+
+static int32_t mmDecodeOption(SJson *pJson, SMnodeOpt *pOption) {
+  int32_t code = 0;
+
+  tjsonGetInt32ValueFromDouble(pJson, "deployed", pOption->deploy, code);
+  if (code < 0) return -1;
+  tjsonGetInt32ValueFromDouble(pJson, "selfIndex", pOption->selfIndex, code);
+  if (code < 0) return 0;
+
+  SJson *replicas = tjsonGetObjectItem(pJson, "replicas");
+  if (replicas == NULL) return 0;
+  pOption->numOfReplicas = tjsonGetArraySize(replicas);
+
+  for (int32_t i = 0; i < pOption->numOfReplicas; ++i) {
+    SJson *replica = tjsonGetArrayItem(replicas, i);
+    if (replica == NULL) return -1;
+
+    SReplica *pReplica = pOption->replicas + i;
+    tjsonGetInt32ValueFromDouble(replica, "id", pReplica->id, code);
+    if (code < 0) return -1;
+    code = tjsonGetStringValue(replica, "fqdn", pReplica->fqdn);
+    if (code < 0) return -1;
+    tjsonGetUInt16ValueFromDouble(replica, "port", pReplica->port, code);
+    if (code < 0) return -1;
+  }
+
+  return 0;
+}
 
 int32_t mmReadFile(const char *path, SMnodeOpt *pOption) {
-  int32_t   code = TSDB_CODE_INVALID_JSON_FORMAT;
+  int32_t   code = -1;
-  int32_t   len = 0;
-  int32_t   maxLen = 4096;
-  char     *content = taosMemoryCalloc(1, maxLen + 1);
-  cJSON    *root = NULL;
-  char      file[PATH_MAX] = {0};
   TdFilePtr pFile = NULL;
-
+  char     *pData = NULL;
+  SJson    *pJson = NULL;
+  char      file[PATH_MAX] = {0};
   snprintf(file, sizeof(file), "%s%smnode.json", path, TD_DIRSEP);
+
+  if (taosStatFile(file, NULL, NULL) < 0) {
+    dInfo("mnode file:%s not exist", file);
+    return 0;
+  }
+
   pFile = taosOpenFile(file, TD_FILE_READ);
   if (pFile == NULL) {
-    code = 0;
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to open mnode file:%s since %s", file, terrstr());
     goto _OVER;
   }
 
-  len = (int32_t)taosReadFile(pFile, content, maxLen);
+  int64_t size = 0;
-  if (len <= 0) {
+  if (taosFStatFile(pFile, &size, NULL) < 0) {
-    dError("failed to read %s since content is null", file);
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to fstat mnode file:%s since %s", file, terrstr());
     goto _OVER;
   }
 
-  content[len] = 0;
+  pData = taosMemoryMalloc(size + 1);
-  root = cJSON_Parse(content);
+  if (pData == NULL) {
-  if (root == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    dError("failed to read %s since invalid json format", file);
     goto _OVER;
   }
 
-  cJSON *deployed = cJSON_GetObjectItem(root, "deployed");
+  if (taosReadFile(pFile, pData, size) != size) {
-  if (!deployed || deployed->type != cJSON_Number) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to read %s since deployed not found", file);
+    dError("failed to read mnode file:%s since %s", file, terrstr());
     goto _OVER;
   }
-  pOption->deploy = deployed->valueint;
 
-  cJSON *selfIndex = cJSON_GetObjectItem(root, "selfIndex");
+  pData[size] = '\0';
-  if (selfIndex) {
+
-    if (selfIndex->type != cJSON_Number) {
+  pJson = tjsonParse(pData);
-      dError("failed to read %s since selfIndex not found", file);
+  if (pJson == NULL) {
-      goto _OVER;
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
-    }
+    goto _OVER;
-    pOption->selfIndex = selfIndex->valueint;
   }
 
-  cJSON *replicas = cJSON_GetObjectItem(root, "replicas");
+  if (mmDecodeOption(pJson, pOption) < 0) {
-  if (replicas) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
-    if (replicas->type != cJSON_Array) {
+    goto _OVER;
-      dError("failed to read %s since replicas not found", file);
-      goto _OVER;
-    }
-
-    int32_t numOfReplicas = cJSON_GetArraySize(replicas);
-    if (numOfReplicas <= 0) {
-      dError("failed to read %s since numOfReplicas:%d invalid", file, numOfReplicas);
-      goto _OVER;
-    }
-    pOption->numOfReplicas = numOfReplicas;
-
-    for (int32_t i = 0; i < numOfReplicas; ++i) {
-      SReplica *pReplica = pOption->replicas + i;
-
-      cJSON *replica = cJSON_GetArrayItem(replicas, i);
-      if (replica == NULL) break;
-
-      cJSON *id = cJSON_GetObjectItem(replica, "id");
-      if (id) {
-        if (id->type != cJSON_Number) {
-          dError("failed to read %s since id not found", file);
-          goto _OVER;
-        }
-        if (pReplica) {
-          pReplica->id = id->valueint;
-        }
-      }
-
-      cJSON *fqdn = cJSON_GetObjectItem(replica, "fqdn");
-      if (fqdn) {
-        if (fqdn->type != cJSON_String || fqdn->valuestring == NULL) {
-          dError("failed to read %s since fqdn not found", file);
-          goto _OVER;
-        }
-        if (pReplica) {
-          tstrncpy(pReplica->fqdn, fqdn->valuestring, TSDB_FQDN_LEN);
-        }
-      }
-
-      cJSON *port = cJSON_GetObjectItem(replica, "port");
-      if (port) {
-        if (port->type != cJSON_Number) {
-          dError("failed to read %s since port not found", file);
-          goto _OVER;
-        }
-        if (pReplica) {
-          pReplica->port = (uint16_t)port->valueint;
-        }
-      }
-    }
   }
 
   code = 0;
+  dInfo("succceed to read mnode file %s", file);
 
 _OVER:
-  if (content != NULL) taosMemoryFree(content);
+  if (pData != NULL) taosMemoryFree(pData);
-  if (root != NULL) cJSON_Delete(root);
+  if (pJson != NULL) cJSON_Delete(pJson);
   if (pFile != NULL) taosCloseFile(&pFile);
-  if (code == 0) {
-    dDebug("succcessed to read file %s, deployed:%d", file, pOption->deploy);
-  }
 
-  terrno = code;
+  if (code != 0) {
+    dError("failed to read mnode file:%s since %s", file, terrstr());
+  }
   return code;
 }
 
+static int32_t mmEncodeOption(SJson *pJson, const SMnodeOpt *pOption) {
+  if (pOption->deploy && pOption->numOfReplicas > 0) {
+    if (tjsonAddDoubleToObject(pJson, "selfIndex", pOption->selfIndex) < 0) return -1;
+
+    SJson *replicas = tjsonCreateArray();
+    if (replicas == NULL) return -1;
+    if (tjsonAddItemToObject(pJson, "replicas", replicas) < 0) return -1;
+
+    for (int32_t i = 0; i < pOption->numOfReplicas; ++i) {
+      SJson *replica = tjsonCreateObject();
+      if (replica == NULL) return -1;
+
+      const SReplica *pReplica = pOption->replicas + i;
+      if (tjsonAddDoubleToObject(replica, "id", pReplica->id) < 0) return -1;
+      if (tjsonAddStringToObject(replica, "fqdn", pReplica->fqdn) < 0) return -1;
+      if (tjsonAddDoubleToObject(replica, "port", pReplica->port) < 0) return -1;
+      if (tjsonAddItemToArray(replicas, replica) < 0) return -1;
+    }
+  }
+
+  if (tjsonAddDoubleToObject(pJson, "deployed", pOption->deploy) < 0) return -1;
+
+  return 0;
+}
+
 int32_t mmWriteFile(const char *path, const SMnodeOpt *pOption) {
-  char file[PATH_MAX] = {0};
+  int32_t   code = -1;
-  char realfile[PATH_MAX] = {0};
+  char     *buffer = NULL;
+  SJson    *pJson = NULL;
+  TdFilePtr pFile = NULL;
+  char      file[PATH_MAX] = {0};
+  char      realfile[PATH_MAX] = {0};
   snprintf(file, sizeof(file), "%s%smnode.json.bak", path, TD_DIRSEP);
   snprintf(realfile, sizeof(realfile), "%s%smnode.json", path, TD_DIRSEP);
 
-  TdFilePtr pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  terrno = TSDB_CODE_OUT_OF_MEMORY;
-  if (pFile == NULL) {
+  pJson = tjsonCreateObject();
-    terrno = TAOS_SYSTEM_ERROR(errno);
+  if (pJson == NULL) goto _OVER;
-    dError("failed to write %s since %s", file, terrstr());
+  if (mmEncodeOption(pJson, pOption) != 0) goto _OVER;
-    return -1;
+  buffer = tjsonToString(pJson);
-  }
+  if (buffer == NULL) goto _OVER;
+  terrno = 0;
 
-  int32_t len = 0;
+  pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  int32_t maxLen = 4096;
+  if (pFile == NULL) goto _OVER;
-  char   *content = taosMemoryCalloc(1, maxLen + 1);
 
-  len += snprintf(content + len, maxLen - len, "{\n");
+  int32_t len = strlen(buffer);
-  if (pOption->deploy && pOption->numOfReplicas > 0) {
+  if (taosWriteFile(pFile, buffer, len) <= 0) goto _OVER;
-    len += snprintf(content + len, maxLen - len, "  \"selfIndex\": %d,\n", pOption->selfIndex);
+  if (taosFsyncFile(pFile) < 0) goto _OVER;
-    len += snprintf(content + len, maxLen - len, "  \"replicas\": [{\n");
 
-    for (int32_t i = 0; i < pOption->numOfReplicas; ++i) {
-      const SReplica *pReplica = pOption->replicas + i;
-      if (pReplica != NULL && pReplica->id > 0) {
-        len += snprintf(content + len, maxLen - len, "    \"id\": %d,\n", pReplica->id);
-        len += snprintf(content + len, maxLen - len, "    \"fqdn\": \"%s\",\n", pReplica->fqdn);
-        len += snprintf(content + len, maxLen - len, "    \"port\": %u\n", pReplica->port);
-      }
-      if (i < pOption->numOfReplicas - 1) {
-        len += snprintf(content + len, maxLen - len, "  },{\n");
-      } else {
-        len += snprintf(content + len, maxLen - len, "  }],\n");
-      }
-    }
-  }
-  len += snprintf(content + len, maxLen - len, "  \"deployed\": %d\n", pOption->deploy);
-  len += snprintf(content + len, maxLen - len, "}\n");
-
-  taosWriteFile(pFile, content, len);
-  taosFsyncFile(pFile);
   taosCloseFile(&pFile);
-  taosMemoryFree(content);
+  if (taosRenameFile(file, realfile) != 0) goto _OVER;
 
-  if (taosRenameFile(file, realfile) != 0) {
+  code = 0;
-    terrno = TAOS_SYSTEM_ERROR(errno);
+  dInfo("succeed to write mnode file:%s, deloyed:%d", realfile, pOption->deploy);
-    dError("failed to rename %s since %s", file, terrstr());
+
-    return -1;
+_OVER:
+  if (pJson != NULL) tjsonDelete(pJson);
+  if (buffer != NULL) taosMemoryFree(buffer);
+  if (pFile != NULL) taosCloseFile(&pFile);
+
+  if (code != 0) {
+    if (terrno == 0) terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write mnode file:%s since %s, deloyed:%d", realfile, terrstr(), pOption->deploy);
   }
-
+  return code;
-  dDebug("succeed to write %s, deployed:%d", realfile, pOption->deploy);
-  return 0;
 }

source/dnode/mgmt/mgmt_vnode/src/vmFile.c

@@ -14,6 +14,7 @@
  */
 
 #define _DEFAULT_SOURCE
+#include "tjson.h"
 #include "vmInt.h"
 
 #define MAX_CONTENT_LEN 2 * 1024 * 1024
@@ -45,164 +46,171 @@ SVnodeObj **vmGetVnodeListFromHash(SVnodeMgmt *pMgmt, int32_t *numOfVnodes) {
   return pVnodes;
 }
 
-int32_t vmGetVnodeListFromFile(SVnodeMgmt *pMgmt, SWrapperCfg **ppCfgs, int32_t *numOfVnodes) {
+static int32_t vmDecodeVnodeList(SJson *pJson, SVnodeMgmt *pMgmt, SWrapperCfg **ppCfgs, int32_t *numOfVnodes) {
-  int32_t      code = TSDB_CODE_INVALID_JSON_FORMAT;
+  int32_t      code = -1;
-  int32_t      len = 0;
-  int32_t      maxLen = MAX_CONTENT_LEN;
-  char        *content = taosMemoryCalloc(1, maxLen + 1);
-  cJSON       *root = NULL;
-  FILE        *fp = NULL;
-  char         file[PATH_MAX] = {0};
   SWrapperCfg *pCfgs = NULL;
-  TdFilePtr    pFile = NULL;
+  *ppCfgs = NULL;
 
-  snprintf(file, sizeof(file), "%s%svnodes.json", pMgmt->path, TD_DIRSEP);
+  SJson *vnodes = tjsonGetObjectItem(pJson, "vnodes");
-
+  if (vnodes == NULL) return -1;
-  pFile = taosOpenFile(file, TD_FILE_READ);
-  if (pFile == NULL) {
-    dInfo("file %s not exist", file);
-    code = 0;
-    goto _OVER;
-  }
-
-  if (content == NULL) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    return -1;
-  }
-
-  len = (int32_t)taosReadFile(pFile, content, maxLen);
-  if (len <= 0) {
-    dError("failed to read %s since content is null", file);
-    goto _OVER;
-  }
-
-  content[len] = 0;
-  root = cJSON_Parse(content);
-  if (root == NULL) {
-    dError("failed to read %s since invalid json format", file);
-    goto _OVER;
-  }
-
-  cJSON *vnodes = cJSON_GetObjectItem(root, "vnodes");
-  if (!vnodes || vnodes->type != cJSON_Array) {
-    dError("failed to read %s since vnodes not found", file);
-    goto _OVER;
-  }
 
   int32_t vnodesNum = cJSON_GetArraySize(vnodes);
   if (vnodesNum > 0) {
     pCfgs = taosMemoryCalloc(vnodesNum, sizeof(SWrapperCfg));
-    if (pCfgs == NULL) {
+    if (pCfgs == NULL) return -1;
-      dError("failed to read %s since out of memory", file);
+  }
-      code = TSDB_CODE_OUT_OF_MEMORY;
+
-      goto _OVER;
+  for (int32_t i = 0; i < vnodesNum; ++i) {
-    }
+    SJson *vnode = tjsonGetArrayItem(vnodes, i);
-
+    if (vnode == NULL) goto _OVER;
-    for (int32_t i = 0; i < vnodesNum; ++i) {
+
-      cJSON       *vnode = cJSON_GetArrayItem(vnodes, i);
+    SWrapperCfg *pCfg = &pCfgs[i];
-      SWrapperCfg *pCfg = &pCfgs[i];
+    tjsonGetInt32ValueFromDouble(vnode, "vgId", pCfg->vgId, code);
-
+    if (code < 0) goto _OVER;
-      cJSON *vgId = cJSON_GetObjectItem(vnode, "vgId");
+    tjsonGetInt32ValueFromDouble(vnode, "dropped", pCfg->dropped, code);
-      if (!vgId || vgId->type != cJSON_Number) {
+    if (code < 0) goto _OVER;
-        dError("failed to read %s since vgId not found", file);
+    tjsonGetInt32ValueFromDouble(vnode, "vgVersion", pCfg->vgVersion, code);
-        taosMemoryFree(pCfgs);
+    if (code < 0) goto _OVER;
-        goto _OVER;
+
-      }
+    snprintf(pCfg->path, sizeof(pCfg->path), "%s%svnode%d", pMgmt->path, TD_DIRSEP, pCfg->vgId);
-      pCfg->vgId = vgId->valueint;
-      snprintf(pCfg->path, sizeof(pCfg->path), "%s%svnode%d", pMgmt->path, TD_DIRSEP, pCfg->vgId);
-
-      cJSON *dropped = cJSON_GetObjectItem(vnode, "dropped");
-      if (!dropped || dropped->type != cJSON_Number) {
-        dError("failed to read %s since dropped not found", file);
-        taosMemoryFree(pCfgs);
-        goto _OVER;
-      }
-      pCfg->dropped = dropped->valueint;
-
-      cJSON *vgVersion = cJSON_GetObjectItem(vnode, "vgVersion");
-      if (!vgVersion || vgVersion->type != cJSON_Number) {
-        dError("failed to read %s since vgVersion not found", file);
-        taosMemoryFree(pCfgs);
-        goto _OVER;
-      }
-      pCfg->vgVersion = vgVersion->valueint;
-    }
-
-    *ppCfgs = pCfgs;
   }
 
-  *numOfVnodes = vnodesNum;
   code = 0;
-  dInfo("succcessed to read file %s, numOfVnodes:%d", file, vnodesNum);
+  *ppCfgs = pCfgs;
+  *numOfVnodes = vnodesNum;
 
 _OVER:
-  if (content != NULL) taosMemoryFree(content);
+  if (*ppCfgs == NULL) taosMemoryFree(pCfgs);
-  if (root != NULL) cJSON_Delete(root);
-  if (pFile != NULL) taosCloseFile(&pFile);
-
-  terrno = code;
   return code;
 }
 
-int32_t vmWriteVnodeListToFile(SVnodeMgmt *pMgmt) {
+int32_t vmGetVnodeListFromFile(SVnodeMgmt *pMgmt, SWrapperCfg **ppCfgs, int32_t *numOfVnodes) {
-  int32_t code = 0;
+  int32_t      code = -1;
-  char    file[PATH_MAX] = {0};
+  TdFilePtr    pFile = NULL;
-  char    realfile[PATH_MAX] = {0};
+  char        *pData = NULL;
-  snprintf(file, sizeof(file), "%s%svnodes.json.bak", pMgmt->path, TD_DIRSEP);
+  SJson       *pJson = NULL;
-  snprintf(realfile, sizeof(file), "%s%svnodes.json", pMgmt->path, TD_DIRSEP);
+  char         file[PATH_MAX] = {0};
+  SWrapperCfg *pCfgs = NULL;
+  snprintf(file, sizeof(file), "%s%svnodes.json", pMgmt->path, TD_DIRSEP);
 
-  TdFilePtr pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  if (taosStatFile(file, NULL, NULL) < 0) {
+    dInfo("vnode file:%s not exist", file);
+    return 0;
+  }
+
+  pFile = taosOpenFile(file, TD_FILE_READ);
   if (pFile == NULL) {
     terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to write %s since %s", file, terrstr());
+    dError("failed to open vnode file:%s since %s", file, terrstr());
-    return -1;
-  }
-
-  int32_t     numOfVnodes = 0;
-  SVnodeObj **ppVnodes = vmGetVnodeListFromHash(pMgmt, &numOfVnodes);
-  if (ppVnodes == NULL) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    code = -1;
-    dError("failed to write %s while get vnodelist", file);
     goto _OVER;
   }
 
-  int32_t len = 0;
+  int64_t size = 0;
-  int32_t maxLen = MAX_CONTENT_LEN;
+  if (taosFStatFile(pFile, &size, NULL) < 0) {
-  char   *content = taosMemoryCalloc(1, maxLen + 1);
+    terrno = TAOS_SYSTEM_ERROR(errno);
-  if (content == NULL) {
+    dError("failed to fstat mnode file:%s since %s", file, terrstr());
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    code = -1;
-    dError("failed to write %s while malloc content", file);
     goto _OVER;
   }
 
-  len += snprintf(content + len, maxLen - len, "{\n");
+  pData = taosMemoryMalloc(size + 1);
-  len += snprintf(content + len, maxLen - len, "  \"vnodes\": [\n");
+  if (pData == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    goto _OVER;
+  }
+
+  if (taosReadFile(pFile, pData, size) != size) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to read vnode file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  pData[size] = '\0';
+
+  pJson = tjsonParse(pData);
+  if (pJson == NULL) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
+    goto _OVER;
+  }
+
+  if (vmDecodeVnodeList(pJson, pMgmt, ppCfgs, numOfVnodes) < 0) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
+    goto _OVER;
+  }
+
+  code = 0;
+  dInfo("succceed to read vnode file %s", file);
+
+_OVER:
+  if (pData != NULL) taosMemoryFree(pData);
+  if (pJson != NULL) cJSON_Delete(pJson);
+  if (pFile != NULL) taosCloseFile(&pFile);
+
+  if (code != 0) {
+    dError("failed to read vnode file:%s since %s", file, terrstr());
+  }
+  return code;
+}
+
+static int32_t vmEncodeVnodeList(SJson *pJson, SVnodeObj **ppVnodes, int32_t numOfVnodes) {
+  SJson *vnodes = tjsonCreateArray();
+  if (vnodes == NULL) return -1;
+  if (tjsonAddItemToObject(pJson, "vnodes", vnodes) < 0) return -1;
+
   for (int32_t i = 0; i < numOfVnodes; ++i) {
     SVnodeObj *pVnode = ppVnodes[i];
     if (pVnode == NULL) continue;
 
-    len += snprintf(content + len, maxLen - len, "    {\n");
+    SJson *vnode = tjsonCreateObject();
-    len += snprintf(content + len, maxLen - len, "      \"vgId\": %d,\n", pVnode->vgId);
+    if (vnode == NULL) return -1;
-    len += snprintf(content + len, maxLen - len, "      \"dropped\": %d,\n", pVnode->dropped);
+    if (tjsonAddDoubleToObject(vnode, "vgId", pVnode->vgId) < 0) return -1;
-    len += snprintf(content + len, maxLen - len, "      \"vgVersion\": %d\n", pVnode->vgVersion);
+    if (tjsonAddDoubleToObject(vnode, "dropped", pVnode->dropped) < 0) return -1;
-    if (i < numOfVnodes - 1) {
+    if (tjsonAddDoubleToObject(vnode, "vgVersion", pVnode->vgVersion) < 0) return -1;
-      len += snprintf(content + len, maxLen - len, "    },\n");
+    if (tjsonAddItemToArray(vnodes, vnode) < 0) return -1;
-    } else {
-      len += snprintf(content + len, maxLen - len, "    }\n");
-    }
   }
-  len += snprintf(content + len, maxLen - len, "  ]\n");
+
-  len += snprintf(content + len, maxLen - len, "}\n");
+  return 0;
+}
+
+int32_t vmWriteVnodeListToFile(SVnodeMgmt *pMgmt) {
+  int32_t     code = -1;
+  char       *buffer = NULL;
+  SJson      *pJson = NULL;
+  TdFilePtr   pFile = NULL;
+  SVnodeObj **ppVnodes = NULL;
+  char        file[PATH_MAX] = {0};
+  char        realfile[PATH_MAX] = {0};
+  snprintf(file, sizeof(file), "%s%svnodes.json.bak", pMgmt->path, TD_DIRSEP);
+  snprintf(realfile, sizeof(realfile), "%s%svnodes.json", pMgmt->path, TD_DIRSEP);
+
+  int32_t numOfVnodes = 0;
+  ppVnodes = vmGetVnodeListFromHash(pMgmt, &numOfVnodes);
+  if (ppVnodes == NULL) goto _OVER;
+
+  terrno = TSDB_CODE_OUT_OF_MEMORY;
+  pJson = tjsonCreateObject();
+  if (pJson == NULL) goto _OVER;
+  if (vmEncodeVnodeList(pJson, ppVnodes, numOfVnodes) != 0) goto _OVER;
+  buffer = tjsonToString(pJson);
+  if (buffer == NULL) goto _OVER;
   terrno = 0;
 
-_OVER:
+  pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  taosWriteFile(pFile, content, len);
+  if (pFile == NULL) goto _OVER;
-  taosFsyncFile(pFile);
-  taosCloseFile(&pFile);
-  taosMemoryFree(content);
 
+  int32_t len = strlen(buffer);
+  if (taosWriteFile(pFile, buffer, len) <= 0) goto _OVER;
+  if (taosFsyncFile(pFile) < 0) goto _OVER;
+
+  taosCloseFile(&pFile);
+  if (taosRenameFile(file, realfile) != 0) goto _OVER;
+
+  code = 0;
+  dInfo("succeed to write vnodes file:%s, vnodes:%d", realfile, numOfVnodes);
+
+_OVER:
+  if (pJson != NULL) tjsonDelete(pJson);
+  if (buffer != NULL) taosMemoryFree(buffer);
+  if (pFile != NULL) taosCloseFile(&pFile);
   if (ppVnodes != NULL) {
     for (int32_t i = 0; i < numOfVnodes; ++i) {
       SVnodeObj *pVnode = ppVnodes[i];
@@ -213,14 +221,9 @@ _OVER:
     taosMemoryFree(ppVnodes);
   }
 
-  if (code != 0) return -1;
-
-  dInfo("succeed to write %s, numOfVnodes:%d", realfile, numOfVnodes);
-  code = taosRenameFile(file, realfile);
-  
   if (code != 0) {
-    dError("failed to rename %s to %s", file, realfile);
+    if (terrno == 0) terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write vnodes file:%s since %s, vnodes:%d", realfile, terrstr(), numOfVnodes);
   }
-
   return code;
 }

source/dnode/mgmt/mgmt_vnode/src/vmHandle.c

@@ -137,7 +137,7 @@ static void vmGenerateVnodeCfg(SCreateVnodeReq *pCreate, SVnodeCfg *pCfg) {
     pNode->nodeId = pCreate->replicas[i].id;
     pNode->nodePort = pCreate->replicas[i].port;
     tstrncpy(pNode->nodeFqdn, pCreate->replicas[i].fqdn, TSDB_FQDN_LEN);
-    (void)tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
+    tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
   }
 }
 

source/dnode/mgmt/mgmt_vnode/src/vmInt.c

@@ -79,6 +79,8 @@ int32_t vmOpenVnode(SVnodeMgmt *pMgmt, SWrapperCfg *pCfg, SVnode *pImpl) {
 void vmCloseVnode(SVnodeMgmt *pMgmt, SVnodeObj *pVnode) {
   char path[TSDB_FILENAME_LEN] = {0};
 
+  vnodeProposeCommitOnNeed(pVnode->pImpl);
+
   taosThreadRwlockWrlock(&pMgmt->lock);
   taosHashRemove(pMgmt->hash, &pVnode->vgId, sizeof(int32_t));
   taosThreadRwlockUnlock(&pMgmt->lock);
@@ -341,13 +343,12 @@ static void vmCheckSyncTimeout(SVnodeMgmt *pMgmt) {
   int32_t     numOfVnodes = 0;
   SVnodeObj **ppVnodes = vmGetVnodeListFromHash(pMgmt, &numOfVnodes);
 
-  for (int32_t i = 0; i < numOfVnodes; ++i) {
-    SVnodeObj *pVnode = ppVnodes[i];
-    vnodeSyncCheckTimeout(pVnode->pImpl);
-    vmReleaseVnode(pMgmt, pVnode);
-  }
-
   if (ppVnodes != NULL) {
+    for (int32_t i = 0; i < numOfVnodes; ++i) {
+      SVnodeObj *pVnode = ppVnodes[i];
+      vnodeSyncCheckTimeout(pVnode->pImpl);
+      vmReleaseVnode(pMgmt, pVnode);
+    }
     taosMemoryFree(ppVnodes);
   }
 }

source/dnode/mgmt/node_mgmt/inc/dmMgmt.h

@@ -85,6 +85,7 @@ typedef struct SDnode {
 // dmEnv.c
 SDnode *dmInstance();
 void    dmReportStartup(const char *pName, const char *pDesc);
+int64_t dmGetClusterId();
 
 // dmMgmt.c
 int32_t       dmInitDnode(SDnode *pDnode);

source/dnode/mgmt/node_mgmt/src/dmEnv.c

@@ -16,9 +16,9 @@
 #define _DEFAULT_SOURCE
 #include "dmMgmt.h"
 
-static SDnode global = {0};
+static SDnode globalDnode = {0};
 
-SDnode *dmInstance() { return &global; }
+SDnode *dmInstance() { return &globalDnode; }
 
 static int32_t dmCheckRepeatInit(SDnode *pDnode) {
   if (atomic_val_compare_exchange_8(&pDnode->once, DND_ENV_INIT, DND_ENV_READY) != DND_ENV_INIT) {
@@ -268,3 +268,8 @@ void dmReportStartup(const char *pName, const char *pDesc) {
   tstrncpy(pStartup->desc, pDesc, TSDB_STEP_DESC_LEN);
   dDebug("step:%s, %s", pStartup->name, pStartup->desc);
 }
+
+int64_t dmGetClusterId() {
+  return globalDnode.data.clusterId;
+}
+

source/dnode/mgmt/node_mgmt/src/dmNodes.c

@@ -111,6 +111,7 @@ static int32_t dmStartNodes(SDnode *pDnode) {
 
   dInfo("TDengine initialized successfully");
   dmReportStartup("TDengine", "initialized successfully");
+  
   return 0;
 }
 

source/dnode/mgmt/node_util/inc/dmUtil.h

@@ -100,6 +100,7 @@ typedef struct {
   bool           stopped;
   SEpSet         mnodeEps;
   SArray        *dnodeEps;
+  SArray        *oldDnodeEps;
   SHashObj      *dnodeHash;
   TdThreadRwlock lock;
   SMsgCb         msgCb;
@@ -167,7 +168,7 @@ void    dmUpdateEps(SDnodeData *pData, SArray *pDnodeEps);
 void    dmGetMnodeEpSet(SDnodeData *pData, SEpSet *pEpSet);
 void    dmGetMnodeEpSetForRedirect(SDnodeData *pData, SRpcMsg *pMsg, SEpSet *pEpSet);
 void    dmSetMnodeEpSet(SDnodeData *pData, SEpSet *pEpSet);
-int32_t dmUpdateDnodeInfo(void *pData, int32_t *dnodeId, int64_t *clusterId, char *fqdn, uint16_t *port);
+void    dmUpdateDnodeInfo(void *pData, int32_t *dnodeId, int64_t *clusterId, char *fqdn, uint16_t *port);
 
 #ifdef __cplusplus
 }

source/dnode/mgmt/node_util/src/dmEps.c

@@ -15,6 +15,7 @@
 
 #define _DEFAULT_SOURCE
 #include "dmUtil.h"
+#include "tjson.h"
 #include "tmisce.h"
 
 static void dmPrintEps(SDnodeData *pData);
@@ -40,14 +41,49 @@ static void dmGetDnodeEp(SDnodeData *pData, int32_t dnodeId, char *pEp, char *pF
   taosThreadRwlockUnlock(&pData->lock);
 }
 
+static int32_t dmDecodeEps(SJson *pJson, SDnodeData *pData) {
+  int32_t code = 0;
+
+  tjsonGetInt32ValueFromDouble(pJson, "dnodeId", pData->dnodeId, code);
+  if (code < 0) return -1;
+  tjsonGetNumberValue(pJson, "dnodeVer", pData->dnodeVer, code);
+  if (code < 0) return -1;
+  tjsonGetNumberValue(pJson, "clusterId", pData->clusterId, code);
+  if (code < 0) return -1;
+  tjsonGetInt32ValueFromDouble(pJson, "dropped", pData->dropped, code);
+  if (code < 0) return -1;
+
+  SJson *dnodes = tjsonGetObjectItem(pJson, "dnodes");
+  if (dnodes == NULL) return 0;
+  int32_t numOfDnodes = tjsonGetArraySize(dnodes);
+
+  for (int32_t i = 0; i < numOfDnodes; ++i) {
+    SJson *dnode = tjsonGetArrayItem(dnodes, i);
+    if (dnode == NULL) return -1;
+
+    SDnodeEp dnodeEp = {0};
+    tjsonGetInt32ValueFromDouble(dnode, "id", dnodeEp.id, code);
+    if (code < 0) return -1;
+    code = tjsonGetStringValue(dnode, "fqdn", dnodeEp.ep.fqdn);
+    if (code < 0) return -1;
+    tjsonGetUInt16ValueFromDouble(dnode, "port", dnodeEp.ep.port, code);
+    if (code < 0) return -1;
+    tjsonGetInt8ValueFromDouble(dnode, "isMnode", dnodeEp.isMnode, code);
+    if (code < 0) return -1;
+
+    if (taosArrayPush(pData->dnodeEps, &dnodeEp) == NULL) return -1;
+  }
+
+  return 0;
+}
+
 int32_t dmReadEps(SDnodeData *pData) {
-  int32_t   code = TSDB_CODE_INVALID_JSON_FORMAT;
+  int32_t   code = -1;
-  int32_t   len = 0;
-  int32_t   maxLen = 256 * 1024;
-  char     *content = taosMemoryCalloc(1, maxLen + 1);
-  cJSON    *root = NULL;
-  char      file[PATH_MAX] = {0};
   TdFilePtr pFile = NULL;
+  char     *content = NULL;
+  SJson    *pJson = NULL;
+  char      file[PATH_MAX] = {0};
+  snprintf(file, sizeof(file), "%s%sdnode%sdnode.json", tsDataDir, TD_DIRSEP, TD_DIRSEP);
 
   pData->dnodeEps = taosArrayInit(1, sizeof(SDnodeEp));
   if (pData->dnodeEps == NULL) {
@@ -55,113 +91,63 @@ int32_t dmReadEps(SDnodeData *pData) {
     goto _OVER;
   }
 
-  snprintf(file, sizeof(file), "%s%sdnode%sdnode.json", tsDataDir, TD_DIRSEP, TD_DIRSEP);
+  if (taosStatFile(file, NULL, NULL) < 0) {
-  pFile = taosOpenFile(file, TD_FILE_READ);
+    dInfo("dnode file:%s not exist", file);
-  if (pFile == NULL) {
     code = 0;
     goto _OVER;
   }
 
-  len = (int32_t)taosReadFile(pFile, content, maxLen);
+  pFile = taosOpenFile(file, TD_FILE_READ);
-  if (len <= 0) {
+  if (pFile == NULL) {
-    dError("failed to read %s since content is null", file);
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to open dnode file:%s since %s", file, terrstr());
     goto _OVER;
   }
 
-  content[len] = 0;
+  int64_t size = 0;
-  root = cJSON_Parse(content);
+  if (taosFStatFile(pFile, &size, NULL) < 0) {
-  if (root == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to read %s since invalid json format", file);
+    dError("failed to fstat dnode file:%s since %s", file, terrstr());
     goto _OVER;
   }
 
-  cJSON *dnodeId = cJSON_GetObjectItem(root, "dnodeId");
+  content = taosMemoryMalloc(size + 1);
-  if (!dnodeId || dnodeId->type != cJSON_Number) {
+  if (content == NULL) {
-    dError("failed to read %s since dnodeId not found", file);
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    goto _OVER;
-  }
-  pData->dnodeId = dnodeId->valueint;
-
-  cJSON *dnodeVer = cJSON_GetObjectItem(root, "dnodeVer");
-  if (!dnodeVer || dnodeVer->type != cJSON_String) {
-    dError("failed to read %s since dnodeVer not found", file);
-    goto _OVER;
-  }
-  pData->dnodeVer = atoll(dnodeVer->valuestring);
-
-  cJSON *clusterId = cJSON_GetObjectItem(root, "clusterId");
-  if (!clusterId || clusterId->type != cJSON_String) {
-    dError("failed to read %s since clusterId not found", file);
-    goto _OVER;
-  }
-  pData->clusterId = atoll(clusterId->valuestring);
-
-  cJSON *dropped = cJSON_GetObjectItem(root, "dropped");
-  if (!dropped || dropped->type != cJSON_Number) {
-    dError("failed to read %s since dropped not found", file);
-    goto _OVER;
-  }
-  pData->dropped = dropped->valueint;
-
-  cJSON *dnodes = cJSON_GetObjectItem(root, "dnodes");
-  if (!dnodes || dnodes->type != cJSON_Array) {
-    dError("failed to read %s since dnodes not found", file);
     goto _OVER;
   }
 
-  int32_t numOfDnodes = cJSON_GetArraySize(dnodes);
+  if (taosReadFile(pFile, content, size) != size) {
-  if (numOfDnodes <= 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to read %s since numOfDnodes:%d invalid", file, numOfDnodes);
+    dError("failed to read dnode file:%s since %s", file, terrstr());
     goto _OVER;
   }
 
-  for (int32_t i = 0; i < numOfDnodes; ++i) {
+  content[size] = '\0';
-    cJSON *node = cJSON_GetArrayItem(dnodes, i);
-    if (node == NULL) break;
 
-    SDnodeEp dnodeEp = {0};
+  pJson = tjsonParse(content);
+  if (pJson == NULL) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
+    goto _OVER;
+  }
 
-    cJSON *did = cJSON_GetObjectItem(node, "id");
+  if (dmDecodeEps(pJson, pData) < 0) {
-    if (!did || did->type != cJSON_Number) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
-      dError("failed to read %s since dnodeId not found", file);
+    goto _OVER;
-      goto _OVER;
-    }
-
-    dnodeEp.id = did->valueint;
-
-    cJSON *dnodeFqdn = cJSON_GetObjectItem(node, "fqdn");
-    if (!dnodeFqdn || dnodeFqdn->type != cJSON_String || dnodeFqdn->valuestring == NULL) {
-      dError("failed to read %s since dnodeFqdn not found", file);
-      goto _OVER;
-    }
-    tstrncpy(dnodeEp.ep.fqdn, dnodeFqdn->valuestring, TSDB_FQDN_LEN);
-
-    cJSON *dnodePort = cJSON_GetObjectItem(node, "port");
-    if (!dnodePort || dnodePort->type != cJSON_Number) {
-      dError("failed to read %s since dnodePort not found", file);
-      goto _OVER;
-    }
-
-    dnodeEp.ep.port = dnodePort->valueint;
-
-    cJSON *isMnode = cJSON_GetObjectItem(node, "isMnode");
-    if (!isMnode || isMnode->type != cJSON_Number) {
-      dError("failed to read %s since isMnode not found", file);
-      goto _OVER;
-    }
-    dnodeEp.isMnode = isMnode->valueint;
-
-    taosArrayPush(pData->dnodeEps, &dnodeEp);
   }
 
   code = 0;
-  dDebug("succcessed to read file %s", file);
+  dInfo("succceed to read mnode file %s", file);
 
 _OVER:
   if (content != NULL) taosMemoryFree(content);
-  if (root != NULL) cJSON_Delete(root);
+  if (pJson != NULL) cJSON_Delete(pJson);
   if (pFile != NULL) taosCloseFile(&pFile);
 
+  if (code != 0) {
+    dError("failed to read dnode file:%s since %s", file, terrstr());
+  }
+
   if (taosArrayGetSize(pData->dnodeEps) == 0) {
     SDnodeEp dnodeEp = {0};
     dnodeEp.isMnode = 1;
@@ -177,85 +163,76 @@ _OVER:
     return -1;
   }
 
-  terrno = code;
   return code;
 }
 
-int32_t dmWriteEps(SDnodeData *pData) {
+static int32_t dmEncodeEps(SJson *pJson, SDnodeData *pData) {
-  int32_t   code = -1;
+  if (tjsonAddDoubleToObject(pJson, "dnodeId", pData->dnodeId) < 0) return -1;
-  char     *content = NULL;
+  if (tjsonAddIntegerToObject(pJson, "dnodeVer", pData->dnodeVer) < 0) return -1;
-  TdFilePtr pFile = NULL;
+  if (tjsonAddIntegerToObject(pJson, "clusterId", pData->clusterId) < 0) return -1;
+  if (tjsonAddDoubleToObject(pJson, "dropped", pData->dropped) < 0) return -1;
 
-  char file[PATH_MAX] = {0};
+  SJson *dnodes = tjsonCreateArray();
-  char realfile[PATH_MAX] = {0};
+  if (dnodes == NULL) return -1;
-  snprintf(file, sizeof(file), "%s%sdnode%sdnode.json.bak", tsDataDir, TD_DIRSEP, TD_DIRSEP);
+  if (tjsonAddItemToObject(pJson, "dnodes", dnodes) < 0) return -1;
-  snprintf(realfile, sizeof(realfile), "%s%sdnode%sdnode.json", tsDataDir, TD_DIRSEP, TD_DIRSEP);
-
-  pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  if (pFile == NULL) {
-    dError("failed to open %s since %s", file, strerror(errno));
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    goto _OVER;
-  }
-
-  int32_t len = 0;
-  int32_t maxLen = 256 * 1024;
-  content = taosMemoryCalloc(1, maxLen + 1);
-
-  len += snprintf(content + len, maxLen - len, "{\n");
-  len += snprintf(content + len, maxLen - len, "  \"dnodeId\": %d,\n", pData->dnodeId);
-  len += snprintf(content + len, maxLen - len, "  \"dnodeVer\": \"%" PRId64 "\",\n", pData->dnodeVer);
-  len += snprintf(content + len, maxLen - len, "  \"clusterId\": \"%" PRId64 "\",\n", pData->clusterId);
-  len += snprintf(content + len, maxLen - len, "  \"dropped\": %d,\n", pData->dropped);
-  len += snprintf(content + len, maxLen - len, "  \"dnodes\": [{\n");
 
   int32_t numOfEps = (int32_t)taosArrayGetSize(pData->dnodeEps);
   for (int32_t i = 0; i < numOfEps; ++i) {
     SDnodeEp *pDnodeEp = taosArrayGet(pData->dnodeEps, i);
-    len += snprintf(content + len, maxLen - len, "    \"id\": %d,\n", pDnodeEp->id);
+    SJson    *dnode = tjsonCreateObject();
-    len += snprintf(content + len, maxLen - len, "    \"fqdn\": \"%s\",\n", pDnodeEp->ep.fqdn);
+    if (dnode == NULL) return -1;
-    len += snprintf(content + len, maxLen - len, "    \"port\": %u,\n", pDnodeEp->ep.port);
-    len += snprintf(content + len, maxLen - len, "    \"isMnode\": %d\n", pDnodeEp->isMnode);
-    if (i < numOfEps - 1) {
-      len += snprintf(content + len, maxLen - len, "  },{\n");
-    } else {
-      len += snprintf(content + len, maxLen - len, "  }]\n");
-    }
-  }
-  len += snprintf(content + len, maxLen - len, "}\n");
 
-  if (taosWriteFile(pFile, content, len) != len) {
+    if (tjsonAddDoubleToObject(dnode, "id", pDnodeEp->id) < 0) return -1;
-    dError("failed to write %s since %s", file, strerror(errno));
+    if (tjsonAddStringToObject(dnode, "fqdn", pDnodeEp->ep.fqdn) < 0) return -1;
-    terrno = TAOS_SYSTEM_ERROR(errno);
+    if (tjsonAddDoubleToObject(dnode, "port", pDnodeEp->ep.port) < 0) return -1;
-    goto _OVER;
+    if (tjsonAddDoubleToObject(dnode, "isMnode", pDnodeEp->isMnode) < 0) return -1;
+    if (tjsonAddItemToArray(dnodes, dnode) < 0) return -1;
   }
 
-  if (taosFsyncFile(pFile) < 0) {
+  return 0;
-    dError("failed to fsync %s since %s", file, strerror(errno));
+}
-    terrno = TAOS_SYSTEM_ERROR(errno);
+
-    goto _OVER;
+int32_t dmWriteEps(SDnodeData *pData) {
-  }
+  int32_t   code = -1;
+  char     *buffer = NULL;
+  SJson    *pJson = NULL;
+  TdFilePtr pFile = NULL;
+  char      file[PATH_MAX] = {0};
+  char      realfile[PATH_MAX] = {0};
+  snprintf(file, sizeof(file), "%s%sdnode%sdnode.json.bak", tsDataDir, TD_DIRSEP, TD_DIRSEP);
+  snprintf(realfile, sizeof(realfile), "%s%sdnode%sdnode.json", tsDataDir, TD_DIRSEP, TD_DIRSEP);
+
+  terrno = TSDB_CODE_OUT_OF_MEMORY;
+  pJson = tjsonCreateObject();
+  if (pJson == NULL) goto _OVER;
+  if (dmEncodeEps(pJson, pData) != 0) goto _OVER;
+  buffer = tjsonToString(pJson);
+  if (buffer == NULL) goto _OVER;
+  terrno = 0;
+
+  pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  if (pFile == NULL) goto _OVER;
+
+  int32_t len = strlen(buffer);
+  if (taosWriteFile(pFile, buffer, len) <= 0) goto _OVER;
+  if (taosFsyncFile(pFile) < 0) goto _OVER;
 
   taosCloseFile(&pFile);
-  taosMemoryFreeClear(content);
+  if (taosRenameFile(file, realfile) != 0) goto _OVER;
-
-  if (taosRenameFile(file, realfile) != 0) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to rename %s since %s", file, terrstr());
-    goto _OVER;
-  }
 
   code = 0;
   pData->updateTime = taosGetTimestampMs();
-  dInfo("succeed to write %s, dnodeVer:%" PRId64, realfile, pData->dnodeVer);
+  dInfo("succeed to write dnode file:%s, dnodeVer:%" PRId64, realfile, pData->dnodeVer);
 
 _OVER:
-  if (content != NULL) taosMemoryFreeClear(content);
+  if (pJson != NULL) tjsonDelete(pJson);
+  if (buffer != NULL) taosMemoryFree(buffer);
   if (pFile != NULL) taosCloseFile(&pFile);
-  if (code != 0) {
-    dError("failed to write file %s since %s", realfile, terrstr());
-  }
 
+  if (code != 0) {
+    if (terrno == 0) terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write dnode file:%s since %s, dnodeVer:%" PRId64, realfile, terrstr(), pData->dnodeVer);
+  }
   return code;
 }
 
@@ -355,40 +332,48 @@ void dmSetMnodeEpSet(SDnodeData *pData, SEpSet *pEpSet) {
   }
 }
 
-int32_t dmUpdateDnodeInfo(void *data, int32_t *dnodeId, int64_t *clusterId, char *fqdn, uint16_t *port) {
+void dmUpdateDnodeInfo(void *data, int32_t *did, int64_t *clusterId, char *fqdn, uint16_t *port) {
   SDnodeData *pData = data;
-  int32_t     ret = -1;
+  int32_t     dnodeId = -1;
+  if (did != NULL) dnodeId = *did;
+
   taosThreadRwlockRdlock(&pData->lock);
-  if (*dnodeId <= 0) {
+
-    for (int32_t i = 0; i < (int32_t)taosArrayGetSize(pData->dnodeEps); ++i) {
+  if (pData->oldDnodeEps != NULL) {
-      SDnodeEp *pDnodeEp = taosArrayGet(pData->dnodeEps, i);
+    int32_t size = (int32_t)taosArrayGetSize(pData->oldDnodeEps);
+    for (int32_t i = 0; i < size; ++i) {
+      SDnodeEp *pDnodeEp = taosArrayGet(pData->oldDnodeEps, i);
       if (strcmp(pDnodeEp->ep.fqdn, fqdn) == 0 && pDnodeEp->ep.port == *port) {
-        dInfo("dnode:%s:%u, update dnodeId from %d to %d", fqdn, *port, *dnodeId, pDnodeEp->id);
+        dInfo("dnode:%d, update ep:%s:%u to %s:%u", dnodeId, fqdn, *port, pDnodeEp->ep.fqdn, pDnodeEp->ep.port);
-        *dnodeId = pDnodeEp->id;
-        *clusterId = pData->clusterId;
-        ret = 0;
-      }
-    }
-    if (ret != 0) {
-      dInfo("dnode:%s:%u, failed to update dnodeId:%d", fqdn, *port, *dnodeId);
-    }
-  } else {
-    SDnodeEp *pDnodeEp = taosHashGet(pData->dnodeHash, dnodeId, sizeof(int32_t));
-    if (pDnodeEp) {
-      if (strcmp(pDnodeEp->ep.fqdn, fqdn) != 0) {
-        dInfo("dnode:%d, update port from %s to %s", *dnodeId, fqdn, pDnodeEp->ep.fqdn);
         tstrncpy(fqdn, pDnodeEp->ep.fqdn, TSDB_FQDN_LEN);
-      }
-      if (pDnodeEp->ep.port != *port) {
-        dInfo("dnode:%d, update port from %u to %u", *dnodeId, *port, pDnodeEp->ep.port);
         *port = pDnodeEp->ep.port;
       }
-      *clusterId = pData->clusterId;
-      ret = 0;
-    } else {
-      dInfo("dnode:%d, failed to update dnode info", *dnodeId);
     }
   }
+
+  if (did != NULL && dnodeId <= 0) {
+    int32_t size = (int32_t)taosArrayGetSize(pData->dnodeEps);
+    for (int32_t i = 0; i < size; ++i) {
+      SDnodeEp *pDnodeEp = taosArrayGet(pData->dnodeEps, i);
+      if (strcmp(pDnodeEp->ep.fqdn, fqdn) == 0 && pDnodeEp->ep.port == *port) {
+        dInfo("dnode:%s:%u, update dnodeId to dnode:%d", fqdn, *port, pDnodeEp->id);
+        *did = pDnodeEp->id;
+        if (clusterId != NULL) *clusterId = pData->clusterId;
+      }
+    }
+  }
+
+  if (dnodeId > 0) {
+    SDnodeEp *pDnodeEp = taosHashGet(pData->dnodeHash, &dnodeId, sizeof(int32_t));
+    if (pDnodeEp) {
+      if (strcmp(pDnodeEp->ep.fqdn, fqdn) != 0 || pDnodeEp->ep.port != *port) {
+        dInfo("dnode:%d, update ep:%s:%u to %s:%u", dnodeId, fqdn, *port, pDnodeEp->ep.fqdn, pDnodeEp->ep.port);
+        tstrncpy(fqdn, pDnodeEp->ep.fqdn, TSDB_FQDN_LEN);
+        *port = pDnodeEp->ep.port;
+      }
+      if (clusterId != NULL) *clusterId = pData->clusterId;
+    }
+  }
+
   taosThreadRwlockUnlock(&pData->lock);
-  return ret;
 }

source/dnode/mgmt/node_util/src/dmFile.c

@@ -15,104 +15,133 @@
 
 #define _DEFAULT_SOURCE
 #include "dmUtil.h"
+#include "tjson.h"
 
 #define MAXLEN 1024
 
-int32_t dmReadFile(const char *path, const char *name, bool *pDeployed) {
+static int32_t dmDecodeFile(SJson *pJson, bool *deployed) {
-  int32_t   code = TSDB_CODE_INVALID_JSON_FORMAT;
+  int32_t code = 0;
-  int64_t   len = 0;
+  int32_t value = 0;
-  char      content[MAXLEN + 1] = {0};
-  cJSON    *root = NULL;
-  char      file[PATH_MAX] = {0};
-  TdFilePtr pFile = NULL;
 
+  tjsonGetInt32ValueFromDouble(pJson, "deployed", value, code);
+  if (code < 0) return -1;
+
+  *deployed = (value != 0);
+  return code;
+}
+
+int32_t dmReadFile(const char *path, const char *name, bool *pDeployed) {
+  int32_t   code = -1;
+  TdFilePtr pFile = NULL;
+  char     *content = NULL;
+  SJson    *pJson = NULL;
+  char      file[PATH_MAX] = {0};
   snprintf(file, sizeof(file), "%s%s%s.json", path, TD_DIRSEP, name);
-  pFile = taosOpenFile(file, TD_FILE_READ);
+
-  if (pFile == NULL) {
+  if (taosStatFile(file, NULL, NULL) < 0) {
+    dInfo("file:%s not exist", file);
     code = 0;
     goto _OVER;
   }
 
-  len = taosReadFile(pFile, content, MAXLEN);
+  pFile = taosOpenFile(file, TD_FILE_READ);
-  if (len <= 0) {
+  if (pFile == NULL) {
-    dError("failed to read %s since content is null", file);
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to open file:%s since %s", file, terrstr());
     goto _OVER;
   }
 
-  root = cJSON_Parse(content);
+  int64_t size = 0;
-  if (root == NULL) {
+  if (taosFStatFile(pFile, &size, NULL) < 0) {
-    dError("failed to read %s since invalid json format", file);
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to fstat file:%s since %s", file, terrstr());
     goto _OVER;
   }
 
-  cJSON *deployed = cJSON_GetObjectItem(root, "deployed");
+  content = taosMemoryMalloc(size + 1);
-  if (!deployed || deployed->type != cJSON_Number) {
+  if (content == NULL) {
-    dError("failed to read %s since deployed not found", file);
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    goto _OVER;
+  }
+
+  if (taosReadFile(pFile, content, size) != size) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to read file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  content[size] = '\0';
+
+  pJson = tjsonParse(content);
+  if (pJson == NULL) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
+    goto _OVER;
+  }
+
+  if (dmDecodeFile(pJson, pDeployed) < 0) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
     goto _OVER;
   }
-  *pDeployed = deployed->valueint != 0;
 
-  dDebug("succcessed to read file %s, deployed:%d", file, *pDeployed);
   code = 0;
+  dInfo("succceed to read mnode file %s", file);
 
 _OVER:
-  if (root != NULL) cJSON_Delete(root);
+  if (content != NULL) taosMemoryFree(content);
+  if (pJson != NULL) cJSON_Delete(pJson);
   if (pFile != NULL) taosCloseFile(&pFile);
 
-  terrno = code;
+  if (code != 0) {
+    dError("failed to read dnode file:%s since %s", file, terrstr());
+  }
   return code;
 }
 
+static int32_t dmEncodeFile(SJson *pJson, bool deployed) {
+  if (tjsonAddDoubleToObject(pJson, "deployed", deployed) < 0) return -1;
+  return 0;
+}
+
 int32_t dmWriteFile(const char *path, const char *name, bool deployed) {
   int32_t   code = -1;
-  int32_t   len = 0;
+  char     *buffer = NULL;
-  char      content[MAXLEN + 1] = {0};
+  SJson    *pJson = NULL;
+  TdFilePtr pFile = NULL;
   char      file[PATH_MAX] = {0};
   char      realfile[PATH_MAX] = {0};
-  TdFilePtr pFile = NULL;
-
   snprintf(file, sizeof(file), "%s%s%s.json", path, TD_DIRSEP, name);
   snprintf(realfile, sizeof(realfile), "%s%s%s.json", path, TD_DIRSEP, name);
 
+  terrno = TSDB_CODE_OUT_OF_MEMORY;
+  pJson = tjsonCreateObject();
+  if (pJson == NULL) goto _OVER;
+  if (dmEncodeFile(pJson, deployed) != 0) goto _OVER;
+  buffer = tjsonToString(pJson);
+  if (buffer == NULL) goto _OVER;
+  terrno = 0;
+
   pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  if (pFile == NULL) {
+  if (pFile == NULL) goto _OVER;
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to write %s since %s", file, terrstr());
-    goto _OVER;
-  }
 
-  len += snprintf(content + len, MAXLEN - len, "{\n");
+  int32_t len = strlen(buffer);
-  len += snprintf(content + len, MAXLEN - len, "  \"deployed\": %d\n", deployed);
+  if (taosWriteFile(pFile, buffer, len) <= 0) goto _OVER;
-  len += snprintf(content + len, MAXLEN - len, "}\n");
+  if (taosFsyncFile(pFile) < 0) goto _OVER;
-
-  if (taosWriteFile(pFile, content, len) != len) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to write file:%s since %s", file, terrstr());
-    goto _OVER;
-  }
-
-  if (taosFsyncFile(pFile) != 0) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to fsync file:%s since %s", file, terrstr());
-    goto _OVER;
-  }
 
   taosCloseFile(&pFile);
+  if (taosRenameFile(file, realfile) != 0) goto _OVER;
 
-  if (taosRenameFile(file, realfile) != 0) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to rename %s since %s", file, terrstr());
-    return -1;
-  }
-
-  dInfo("succeed to write %s, deployed:%d", realfile, deployed);
   code = 0;
+  dInfo("succeed to write file:%s, deloyed:%d", realfile, deployed);
 
 _OVER:
-  if (pFile != NULL) {
+  if (pJson != NULL) tjsonDelete(pJson);
-    taosCloseFile(&pFile);
+  if (buffer != NULL) taosMemoryFree(buffer);
-  }
+  if (pFile != NULL) taosCloseFile(&pFile);
 
+  if (code != 0) {
+    if (terrno == 0) terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write file:%s since %s, deloyed:%d", realfile, terrstr(), deployed);
+  }
   return code;
 }
 

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

@@ -34,6 +34,8 @@ SHashObj *mndDupDbHash(SHashObj *pOld);
 SHashObj *mndDupTopicHash(SHashObj *pOld);
 int32_t   mndValidateUserAuthInfo(SMnode *pMnode, SUserAuthVersion *pUsers, int32_t numOfUses, void **ppRsp,
                                   int32_t *pRspLen);
+int32_t   mndUserRemoveDb(SMnode *pMnode, STrans *pTrans, char *db);
+int32_t   mndUserRemoveTopic(SMnode *pMnode, STrans *pTrans, char *topic);
 
 #ifdef __cplusplus
 }

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

@@ -20,6 +20,8 @@
 
 #define CLUSTER_VER_NUMBE    1
 #define CLUSTER_RESERVE_SIZE 60
+char    tsVersionName[16] = "community";
+int64_t tsExpireTime = 0;
 
 static SSdbRaw *mndClusterActionEncode(SClusterObj *pCluster);
 static SSdbRow *mndClusterActionDecode(SSdbRaw *pRaw);
@@ -291,6 +293,18 @@ static int32_t mndRetrieveClusters(SRpcMsg *pMsg, SShowObj *pShow, SSDataBlock *
     pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
     colDataAppend(pColInfo, numOfRows, (const char *)&pCluster->createdTime, false);
 
+    char ver[12] = {0};
+    STR_WITH_MAXSIZE_TO_VARSTR(ver, tsVersionName, pShow->pMeta->pSchemas[cols].bytes);
+    pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
+    colDataAppend(pColInfo, numOfRows, (const char *)ver, false);
+
+    pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
+    if (tsExpireTime <= 0) {
+      colDataAppendNULL(pColInfo, numOfRows);
+    } else {
+      colDataAppend(pColInfo, numOfRows, (const char *)&tsExpireTime, false);
+    }
+
     sdbRelease(pSdb, pCluster);
     numOfRows++;
   }

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

@@ -742,6 +742,7 @@ SSdbRow *mndConsumerActionDecode(SSdbRaw *pRaw) {
   if (tDecodeSMqConsumerObj(buf, pConsumer) == NULL) {
     goto CM_DECODE_OVER;
   }
+  tmsgUpdateDnodeEpSet(&pConsumer->ep);
 
   terrno = TSDB_CODE_SUCCESS;
 

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

@@ -889,7 +889,7 @@ static int32_t mndProcessGetDbCfgReq(SRpcMsg *pReq) {
   cfgRsp.numOfRetensions = pDb->cfg.numOfRetensions;
   cfgRsp.pRetensions = pDb->cfg.pRetensions;
   cfgRsp.schemaless = pDb->cfg.schemaless;
-
+  cfgRsp.sstTrigger = pDb->cfg.sstTrigger;
   int32_t contLen = tSerializeSDbCfgRsp(NULL, 0, &cfgRsp);
   void   *pRsp = rpcMallocCont(contLen);
   if (pRsp == NULL) {
@@ -1051,17 +1051,7 @@ static int32_t mndDropDb(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb) {
   if (mndDropStreamByDb(pMnode, pTrans, pDb) != 0) goto _OVER;
   if (mndDropSmasByDb(pMnode, pTrans, pDb) != 0) goto _OVER;
   if (mndSetDropDbRedoActions(pMnode, pTrans, pDb) != 0) goto _OVER;
-
+  if (mndUserRemoveDb(pMnode, pTrans, pDb->name) != 0) goto _OVER;
-  SUserObj *pUser = mndAcquireUser(pMnode, pDb->createUser);
-  if (pUser != NULL) {
-    pUser->authVersion++;
-    SSdbRaw *pCommitRaw = mndUserActionEncode(pUser);
-    if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) {
-      mError("trans:%d, failed to append redo log since %s", pTrans->id, terrstr());
-      goto _OVER;
-    }
-    (void)sdbSetRawStatus(pCommitRaw, SDB_STATUS_READY);
-  }
 
   int32_t rspLen = 0;
   void   *pRsp = NULL;

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

@@ -180,6 +180,7 @@ static SSdbRow *mndDnodeActionDecode(SSdbRaw *pRaw) {
   SDB_GET_RESERVE(pRaw, dataPos, TSDB_DNODE_RESERVE_SIZE, _OVER)
 
   terrno = 0;
+  tmsgUpdateDnodeInfo(&pDnode->id, NULL, pDnode->fqdn, &pDnode->port);
 
 _OVER:
   if (terrno != 0) {
@@ -308,7 +309,8 @@ void mndGetDnodeData(SMnode *pMnode, SArray *pDnodeEps) {
   void   *pIter = NULL;
   while (1) {
     SDnodeObj *pDnode = NULL;
-    pIter = sdbFetch(pSdb, SDB_DNODE, pIter, (void **)&pDnode);
+    ESdbStatus objStatus = 0;
+    pIter = sdbFetchAll(pSdb, SDB_DNODE, pIter, (void **)&pDnode, &objStatus, true);
     if (pIter == NULL) break;
 
     SDnodeEp dnodeEp = {0};

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

@@ -293,7 +293,7 @@ static int32_t mndProcessCreateFuncReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  mInfo("func:%s, start to create", createReq.name);
+  mInfo("func:%s, start to create, size:%d", createReq.name, createReq.codeLen);
   if (mndCheckOperPrivilege(pMnode, pReq->info.conn.user, MND_OPER_CREATE_FUNC) != 0) {
     goto _OVER;
   }

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

@@ -15,13 +15,13 @@
 
 #define _DEFAULT_SOURCE
 #include "mndMnode.h"
+#include "mndCluster.h"
 #include "mndDnode.h"
 #include "mndPrivilege.h"
 #include "mndShow.h"
 #include "mndSync.h"
 #include "mndTrans.h"
 #include "tmisce.h"
-#include "mndCluster.h"
 
 #define MNODE_VER_NUMBER   1
 #define MNODE_RESERVE_SIZE 64
@@ -181,9 +181,8 @@ _OVER:
 
 static int32_t mndMnodeActionInsert(SSdb *pSdb, SMnodeObj *pObj) {
   mTrace("mnode:%d, perform insert action, row:%p", pObj->id, pObj);
-  pObj->pDnode = sdbAcquire(pSdb, SDB_DNODE, &pObj->id);
+  pObj->pDnode = sdbAcquireNotReadyObj(pSdb, SDB_DNODE, &pObj->id);
   if (pObj->pDnode == NULL) {
-    terrno = TSDB_CODE_MND_DNODE_NOT_EXIST;
     mError("mnode:%d, failed to perform insert action since %s", pObj->id, terrstr());
     return -1;
   }
@@ -748,7 +747,7 @@ static void mndReloadSyncConfig(SMnode *pMnode) {
       pNode->clusterId = mndGetClusterId(pMnode);
       pNode->nodePort = pObj->pDnode->port;
       tstrncpy(pNode->nodeFqdn, pObj->pDnode->fqdn, TSDB_FQDN_LEN);
-      (void)tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
+      tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
       mInfo("vgId:1, ep:%s:%u dnode:%d", pNode->nodeFqdn, pNode->nodePort, pNode->nodeId);
       if (pObj->pDnode->id == pMnode->selfDnodeId) {
         cfg.myIndex = cfg.replicaNum;
@@ -785,9 +784,9 @@ static void mndReloadSyncConfig(SMnode *pMnode) {
 
     int32_t code = syncReconfig(pMnode->syncMgmt.sync, &cfg);
     if (code != 0) {
-      mError("vgId:1, failed to reconfig mnode sync since %s", terrstr());
+      mError("vgId:1, mnode sync reconfig failed since %s", terrstr());
     } else {
-      mInfo("vgId:1, reconfig mnode sync success");
+      mInfo("vgId:1, mnode sync reconfig success");
     }
   }
 }

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

@@ -202,11 +202,13 @@ static SSdbRow *mndSmaActionDecode(SSdbRaw *pRaw) {
 
 _OVER:
   if (terrno != 0) {
-    mError("sma:%s, failed to decode from raw:%p since %s", pSma == NULL ? "null" : pSma->name, pRaw, terrstr());
+    if (pSma != NULL) {
-    taosMemoryFreeClear(pSma->expr);
+      mError("sma:%s, failed to decode from raw:%p since %s", pSma->name, pRaw, terrstr());
-    taosMemoryFreeClear(pSma->tagsFilter);
+      taosMemoryFreeClear(pSma->expr);
-    taosMemoryFreeClear(pSma->sql);
+      taosMemoryFreeClear(pSma->tagsFilter);
-    taosMemoryFreeClear(pSma->ast);
+      taosMemoryFreeClear(pSma->sql);
+      taosMemoryFreeClear(pSma->ast);
+    }
     taosMemoryFreeClear(pRow);
     return NULL;
   }

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

@@ -760,6 +760,27 @@ static SSdbRow *mndSubActionDecode(SSdbRaw *pRaw) {
     goto SUB_DECODE_OVER;
   }
 
+  // update epset saved in mnode
+  if (pSub->unassignedVgs != NULL) {
+    int32_t size = (int32_t)taosArrayGetSize(pSub->unassignedVgs);
+    for (int32_t i = 0; i < size; ++i) {
+      SMqVgEp *pMqVgEp = taosArrayGet(pSub->unassignedVgs, i);
+      tmsgUpdateDnodeEpSet(&pMqVgEp->epSet);
+    }
+  }
+  if (pSub->consumerHash != NULL) {
+    void *pIter = taosHashIterate(pSub->consumerHash, NULL);
+    while (pIter) {
+      SMqConsumerEp *pConsumerEp = pIter;
+      int32_t        size = (int32_t)taosArrayGetSize(pConsumerEp->vgs);
+      for (int32_t i = 0; i < size; ++i) {
+        SMqVgEp *pMqVgEp = taosArrayGet(pConsumerEp->vgs, i);
+        tmsgUpdateDnodeEpSet(&pMqVgEp->epSet);
+      }
+      pIter = taosHashIterate(pSub->consumerHash, pIter);
+    }
+  }
+
   terrno = TSDB_CODE_SUCCESS;
 
 SUB_DECODE_OVER:

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

@@ -271,9 +271,11 @@ SSyncFSM *mndSyncMakeFsm(SMnode *pMnode) {
 int32_t mndInitSync(SMnode *pMnode) {
   SSyncMgmt *pMgmt = &pMnode->syncMgmt;
   taosThreadMutexInit(&pMgmt->lock, NULL);
+  taosThreadMutexLock(&pMgmt->lock);
   pMgmt->transId = 0;
   pMgmt->transSec = 0;
   pMgmt->transSeq = 0;
+  taosThreadMutexUnlock(&pMgmt->lock);
 
   SSyncInfo syncInfo = {
       .snapshotStrategy = SYNC_STRATEGY_STANDARD_SNAPSHOT,
@@ -301,7 +303,7 @@ int32_t mndInitSync(SMnode *pMnode) {
     pNode->nodeId = pMgmt->replicas[i].id;
     pNode->nodePort = pMgmt->replicas[i].port;
     tstrncpy(pNode->nodeFqdn, pMgmt->replicas[i].fqdn, sizeof(pNode->nodeFqdn));
-    (void)tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
+    tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
     mInfo("vgId:1, index:%d ep:%s:%u dnode:%d cluster:%" PRId64, i, pNode->nodeFqdn, pNode->nodePort, pNode->nodeId,
           pNode->clusterId);
   }
@@ -369,6 +371,7 @@ int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) {
   if (pMgmt->transId != 0) {
     mError("trans:%d, can't be proposed since trans:%d already waiting for confirm", transId, pMgmt->transId);
     taosThreadMutexUnlock(&pMgmt->lock);
+    rpcFreeCont(req.pCont);
     terrno = TSDB_CODE_MND_LAST_TRANS_NOT_FINISHED;
     return terrno;
   }

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

@@ -131,7 +131,7 @@ static int32_t mndProcessTelemTimer(SRpcMsg* pReq) {
   taosThreadMutexUnlock(&pMgmt->lock);
 
   if (pCont != NULL) {
-    if (taosSendHttpReport(tsTelemServer, tsTelemPort, pCont, strlen(pCont), HTTP_FLAT) != 0) {
+    if (taosSendHttpReport(tsTelemServer, tsTelemUri, tsTelemPort, pCont, strlen(pCont), HTTP_FLAT) != 0) {
       mError("failed to send telemetry report");
     } else {
       mInfo("succeed to send telemetry report");

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

@@ -604,22 +604,19 @@ _OVER:
 }
 
 static int32_t mndDropTopic(SMnode *pMnode, STrans *pTrans, SRpcMsg *pReq, SMqTopicObj *pTopic) {
+  int32_t code = -1;
+  if (mndUserRemoveTopic(pMnode, pTrans, pTopic->name) != 0) goto _OVER;
+
   SSdbRaw *pCommitRaw = mndTopicActionEncode(pTopic);
-  if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) {
+  if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) goto _OVER;
-    mError("trans:%d, failed to append commit log since %s", pTrans->id, terrstr());
-    mndTransDrop(pTrans);
-    return -1;
-  }
   (void)sdbSetRawStatus(pCommitRaw, SDB_STATUS_DROPPED);
 
-  if (mndTransPrepare(pMnode, pTrans) != 0) {
+  if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
-    mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
+  code = 0;
-    mndTransDrop(pTrans);
-    return -1;
-  }
 
+_OVER:
   mndTransDrop(pTrans);
-  return 0;
+  return code;
 }
 
 static int32_t mndProcessDropTopicReq(SRpcMsg *pReq) {
@@ -890,6 +887,7 @@ int32_t mndCheckTopicExist(SMnode *pMnode, SDbObj *pDb) {
   return 0;
 }
 
+#if 0
 int32_t mndDropTopicByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb) {
   int32_t code = 0;
   SSdb   *pSdb = pMnode->pSdb;
@@ -917,3 +915,4 @@ int32_t mndDropTopicByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb) {
 
   return code;
 }
+#endif

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

@@ -329,6 +329,7 @@ static SSdbRow *mndTransActionDecode(SSdbRaw *pRaw) {
       action.pRaw = NULL;
     } else if (action.actionType == TRANS_ACTION_MSG) {
       SDB_GET_BINARY(pRaw, dataPos, (void *)&action.epSet, sizeof(SEpSet), _OVER);
+      tmsgUpdateDnodeEpSet(&action.epSet);
       SDB_GET_INT16(pRaw, dataPos, &action.msgType, _OVER)
       SDB_GET_INT8(pRaw, dataPos, &unused /*&action.msgSent*/, _OVER)
       SDB_GET_INT8(pRaw, dataPos, &unused /*&action.msgReceived*/, _OVER)

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

@@ -285,14 +285,35 @@ static int32_t mndUserActionInsert(SSdb *pSdb, SUserObj *pUser) {
   return 0;
 }
 
-static int32_t mndUserActionDelete(SSdb *pSdb, SUserObj *pUser) {
+static int32_t mndUserDupObj(SUserObj *pUser, SUserObj *pNew) {
-  mTrace("user:%s, perform delete action, row:%p", pUser->user, pUser);
+  memcpy(pNew, pUser, sizeof(SUserObj));
+  pNew->authVersion++;
+  pNew->updateTime = taosGetTimestampMs();
+
+  taosRLockLatch(&pUser->lock);
+  pNew->readDbs = mndDupDbHash(pUser->readDbs);
+  pNew->writeDbs = mndDupDbHash(pUser->writeDbs);
+  pNew->topics = mndDupTopicHash(pUser->topics);
+  taosRUnLockLatch(&pUser->lock);
+
+  if (pNew->readDbs == NULL || pNew->writeDbs == NULL || pNew->topics == NULL) {
+    return -1;
+  }
+  return 0;
+}
+
+static void mndUserFreeObj(SUserObj *pUser) {
   taosHashCleanup(pUser->readDbs);
   taosHashCleanup(pUser->writeDbs);
   taosHashCleanup(pUser->topics);
   pUser->readDbs = NULL;
   pUser->writeDbs = NULL;
   pUser->topics = NULL;
+}
+
+static int32_t mndUserActionDelete(SSdb *pSdb, SUserObj *pUser) {
+  mTrace("user:%s, perform delete action, row:%p", pUser->user, pUser);
+  mndUserFreeObj(pUser);
   return 0;
 }
 
@@ -516,19 +537,7 @@ static int32_t mndProcessAlterUserReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  memcpy(&newUser, pUser, sizeof(SUserObj));
+  if (mndUserDupObj(pUser, &newUser) != 0) goto _OVER;
-  newUser.authVersion++;
-  newUser.updateTime = taosGetTimestampMs();
-
-  taosRLockLatch(&pUser->lock);
-  newUser.readDbs = mndDupDbHash(pUser->readDbs);
-  newUser.writeDbs = mndDupDbHash(pUser->writeDbs);
-  newUser.topics = mndDupTopicHash(pUser->topics);
-  taosRUnLockLatch(&pUser->lock);
-
-  if (newUser.readDbs == NULL || newUser.writeDbs == NULL || newUser.topics == NULL) {
-    goto _OVER;
-  }
 
   if (alterReq.alterType == TSDB_ALTER_USER_PASSWD) {
     char pass[TSDB_PASSWORD_LEN + 1] = {0};
@@ -654,9 +663,7 @@ _OVER:
 
   mndReleaseUser(pMnode, pOperUser);
   mndReleaseUser(pMnode, pUser);
-  taosHashCleanup(newUser.writeDbs);
+  mndUserFreeObj(&newUser);
-  taosHashCleanup(newUser.readDbs);
-  taosHashCleanup(newUser.topics);
 
   return code;
 }
@@ -1007,3 +1014,74 @@ _OVER:
   tFreeSUserAuthBatchRsp(&batchRsp);
   return code;
 }
+
+int32_t mndUserRemoveDb(SMnode *pMnode, STrans *pTrans, char *db) {
+  int32_t   code = 0;
+  SSdb     *pSdb = pMnode->pSdb;
+  int32_t   len = strlen(db) + 1;
+  void     *pIter = NULL;
+  SUserObj *pUser = NULL;
+  SUserObj  newUser = {0};
+
+  while (1) {
+    pIter = sdbFetch(pSdb, SDB_USER, pIter, (void **)&pUser);
+    if (pIter == NULL) break;
+
+    code = -1;
+    if (mndUserDupObj(pUser, &newUser) != 0) break;
+
+    bool inRead = (taosHashGet(newUser.readDbs, db, len) != NULL);
+    bool inWrite = (taosHashGet(newUser.writeDbs, db, len) != NULL);
+    if (inRead || inWrite) {
+      (void)taosHashRemove(newUser.readDbs, db, len);
+      (void)taosHashRemove(newUser.writeDbs, db, len);
+
+      SSdbRaw *pCommitRaw = mndUserActionEncode(&newUser);
+      if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) break;
+      (void)sdbSetRawStatus(pCommitRaw, SDB_STATUS_READY);
+    }
+
+    mndUserFreeObj(&newUser);
+    sdbRelease(pSdb, pUser);
+    code = 0;
+  }
+
+  if (pUser != NULL) sdbRelease(pSdb, pUser);
+  if (pIter != NULL) sdbCancelFetch(pSdb, pIter);
+  mndUserFreeObj(&newUser);
+  return code;
+}
+
+int32_t mndUserRemoveTopic(SMnode *pMnode, STrans *pTrans, char *topic) {
+  int32_t   code = 0;
+  SSdb     *pSdb = pMnode->pSdb;
+  int32_t   len = strlen(topic) + 1;
+  void     *pIter = NULL;
+  SUserObj *pUser = NULL;
+  SUserObj  newUser = {0};
+
+  while (1) {
+    pIter = sdbFetch(pSdb, SDB_USER, pIter, (void **)&pUser);
+    if (pIter == NULL) break;
+
+    code = -1;
+    if (mndUserDupObj(pUser, &newUser) != 0) break;
+
+    bool inTopic = (taosHashGet(newUser.topics, topic, len) != NULL);
+    if (inTopic) {
+      (void)taosHashRemove(newUser.topics, topic, len);
+      SSdbRaw *pCommitRaw = mndUserActionEncode(&newUser);
+      if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) break;
+      (void)sdbSetRawStatus(pCommitRaw, SDB_STATUS_READY);
+    }
+
+    mndUserFreeObj(&newUser);
+    sdbRelease(pSdb, pUser);
+    code = 0;
+  }
+
+  if (pUser != NULL) sdbRelease(pSdb, pUser);
+  if (pIter != NULL) sdbCancelFetch(pSdb, pIter);
+  mndUserFreeObj(&newUser);
+  return code;
+}

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

@@ -1441,10 +1441,10 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
 
   {
     SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
-    if (pRaw == NULL) return -1;
+    if (pRaw == NULL) goto _OVER;
     if (mndTransAppendCommitlog(pTrans, pRaw) != 0) {
       sdbFreeRaw(pRaw);
-      return -1;
+      goto _OVER;
     }
     (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
   }

source/dnode/mnode/sdb/inc/sdb.h

@@ -291,6 +291,7 @@ int32_t sdbWriteWithoutFree(SSdb *pSdb, SSdbRaw *pRaw);
  * @return void* The object of the row.
  */
 void *sdbAcquire(SSdb *pSdb, ESdbType type, const void *pKey);
+void *sdbAcquireNotReadyObj(SSdb *pSdb, ESdbType type, const void *pKey);
 
 /**
  * @brief Release a row from sdb.

source/dnode/mnode/sdb/src/sdbFile.c

@@ -228,11 +228,12 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
   int32_t readLen = 0;
   int64_t ret = 0;
   char    file[PATH_MAX] = {0};
+  int32_t bufLen = TSDB_MAX_MSG_SIZE;
 
   snprintf(file, sizeof(file), "%s%ssdb.data", pSdb->currDir, TD_DIRSEP);
   mInfo("start to read sdb file:%s", file);
 
-  SSdbRaw *pRaw = taosMemoryMalloc(TSDB_MAX_MSG_SIZE + 100);
+  SSdbRaw *pRaw = taosMemoryMalloc(bufLen + 100);
   if (pRaw == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     mError("failed read sdb file since %s", terrstr());
@@ -243,7 +244,7 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
   if (pFile == NULL) {
     taosMemoryFree(pRaw);
     terrno = TAOS_SYSTEM_ERROR(errno);
-    mDebug("failed to read sdb file:%s since %s", file, terrstr());
+    mInfo("read sdb file:%s finished since %s", file, terrstr());
     return 0;
   }
 
@@ -275,14 +276,15 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
     }
 
     readLen = pRaw->dataLen + sizeof(int32_t);
-    if (readLen >= pRaw->dataLen) {
+    if (readLen >= bufLen) {
-      SSdbRaw *pNewRaw = taosMemoryMalloc(pRaw->dataLen + TSDB_MAX_MSG_SIZE);
+      bufLen = pRaw->dataLen * 2;
+      SSdbRaw *pNewRaw = taosMemoryMalloc(bufLen + 100);
       if (pNewRaw == NULL) {
         terrno = TSDB_CODE_OUT_OF_MEMORY;
-        mError("failed read sdb file since malloc new sdbRaw size:%d failed", pRaw->dataLen + TSDB_MAX_MSG_SIZE);
+        mError("failed read sdb file since malloc new sdbRaw size:%d failed", bufLen);
         goto _OVER;
       }
-      mInfo("malloc new sdbRaw size:%d, type:%d", pRaw->dataLen + TSDB_MAX_MSG_SIZE, pRaw->type);
+      mInfo("malloc new sdb raw size:%d, type:%d", bufLen, pRaw->type);
       memcpy(pNewRaw, pRaw, sizeof(SSdbRaw));
       sdbFreeRaw(pRaw);
       pRaw = pNewRaw;
@@ -636,15 +638,20 @@ int32_t sdbStartWrite(SSdb *pSdb, SSdbIter **ppIter) {
 }
 
 int32_t sdbStopWrite(SSdb *pSdb, SSdbIter *pIter, bool isApply, int64_t index, int64_t term, int64_t config) {
-  int32_t code = 0;
+  int32_t code = -1;
 
   if (!isApply) {
     mInfo("sdbiter:%p, not apply to sdb", pIter);
-    sdbCloseIter(pIter);
+    code = 0;
-    return 0;
+    goto _OVER;
+  }
+
+  if (taosFsyncFile(pIter->file) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    mError("sdbiter:%p, failed to fasync file %s since %s", pIter, pIter->name, terrstr());
+    goto _OVER;
   }
 
-  taosFsyncFile(pIter->file);
   taosCloseFile(&pIter->file);
   pIter->file = NULL;
 
@@ -653,14 +660,12 @@ int32_t sdbStopWrite(SSdb *pSdb, SSdbIter *pIter, bool isApply, int64_t index, i
   if (taosRenameFile(pIter->name, datafile) != 0) {
     terrno = TAOS_SYSTEM_ERROR(errno);
     mError("sdbiter:%p, failed to rename file %s to %s since %s", pIter, pIter->name, datafile, terrstr());
-    sdbCloseIter(pIter);
+    goto _OVER;
-    return -1;
   }
 
   if (sdbReadFile(pSdb) != 0) {
     mError("sdbiter:%p, failed to read from %s since %s", pIter, datafile, terrstr());
-    sdbCloseIter(pIter);
+    goto _OVER;
-    return -1;
   }
 
   if (config > 0) {
@@ -674,8 +679,11 @@ int32_t sdbStopWrite(SSdb *pSdb, SSdbIter *pIter, bool isApply, int64_t index, i
   }
 
   mInfo("sdbiter:%p, success applyed to sdb", pIter);
+  code = 0;
+
+_OVER:
   sdbCloseIter(pIter);
-  return 0;
+  return code;
 }
 
 int32_t sdbDoWrite(SSdb *pSdb, SSdbIter *pIter, void *pBuf, int32_t len) {

source/dnode/mnode/sdb/src/sdbHash.c

@@ -270,7 +270,7 @@ int32_t sdbWrite(SSdb *pSdb, SSdbRaw *pRaw) {
   return code;
 }
 
-void *sdbAcquire(SSdb *pSdb, ESdbType type, const void *pKey) {
+void *sdbAcquireAll(SSdb *pSdb, ESdbType type, const void *pKey, bool onlyReady) {
   terrno = 0;
 
   SHashObj *hash = sdbGetHash(pSdb, type);
@@ -306,10 +306,24 @@ void *sdbAcquire(SSdb *pSdb, ESdbType type, const void *pKey) {
       break;
   }
 
+  if (pRet == NULL) {
+    if (!onlyReady) {
+      terrno = 0;
+      atomic_add_fetch_32(&pRow->refCount, 1);
+      pRet = pRow->pObj;
+      sdbPrintOper(pSdb, pRow, "acquire");
+    }
+  }
+
   sdbUnLock(pSdb, type);
   return pRet;
 }
 
+void *sdbAcquire(SSdb *pSdb, ESdbType type, const void *pKey) { return sdbAcquireAll(pSdb, type, pKey, true); }
+void *sdbAcquireNotReadyObj(SSdb *pSdb, ESdbType type, const void *pKey) {
+  return sdbAcquireAll(pSdb, type, pKey, false);
+}
+
 static void sdbCheckRow(SSdb *pSdb, SSdbRow *pRow) {
   int32_t type = pRow->type;
   sdbWriteLock(pSdb, type);

source/dnode/snode/src/snode.c

@@ -124,6 +124,7 @@ FAIL:
 }
 
 void sndClose(SSnode *pSnode) {
+  streamMetaCommit(pSnode->pMeta);
   streamMetaClose(pSnode->pMeta);
   taosMemoryFree(pSnode->path);
   taosMemoryFree(pSnode);

source/dnode/vnode/inc/vnode.h

@@ -89,6 +89,7 @@ int32_t vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg);
 int32_t vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo);
 void    vnodeProposeWriteMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs);
 void    vnodeApplyWriteMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs);
+void    vnodeProposeCommitOnNeed(SVnode *pVnode);
 
 // meta
 typedef struct SMeta       SMeta;  // todo: remove
@@ -152,6 +153,8 @@ typedef struct SMTbCursor SMTbCursor;
 SMTbCursor *metaOpenTbCursor(SMeta *pMeta);
 void        metaCloseTbCursor(SMTbCursor *pTbCur);
 int32_t     metaTbCursorNext(SMTbCursor *pTbCur);
+int32_t     metaTbCursorPrev(SMTbCursor *pTbCur);
+
 #endif
 
 // tsdb

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

@@ -86,6 +86,7 @@ int32_t vnodeGetBatchMeta(SVnode* pVnode, SRpcMsg* pMsg);
 // vnodeCommit.c
 int32_t vnodeBegin(SVnode* pVnode);
 int32_t vnodeShouldCommit(SVnode* pVnode);
+void    vnodeUpdCommitSched(SVnode* pVnode);
 void    vnodeRollback(SVnode* pVnode);
 int32_t vnodeSaveInfo(const char* dir, const SVnodeInfo* pCfg);
 int32_t vnodeCommitInfo(const char* dir, const SVnodeInfo* pInfo);
@@ -103,6 +104,7 @@ void    vnodeSyncClose(SVnode* pVnode);
 void    vnodeRedirectRpcMsg(SVnode* pVnode, SRpcMsg* pMsg, int32_t code);
 bool    vnodeIsLeader(SVnode* pVnode);
 bool    vnodeIsRoleLeader(SVnode* pVnode);
+int     vnodeShouldCommit(SVnode* pVnode);
 
 #ifdef __cplusplus
 }

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

@@ -332,6 +332,11 @@ struct STsdbKeepCfg {
   int32_t keep2;
 };
 
+typedef struct SVCommitSched {
+  int64_t commitMs;
+  int64_t maxWaitMs;
+} SVCommitSched;
+
 struct SVnode {
   char*         path;
   SVnodeCfg     config;
@@ -350,6 +355,7 @@ struct SVnode {
   STQ*          pTq;
   SSink*        pSink;
   tsem_t        canCommit;
+  SVCommitSched commitSched;
   int64_t       sync;
   TdThreadMutex lock;
   bool          blocked;

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

@@ -203,6 +203,7 @@ _err:
 
 int metaClose(SMeta *pMeta) {
   if (pMeta) {
+    if (pMeta->pEnv) tdbAbort(pMeta->pEnv, pMeta->txn);
     if (pMeta->pCache) metaCacheClose(pMeta);
     if (pMeta->pIdx) metaCloseIdx(pMeta);
     if (pMeta->pStreamDb) tdbTbClose(pMeta->pStreamDb);

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

@@ -311,7 +311,7 @@ void metaCloseTbCursor(SMTbCursor *pTbCur) {
   }
 }
 
-int metaTbCursorNext(SMTbCursor *pTbCur) {
+int32_t metaTbCursorNext(SMTbCursor *pTbCur) {
   int    ret;
   void  *pBuf;
   STbCfg tbCfg;
@@ -335,6 +335,31 @@ int metaTbCursorNext(SMTbCursor *pTbCur) {
   return 0;
 }
 
+int32_t metaTbCursorPrev(SMTbCursor *pTbCur) {
+  int    ret;
+  void  *pBuf;
+  STbCfg tbCfg;
+
+  for (;;) {
+    ret = tdbTbcPrev(pTbCur->pDbc, &pTbCur->pKey, &pTbCur->kLen, &pTbCur->pVal, &pTbCur->vLen);
+    if (ret < 0) {
+      return -1;
+    }
+
+    tDecoderClear(&pTbCur->mr.coder);
+
+    metaGetTableEntryByVersion(&pTbCur->mr, ((SUidIdxVal *)pTbCur->pVal)[0].version, *(tb_uid_t *)pTbCur->pKey);
+    if (pTbCur->mr.me.type == TSDB_SUPER_TABLE) {
+      continue;
+    }
+
+    break;
+  }
+
+  return 0;
+}
+
+
 SSchemaWrapper *metaGetTableSchema(SMeta *pMeta, tb_uid_t uid, int32_t sver, int lock) {
   void           *pData = NULL;
   int             nData = 0;

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

@@ -15,4 +15,11 @@
 
 #include "tq.h"
 
-int tqCommit(STQ* pTq) { return tqOffsetCommitFile(pTq->pOffsetStore); }
+int tqCommit(STQ* pTq) {
+  if (streamMetaCommit(pTq->pStreamMeta) < 0) {
+    tqError("vgId:%d, failed to commit stream meta since %s", TD_VID(pTq->pVnode), terrstr());
+    return -1;
+  }
+
+  return tqOffsetCommitFile(pTq->pOffsetStore);
+}

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

@@ -1080,6 +1080,8 @@ static int32_t nextRowIterGet(CacheNextRowIter *pIter, TSDBROW **ppRow) {
 
           iMax[nMax] = i;
           max[nMax++] = pIter->input[i].pRow;
+        } else {
+          pIter->input[i].next = false;
         }
       }
     }

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

@@ -164,7 +164,7 @@ void* tsdbCacherowsReaderClose(void* pReader) {
 
   destroyLastBlockLoadInfo(p->pLoadInfo);
 
-  taosMemoryFree((void*) p->idstr);
+  taosMemoryFree((void*)p->idstr);
   taosMemoryFree(pReader);
   return NULL;
 }
@@ -241,7 +241,11 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32
     taosArrayPush(pLastCols, &p);
   }
 
-  tsdbTakeReadSnap(pr->pVnode->pTsdb, &pr->pReadSnap, "cache-l");
+  code = tsdbTakeReadSnap(pr->pVnode->pTsdb, &pr->pReadSnap, "cache-l");
+  if (code != TSDB_CODE_SUCCESS) {
+    goto _end;
+  }
+
   pr->pDataFReader = NULL;
   pr->pDataFReaderLast = NULL;
 
@@ -252,7 +256,7 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32
 
       code = doExtractCacheRow(pr, lruCache, pKeyInfo->uid, &pRow, &h);
       if (code != TSDB_CODE_SUCCESS) {
-        return code;
+        goto _end;
       }
 
       if (h == NULL) {
@@ -321,7 +325,7 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32
       STableKeyInfo* pKeyInfo = &pr->pTableList[i];
       code = doExtractCacheRow(pr, lruCache, pKeyInfo->uid, &pRow, &h);
       if (code != TSDB_CODE_SUCCESS) {
-        return code;
+        goto _end;
       }
 
       if (h == NULL) {

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

@@ -458,9 +458,8 @@ static int32_t tsdbMergeFileSet(STsdb *pTsdb, SDFileSet *pSetOld, SDFileSet *pSe
       taosMemoryFree(pHeadF);
     }
   } else {
-    nRef = pHeadF->nRef;
+    ASSERT(pHeadF->offset == pSetNew->pHeadF->offset);
-    *pHeadF = *pSetNew->pHeadF;
+    ASSERT(pHeadF->size == pSetNew->pHeadF->size);
-    pHeadF->nRef = nRef;
   }
 
   // data
@@ -481,9 +480,7 @@ static int32_t tsdbMergeFileSet(STsdb *pTsdb, SDFileSet *pSetOld, SDFileSet *pSe
       taosMemoryFree(pDataF);
     }
   } else {
-    nRef = pDataF->nRef;
+    pDataF->size = pSetNew->pDataF->size;
-    *pDataF = *pSetNew->pDataF;
-    pDataF->nRef = nRef;
   }
 
   // sma
@@ -504,9 +501,7 @@ static int32_t tsdbMergeFileSet(STsdb *pTsdb, SDFileSet *pSetOld, SDFileSet *pSe
       taosMemoryFree(pSmaF);
     }
   } else {
-    nRef = pSmaF->nRef;
+    pSmaF->size = pSetNew->pSmaF->size;
-    *pSmaF = *pSetNew->pSmaF;
-    pSmaF->nRef = nRef;
   }
 
   // stt

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

@@ -1758,11 +1758,14 @@ static int32_t doMergeBufAndFileRows(STsdbReader* pReader, STableBlockScanInfo*
     }
 
     if (minKey == k.ts) {
+      STSchema* pSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, pBlockScanInfo->uid);
+      if (pSchema == NULL) {
+        return terrno;
+      }
       if (init) {
-        tRowMerge(&merge, pRow);
+        tRowMergerAdd(&merge, pRow, pSchema);
       } else {
         init = true;
-        STSchema* pSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, pBlockScanInfo->uid);
         int32_t   code = tRowMergerInit(&merge, pRow, pSchema);
         if (code != TSDB_CODE_SUCCESS) {
           return code;
@@ -2333,32 +2336,33 @@ static int32_t buildComposedDataBlockImpl(STsdbReader* pReader, STableBlockScanI
                                           SBlockData* pBlockData, SLastBlockReader* pLastBlockReader) {
   SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo;
 
+  TSDBROW *pRow = NULL, *piRow = NULL;
   int64_t key = (pBlockData->nRow > 0 && (!pDumpInfo->allDumped)) ? pBlockData->aTSKEY[pDumpInfo->rowIndex] : INT64_MIN;
-  if (pBlockScanInfo->iter.hasVal && pBlockScanInfo->iiter.hasVal) {
+  if (pBlockScanInfo->iter.hasVal) {
-    return doMergeMultiLevelRows(pReader, pBlockScanInfo, pBlockData, pLastBlockReader);
+    pRow = getValidMemRow(&pBlockScanInfo->iter, pBlockScanInfo->delSkyline, pReader);
-  } else {
-    TSDBROW *pRow = NULL, *piRow = NULL;
-    if (pBlockScanInfo->iter.hasVal) {
-      pRow = getValidMemRow(&pBlockScanInfo->iter, pBlockScanInfo->delSkyline, pReader);
-    }
-
-    if (pBlockScanInfo->iiter.hasVal) {
-      piRow = getValidMemRow(&pBlockScanInfo->iiter, pBlockScanInfo->delSkyline, pReader);
-    }
-
-    // imem + file + last block
-    if (pBlockScanInfo->iiter.hasVal) {
-      return doMergeBufAndFileRows(pReader, pBlockScanInfo, piRow, &pBlockScanInfo->iiter, key, pLastBlockReader);
-    }
-
-    // mem + file + last block
-    if (pBlockScanInfo->iter.hasVal) {
-      return doMergeBufAndFileRows(pReader, pBlockScanInfo, pRow, &pBlockScanInfo->iter, key, pLastBlockReader);
-    }
-
-    // files data blocks + last block
-    return mergeFileBlockAndLastBlock(pReader, pLastBlockReader, key, pBlockScanInfo, pBlockData);
   }
+
+  if (pBlockScanInfo->iiter.hasVal) {
+    piRow = getValidMemRow(&pBlockScanInfo->iiter, pBlockScanInfo->delSkyline, pReader);
+  }
+
+  // two levels of mem-table does contain the valid rows
+  if (pRow != NULL && piRow != NULL) {
+    return doMergeMultiLevelRows(pReader, pBlockScanInfo, pBlockData, pLastBlockReader);
+  }
+
+  // imem + file + last block
+  if (pBlockScanInfo->iiter.hasVal) {
+    return doMergeBufAndFileRows(pReader, pBlockScanInfo, piRow, &pBlockScanInfo->iiter, key, pLastBlockReader);
+  }
+
+  // mem + file + last block
+  if (pBlockScanInfo->iter.hasVal) {
+    return doMergeBufAndFileRows(pReader, pBlockScanInfo, pRow, &pBlockScanInfo->iter, key, pLastBlockReader);
+  }
+
+  // files data blocks + last block
+  return mergeFileBlockAndLastBlock(pReader, pLastBlockReader, key, pBlockScanInfo, pBlockData);
 }
 
 static int32_t loadNeighborIfOverlap(SFileDataBlockInfo* pBlockInfo, STableBlockScanInfo* pBlockScanInfo,
@@ -2401,6 +2405,19 @@ static int32_t loadNeighborIfOverlap(SFileDataBlockInfo* pBlockInfo, STableBlock
   return code;
 }
 
+static void updateComposedBlockInfo(STsdbReader* pReader, double el, STableBlockScanInfo* pBlockScanInfo) {
+  SSDataBlock* pResBlock = pReader->pResBlock;
+
+  pResBlock->info.id.uid = (pBlockScanInfo != NULL) ? pBlockScanInfo->uid : 0;
+  pResBlock->info.dataLoad = 1;
+  blockDataUpdateTsWindow(pResBlock, pReader->suppInfo.slotId[0]);
+
+  setComposedBlockFlag(pReader, true);
+
+  pReader->cost.composedBlocks += 1;
+  pReader->cost.buildComposedBlockTime += el;
+}
+
 static int32_t buildComposedDataBlock(STsdbReader* pReader) {
   int32_t code = TSDB_CODE_SUCCESS;
 
@@ -2412,6 +2429,7 @@ static int32_t buildComposedDataBlock(STsdbReader* pReader) {
   bool    asc = ASCENDING_TRAVERSE(pReader->order);
   int64_t st = taosGetTimestampUs();
   int32_t step = asc ? 1 : -1;
+  double  el = 0;
 
   STableBlockScanInfo* pBlockScanInfo = NULL;
   if (pBlockInfo != NULL) {
@@ -2473,10 +2491,8 @@ static int32_t buildComposedDataBlock(STsdbReader* pReader) {
       }
     }
 
-    bool hasBlockLData = hasDataInLastBlock(pLastBlockReader);
-
     // no data in last block and block, no need to proceed.
-    if ((hasBlockData == false) && (hasBlockLData == false)) {
+    if (hasBlockData == false) {
       break;
     }
 
@@ -2495,15 +2511,8 @@ static int32_t buildComposedDataBlock(STsdbReader* pReader) {
   }
 
 _end:
-  pResBlock->info.id.uid = (pBlockScanInfo != NULL) ? pBlockScanInfo->uid : 0;
+  el = (taosGetTimestampUs() - st) / 1000.0;
-  pResBlock->info.dataLoad = 1;
+  updateComposedBlockInfo(pReader, el, pBlockScanInfo);
-  blockDataUpdateTsWindow(pResBlock, pReader->suppInfo.slotId[0]);
-
-  setComposedBlockFlag(pReader, true);
-  double el = (taosGetTimestampUs() - st) / 1000.0;
-
-  pReader->cost.composedBlocks += 1;
-  pReader->cost.buildComposedBlockTime += el;
 
   if (pResBlock->info.rows > 0) {
     tsdbDebug("%p uid:%" PRIu64 ", composed data block created, brange:%" PRIu64 "-%" PRIu64
@@ -2748,6 +2757,8 @@ static int32_t doLoadLastBlockSequentially(STsdbReader* pReader) {
     return code;
   }
 
+  SSDataBlock* pResBlock = pReader->pResBlock;
+
   while (1) {
     // load the last data block of current table
     STableBlockScanInfo* pScanInfo = *(STableBlockScanInfo**)pStatus->pTableIter;
@@ -2758,15 +2769,33 @@ static int32_t doLoadLastBlockSequentially(STsdbReader* pReader) {
       if (!hasNexTable) {
         return TSDB_CODE_SUCCESS;
       }
+
       continue;
     }
 
-    code = doBuildDataBlock(pReader);
+    int64_t st = taosGetTimestampUs();
-    if (code != TSDB_CODE_SUCCESS) {
+    while (1) {
-      return code;
+      bool hasBlockLData = hasDataInLastBlock(pLastBlockReader);
+
+      // no data in last block and block, no need to proceed.
+      if (hasBlockLData == false) {
+        break;
+      }
+
+      buildComposedDataBlockImpl(pReader, pScanInfo, &pReader->status.fileBlockData, pLastBlockReader);
+      if (pResBlock->info.rows >= pReader->capacity) {
+        break;
+      }
     }
 
-    if (pReader->pResBlock->info.rows > 0) {
+    double el = (taosGetTimestampUs() - st) / 1000.0;
+    updateComposedBlockInfo(pReader, el, pScanInfo);
+
+    if (pResBlock->info.rows > 0) {
+      tsdbDebug("%p uid:%" PRIu64 ", composed data block created, brange:%" PRIu64 "-%" PRIu64
+                " rows:%d, elapsed time:%.2f ms %s",
+                pReader, pResBlock->info.id.uid, pResBlock->info.window.skey, pResBlock->info.window.ekey,
+                pResBlock->info.rows, el, pReader->idStr);
       return TSDB_CODE_SUCCESS;
     }
 
@@ -2809,7 +2838,37 @@ static int32_t doBuildDataBlock(STsdbReader* pReader) {
   TSDBKEY keyInBuf = getCurrentKeyInBuf(pScanInfo, pReader);
 
   if (pBlockInfo == NULL) {  // build data block from last data file
-    code = buildComposedDataBlock(pReader);
+    SBlockData* pBData = &pReader->status.fileBlockData;
+    tBlockDataReset(pBData);
+
+    SSDataBlock* pResBlock = pReader->pResBlock;
+    tsdbDebug("load data in last block firstly, due to desc scan data, %s", pReader->idStr);
+
+    int64_t st = taosGetTimestampUs();
+
+    while (1) {
+      bool hasBlockLData = hasDataInLastBlock(pLastBlockReader);
+
+      // no data in last block and block, no need to proceed.
+      if (hasBlockLData == false) {
+        break;
+      }
+
+      buildComposedDataBlockImpl(pReader, pScanInfo, &pReader->status.fileBlockData, pLastBlockReader);
+      if (pResBlock->info.rows >= pReader->capacity) {
+        break;
+      }
+    }
+
+    double el = (taosGetTimestampUs() - st) / 1000.0;
+    updateComposedBlockInfo(pReader, el, pScanInfo);
+
+    if (pResBlock->info.rows > 0) {
+      tsdbDebug("%p uid:%" PRIu64 ", composed data block created, brange:%" PRIu64 "-%" PRIu64
+                    " rows:%d, elapsed time:%.2f ms %s",
+                pReader, pResBlock->info.id.uid, pResBlock->info.window.skey, pResBlock->info.window.ekey,
+                pResBlock->info.rows, el, pReader->idStr);
+    }
   } else if (fileBlockShouldLoad(pReader, pBlockInfo, pBlock, pScanInfo, keyInBuf, pLastBlockReader)) {
     code = doLoadFileBlockData(pReader, pBlockIter, &pStatus->fileBlockData, pScanInfo->uid);
     if (code != TSDB_CODE_SUCCESS) {
@@ -2828,10 +2887,38 @@ static int32_t doBuildDataBlock(STsdbReader* pReader) {
       // only return the rows in last block
       int64_t tsLast = getCurrentKeyInLastBlock(pLastBlockReader);
       ASSERT(tsLast >= pBlock->maxKey.ts);
-      tBlockDataReset(&pReader->status.fileBlockData);
 
+      SBlockData* pBData = &pReader->status.fileBlockData;
+      tBlockDataReset(pBData);
+
+      SSDataBlock* pResBlock = pReader->pResBlock;
       tsdbDebug("load data in last block firstly, due to desc scan data, %s", pReader->idStr);
-      code = buildComposedDataBlock(pReader);
+
+      int64_t st = taosGetTimestampUs();
+
+      while (1) {
+        bool hasBlockLData = hasDataInLastBlock(pLastBlockReader);
+
+        // no data in last block and block, no need to proceed.
+        if (hasBlockLData == false) {
+          break;
+        }
+
+        buildComposedDataBlockImpl(pReader, pScanInfo, &pReader->status.fileBlockData, pLastBlockReader);
+        if (pResBlock->info.rows >= pReader->capacity) {
+          break;
+        }
+      }
+
+      double el = (taosGetTimestampUs() - st) / 1000.0;
+      updateComposedBlockInfo(pReader, el, pScanInfo);
+
+      if (pResBlock->info.rows > 0) {
+        tsdbDebug("%p uid:%" PRIu64 ", composed data block created, brange:%" PRIu64 "-%" PRIu64
+                      " rows:%d, elapsed time:%.2f ms %s",
+                  pReader, pResBlock->info.id.uid, pResBlock->info.window.skey, pResBlock->info.window.ekey,
+                  pResBlock->info.rows, el, pReader->idStr);
+      }
     } else {  // whole block is required, return it directly
       SDataBlockInfo* pInfo = &pReader->pResBlock->info;
       pInfo->rows = pBlock->nRow;
@@ -3800,7 +3887,7 @@ int32_t tsdbReaderOpen(SVnode* pVnode, SQueryTableDataCond* pCond, void* pTableL
     if (pReader->type == TIMEWINDOW_RANGE_CONTAINED) {
       code = doOpenReaderImpl(pReader);
       if (code != TSDB_CODE_SUCCESS) {
-        return code;
+        goto _err;
       }
     } else {
       STsdbReader* pPrevReader = pReader->innerReader[0];
@@ -3821,7 +3908,7 @@ int32_t tsdbReaderOpen(SVnode* pVnode, SQueryTableDataCond* pCond, void* pTableL
 
       code = doOpenReaderImpl(pPrevReader);
       if (code != TSDB_CODE_SUCCESS) {
-        return code;
+        goto _err;
       }
     }
   }

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

@@ -682,6 +682,18 @@ int32_t tRowMergerInit2(SRowMerger *pMerger, STSchema *pResTSchema, TSDBROW *pRo
     }
 
     tsdbRowGetColVal(pRow, pTSchema, jCol++, pColVal);
+    if ((!COL_VAL_IS_NONE(pColVal)) && (!COL_VAL_IS_NULL(pColVal)) && IS_VAR_DATA_TYPE(pColVal->type)) {
+      uint8_t *pVal = pColVal->value.pData;
+    
+      pColVal->value.pData = NULL;
+      code = tRealloc(&pColVal->value.pData, pColVal->value.nData);
+      if (code) goto _exit;
+
+      if (pColVal->value.nData) {
+        memcpy(pColVal->value.pData, pVal, pColVal->value.nData);
+      }
+    }
+
     if (taosArrayPush(pMerger->pArray, pColVal) == NULL) {
       code = TSDB_CODE_OUT_OF_MEMORY;
       goto _exit;
@@ -719,13 +731,59 @@ int32_t tRowMergerAdd(SRowMerger *pMerger, TSDBROW *pRow, STSchema *pTSchema) {
     tsdbRowGetColVal(pRow, pTSchema, jCol++, pColVal);
 
     if (key.version > pMerger->version) {
+#if 0
       if (!COL_VAL_IS_NONE(pColVal)) {
-        taosArraySet(pMerger->pArray, iCol, pColVal);
+        if ((!COL_VAL_IS_NULL(pColVal)) && IS_VAR_DATA_TYPE(pColVal->type)) {
+          SColVal *tColVal = taosArrayGet(pMerger->pArray, iCol);
+          code = tRealloc(&tColVal->value.pData, pColVal->value.nData);
+          if (code) return code;
+
+          tColVal->value.nData = pColVal->value.nData;
+          if (pColVal->value.nData) {
+            memcpy(tColVal->value.pData, pColVal->value.pData, pColVal->value.nData);
+          }
+          tColVal->flag = 0;
+        } else {
+          taosArraySet(pMerger->pArray, iCol, pColVal);
+        }
+      }
+#endif
+      if (!COL_VAL_IS_NONE(pColVal)) {
+        if (IS_VAR_DATA_TYPE(pColVal->type)) {
+          SColVal *pTColVal = taosArrayGet(pMerger->pArray, iCol);
+          if (!COL_VAL_IS_NULL(pColVal)) {
+            code = tRealloc(&pTColVal->value.pData, pColVal->value.nData);
+            if (code) return code;
+
+            pTColVal->value.nData = pColVal->value.nData;
+            if (pTColVal->value.nData) {
+              memcpy(pTColVal->value.pData,  pColVal->value.pData, pTColVal->value.nData);
+            }
+            pTColVal->flag = 0;
+          } else {
+            tFree(pTColVal->value.pData);
+            pTColVal->value.pData = NULL;
+            taosArraySet(pMerger->pArray, iCol, pColVal);
+          }
+        } else {
+          taosArraySet(pMerger->pArray, iCol, pColVal);
+        }
       }
     } else if (key.version < pMerger->version) {
       SColVal *tColVal = (SColVal *)taosArrayGet(pMerger->pArray, iCol);
       if (COL_VAL_IS_NONE(tColVal) && !COL_VAL_IS_NONE(pColVal)) {
-        taosArraySet(pMerger->pArray, iCol, pColVal);
+        if ((!COL_VAL_IS_NULL(pColVal)) && IS_VAR_DATA_TYPE(pColVal->type)) {
+          code = tRealloc(&tColVal->value.pData, pColVal->value.nData);
+          if (code) return code;
+
+          tColVal->value.nData = pColVal->value.nData;        
+          if (pColVal->value.nData) {
+            memcpy(tColVal->value.pData, pColVal->value.pData, pColVal->value.nData);
+          }
+          tColVal->flag = 0;
+        } else {
+          taosArraySet(pMerger->pArray, iCol, pColVal);
+        }
       }
     } else {
       ASSERT(0 && "dup versions not allowed");
@@ -765,6 +823,18 @@ int32_t tRowMergerInit(SRowMerger *pMerger, TSDBROW *pRow, STSchema *pTSchema) {
   // other
   for (int16_t iCol = 1; iCol < pTSchema->numOfCols; iCol++) {
     tsdbRowGetColVal(pRow, pTSchema, iCol, pColVal);
+    if ((!COL_VAL_IS_NONE(pColVal)) && (!COL_VAL_IS_NULL(pColVal)) && IS_VAR_DATA_TYPE(pColVal->type)) {
+      uint8_t *pVal = pColVal->value.pData;
+    
+      pColVal->value.pData = NULL;
+      code = tRealloc(&pColVal->value.pData, pColVal->value.nData);
+      if (code) goto _exit;
+
+      if (pColVal->value.nData) {
+        memcpy(pColVal->value.pData, pVal, pColVal->value.nData);
+      }
+    }
+    
     if (taosArrayPush(pMerger->pArray, pColVal) == NULL) {
       code = TSDB_CODE_OUT_OF_MEMORY;
       goto _exit;
@@ -775,7 +845,16 @@ _exit:
   return code;
 }
 
-void tRowMergerClear(SRowMerger *pMerger) { taosArrayDestroy(pMerger->pArray); }
+void tRowMergerClear(SRowMerger *pMerger) { 
+  for (int32_t iCol = 1; iCol < pMerger->pTSchema->numOfCols; iCol++) {
+    SColVal *pTColVal = taosArrayGet(pMerger->pArray, iCol);
+    if (IS_VAR_DATA_TYPE(pTColVal->type)) {
+      tFree(pTColVal->value.pData);
+    }
+  }
+
+  taosArrayDestroy(pMerger->pArray); 
+}
 
 int32_t tRowMerge(SRowMerger *pMerger, TSDBROW *pRow) {
   int32_t  code = 0;
@@ -789,12 +868,47 @@ int32_t tRowMerge(SRowMerger *pMerger, TSDBROW *pRow) {
 
     if (key.version > pMerger->version) {
       if (!COL_VAL_IS_NONE(pColVal)) {
-        taosArraySet(pMerger->pArray, iCol, pColVal);
+        if (IS_VAR_DATA_TYPE(pColVal->type)) {
+          SColVal *pTColVal = taosArrayGet(pMerger->pArray, iCol);
+          if (!COL_VAL_IS_NULL(pColVal)) {
+            code = tRealloc(&pTColVal->value.pData, pColVal->value.nData);
+            if (code) goto _exit;
+
+            pTColVal->value.nData = pColVal->value.nData;
+            if (pTColVal->value.nData) {
+              memcpy(pTColVal->value.pData,  pColVal->value.pData, pTColVal->value.nData);
+            }
+            pTColVal->flag = 0;
+          } else {
+            tFree(pTColVal->value.pData);
+            pTColVal->value.pData = NULL;
+            taosArraySet(pMerger->pArray, iCol, pColVal);
+          }
+        } else {
+          taosArraySet(pMerger->pArray, iCol, pColVal);
+        }
       }
     } else if (key.version < pMerger->version) {
       SColVal *tColVal = (SColVal *)taosArrayGet(pMerger->pArray, iCol);
       if (COL_VAL_IS_NONE(tColVal) && !COL_VAL_IS_NONE(pColVal)) {
-        taosArraySet(pMerger->pArray, iCol, pColVal);
+        if (IS_VAR_DATA_TYPE(pColVal->type)) {
+          if (!COL_VAL_IS_NULL(pColVal)) {
+            code = tRealloc(&tColVal->value.pData, pColVal->value.nData);
+            if (code) goto _exit;
+
+            tColVal->value.nData = pColVal->value.nData;
+            if (tColVal->value.nData) {
+              memcpy(tColVal->value.pData,  pColVal->value.pData, tColVal->value.nData);
+            }
+            tColVal->flag = 0;
+          } else {
+            tFree(tColVal->value.pData);
+            tColVal->value.pData = NULL;
+            taosArraySet(pMerger->pArray, iCol, pColVal);
+          }
+        } else {
+          taosArraySet(pMerger->pArray, iCol, pColVal);
+        }
       }
     } else {
       ASSERT(0);

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

@@ -58,7 +58,25 @@ int vnodeBegin(SVnode *pVnode) {
   return 0;
 }
 
+void vnodeUpdCommitSched(SVnode *pVnode) {
+  int64_t randNum = taosRand();
+  pVnode->commitSched.commitMs = taosGetMonoTimestampMs();
+  pVnode->commitSched.maxWaitMs = tsVndCommitMaxIntervalMs + (randNum % tsVndCommitMaxIntervalMs);
+}
+
 int vnodeShouldCommit(SVnode *pVnode) {
+  if (!pVnode->inUse || !osDataSpaceAvailable()) {
+    return false;
+  }
+
+  SVCommitSched *pSched = &pVnode->commitSched;
+  int64_t nowMs = taosGetMonoTimestampMs();
+
+  return (((pVnode->inUse->size > pVnode->inUse->node.size) && (pSched->commitMs + SYNC_VND_COMMIT_MIN_MS < nowMs)) ||
+          (pVnode->inUse->size > 0 && pSched->commitMs + pSched->maxWaitMs < nowMs));
+}
+
+int vnodeShouldCommitOld(SVnode *pVnode) {
   if (pVnode->inUse) {
     return osDataSpaceAvailable() && (pVnode->inUse->size > pVnode->inUse->node.size);
   }
@@ -247,6 +265,7 @@ _exit:
   taosMemoryFree(pInfo);
   return code;
 }
+
 int vnodeAsyncCommit(SVnode *pVnode) {
   int32_t code = 0;
 
@@ -294,7 +313,9 @@ static int vnodeCommitImpl(SCommitInfo *pInfo) {
   SVnode *pVnode = pInfo->pVnode;
 
   vInfo("vgId:%d, start to commit, commitId:%" PRId64 " version:%" PRId64 " term: %" PRId64, TD_VID(pVnode),
-        pVnode->state.commitID, pVnode->state.applied, pVnode->state.applyTerm);
+        pInfo->info.state.commitID, pInfo->info.state.committed, pInfo->info.state.commitTerm);
+
+  vnodeUpdCommitSched(pVnode);
 
   // persist wal before starting
   if (walPersist(pVnode->pWal) < 0) {
@@ -308,8 +329,7 @@ static int vnodeCommitImpl(SCommitInfo *pInfo) {
     snprintf(dir, TSDB_FILENAME_LEN, "%s", pVnode->path);
   }
 
-  // walBeginSnapshot(pVnode->pWal, pVnode->state.applied);
+  syncBeginSnapshot(pVnode->sync, pInfo->info.state.committed);
-  syncBeginSnapshot(pVnode->sync, pVnode->state.applied);
 
   // commit each sub-system
   code = tsdbCommit(pVnode->pTsdb, pInfo);
@@ -351,7 +371,6 @@ static int vnodeCommitImpl(SCommitInfo *pInfo) {
     return -1;
   }
 
-  // walEndSnapshot(pVnode->pWal);
   syncEndSnapshot(pVnode->sync);
 
 _exit:

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

@@ -86,7 +86,7 @@ int32_t vnodeAlter(const char *path, SAlterVnodeReplicaReq *pReq, STfs *pTfs) {
     pNode->nodeId = pReq->replicas[i].id;
     pNode->nodePort = pReq->replicas[i].port;
     tstrncpy(pNode->nodeFqdn, pReq->replicas[i].fqdn, sizeof(pNode->nodeFqdn));
-    (void)tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
+    tmsgUpdateDnodeInfo(&pNode->nodeId, &pNode->clusterId, pNode->nodeFqdn, &pNode->nodePort);
     vInfo("vgId:%d, replica:%d ep:%s:%u dnode:%d", pReq->vgId, i, pNode->nodeFqdn, pNode->nodePort, pNode->nodeId);
   }
 
@@ -160,6 +160,8 @@ SVnode *vnodeOpen(const char *path, STfs *pTfs, SMsgCb msgCb) {
   taosThreadMutexInit(&pVnode->mutex, NULL);
   taosThreadCondInit(&pVnode->poolNotEmpty, NULL);
 
+  vnodeUpdCommitSched(pVnode);
+
   int8_t rollback = vnodeShouldRollback(pVnode);
 
   // open buffer pool
@@ -254,7 +256,7 @@ void vnodePostClose(SVnode *pVnode) { vnodeSyncPostClose(pVnode); }
 
 void vnodeClose(SVnode *pVnode) {
   if (pVnode) {
-    vnodeSyncCommit(pVnode);
+    tsem_wait(&pVnode->canCommit);
     vnodeSyncClose(pVnode);
     vnodeQueryClose(pVnode);
     walClose(pVnode->pWal);
@@ -263,6 +265,8 @@ void vnodeClose(SVnode *pVnode) {
     smaClose(pVnode->pSma);
     metaClose(pVnode->pMeta);
     vnodeCloseBufPool(pVnode);
+    tsem_post(&pVnode->canCommit);
+
     // destroy handle
     tsem_destroy(&(pVnode->canCommit));
     tsem_destroy(&pVnode->syncSem);

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

@@ -406,8 +406,10 @@ static int32_t vnodeSnapWriteInfo(SVSnapWriter *pWriter, uint8_t *pData, uint32_
     snprintf(dir, TSDB_FILENAME_LEN, "%s", pWriter->pVnode->path);
   }
 
-  SVnode *pVnode = pWriter->pVnode;
+  SVnodeStats vndStats = pWriter->info.config.vndStats;
+  SVnode     *pVnode = pWriter->pVnode;
   pWriter->info.config = pVnode->config;
+  pWriter->info.config.vndStats = vndStats;
   vDebug("vgId:%d, save config while write snapshot", pWriter->pVnode->config.vgId);
   if (vnodeSaveInfo(dir, &pWriter->info) < 0) {
     code = terrno;

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

@@ -203,6 +203,7 @@ int32_t vnodeProcessWriteMsg(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
   // skip header
   pReq = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
   len = pMsg->contLen - sizeof(SMsgHead);
+  bool needCommit = false;
 
   switch (pMsg->msgType) {
     /* META */
@@ -299,9 +300,8 @@ int32_t vnodeProcessWriteMsg(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
       vnodeProcessAlterConfigReq(pVnode, version, pReq, len, pRsp);
       break;
     case TDMT_VND_COMMIT:
-      vnodeSyncCommit(pVnode);
+      needCommit = true;
-      vnodeBegin(pVnode);
+      break;
-      goto _exit;
     default:
       vError("vgId:%d, unprocessed msg, %d", TD_VID(pVnode), pMsg->msgType);
       return -1;
@@ -318,7 +318,7 @@ int32_t vnodeProcessWriteMsg(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
   }
 
   // commit if need
-  if (vnodeShouldCommit(pVnode)) {
+  if (needCommit) {
     vInfo("vgId:%d, commit at version %" PRId64, TD_VID(pVnode), version);
     vnodeAsyncCommit(pVnode);
 

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

@@ -101,6 +101,64 @@ static void vnodeHandleProposeError(SVnode *pVnode, SRpcMsg *pMsg, int32_t code)
   }
 }
 
+static int32_t inline vnodeProposeMsg(SVnode *pVnode, SRpcMsg *pMsg, bool isWeak) {
+  int64_t seq = 0;
+
+  taosThreadMutexLock(&pVnode->lock);
+  int32_t code = syncPropose(pVnode->sync, pMsg, isWeak, &seq);
+  bool    wait = (code == 0 && vnodeIsMsgBlock(pMsg->msgType));
+  if (wait) {
+    ASSERT(!pVnode->blocked);
+    pVnode->blocked = true;
+    pVnode->blockSec = taosGetTimestampSec();
+    pVnode->blockSeq = seq;
+#if 0
+    pVnode->blockInfo = pMsg->info;
+#endif
+  }
+  taosThreadMutexUnlock(&pVnode->lock);
+
+  if (code > 0) {
+    vnodeHandleWriteMsg(pVnode, pMsg);
+  } else if (code < 0) {
+    if (terrno != 0) code = terrno;
+    vnodeHandleProposeError(pVnode, pMsg, code);
+  }
+
+  if (wait) vnodeWaitBlockMsg(pVnode, pMsg);
+  return code;
+}
+
+void vnodeProposeCommitOnNeed(SVnode *pVnode) {
+  if (!vnodeShouldCommit(pVnode)) {
+    return;
+  }
+
+  int32_t   contLen = sizeof(SMsgHead);
+  SMsgHead *pHead = rpcMallocCont(contLen);
+  pHead->contLen = contLen;
+  pHead->vgId = pVnode->config.vgId;
+
+  SRpcMsg rpcMsg = {0};
+  rpcMsg.msgType = TDMT_VND_COMMIT;
+  rpcMsg.contLen = contLen;
+  rpcMsg.pCont = pHead;
+  rpcMsg.info.noResp = 1;
+
+  bool isWeak = false;
+  if (vnodeProposeMsg(pVnode, &rpcMsg, isWeak) < 0) {
+    vTrace("vgId:%d, failed to propose vnode commit since %s", pVnode->config.vgId, terrstr());
+    goto _out;
+  }
+
+  vInfo("vgId:%d, proposed vnode commit", pVnode->config.vgId);
+
+_out:
+  vnodeUpdCommitSched(pVnode);
+  rpcFreeCont(rpcMsg.pCont);
+  rpcMsg.pCont = NULL;
+}
+
 #if BATCH_ENABLE
 
 static void inline vnodeProposeBatchMsg(SVnode *pVnode, SRpcMsg **pMsgArr, bool *pIsWeakArr, int32_t *arrSize) {
@@ -178,6 +236,8 @@ void vnodeProposeWriteMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs)
       continue;
     }
 
+    vnodeProposeCommitOnNeed(pVnode);
+
     code = vnodePreProcessWriteMsg(pVnode, pMsg);
     if (code != 0) {
       vGError("vgId:%d, msg:%p failed to pre-process since %s", vgId, pMsg, terrstr());
@@ -205,34 +265,6 @@ void vnodeProposeWriteMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs)
 
 #else
 
-static int32_t inline vnodeProposeMsg(SVnode *pVnode, SRpcMsg *pMsg, bool isWeak) {
-  int64_t seq = 0;
-
-  taosThreadMutexLock(&pVnode->lock);
-  int32_t code = syncPropose(pVnode->sync, pMsg, isWeak, &seq);
-  bool    wait = (code == 0 && vnodeIsMsgBlock(pMsg->msgType));
-  if (wait) {
-    ASSERT(!pVnode->blocked);
-    pVnode->blocked = true;
-    pVnode->blockSec = taosGetTimestampSec();
-    pVnode->blockSeq = seq;
-#if 0
-    pVnode->blockInfo = pMsg->info;
-#endif
-  }
-  taosThreadMutexUnlock(&pVnode->lock);
-
-  if (code > 0) {
-    vnodeHandleWriteMsg(pVnode, pMsg);
-  } else if (code < 0) {
-    if (terrno != 0) code = terrno;
-    vnodeHandleProposeError(pVnode, pMsg, code);
-  }
-
-  if (wait) vnodeWaitBlockMsg(pVnode, pMsg);
-  return code;
-}
-
 void vnodeProposeWriteMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs) {
   SVnode  *pVnode = pInfo->ahandle;
   int32_t  vgId = pVnode->config.vgId;
@@ -256,6 +288,8 @@ void vnodeProposeWriteMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs)
       continue;
     }
 
+    vnodeProposeCommitOnNeed(pVnode);
+
     code = vnodePreProcessWriteMsg(pVnode, pMsg);
     if (code != 0) {
       vGError("vgId:%d, msg:%p failed to pre-process since %s", vgId, pMsg, terrstr());

source/libs/catalog/inc/catalogInt.h

@@ -805,6 +805,7 @@ int32_t ctgMakeVgArray(SDBVgInfo* dbInfo);
 int32_t ctgAcquireVgMetaFromCache(SCatalog *pCtg, const char *dbFName, const char *tbName, SCtgDBCache **pDb, SCtgTbCache **pTb);
 int32_t ctgCopyTbMeta(SCatalog *pCtg, SCtgTbMetaCtx *ctx, SCtgDBCache **pDb, SCtgTbCache **pTb, STableMeta **pTableMeta, char* dbFName);
 void    ctgReleaseVgMetaToCache(SCatalog *pCtg, SCtgDBCache *dbCache, SCtgTbCache *pCache);
+void    ctgReleaseTbMetaToCache(SCatalog *pCtg, SCtgDBCache *dbCache, SCtgTbCache *pCache);
 
 extern SCatalogMgmt gCtgMgmt;
 extern SCtgDebug    gCTGDebug;

source/libs/catalog/src/catalog.c

@@ -598,10 +598,16 @@ int32_t ctgGetCachedTbVgMeta(SCatalog* pCtg, const SName* pTableName, SVgroupInf
 
   CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, pTableName, pVgroup));
 
+  ctgRUnlockVgInfo(dbCache);
+
   SCtgTbMetaCtx ctx = {0};
   ctx.pName = (SName*)pTableName;
   ctx.flag = CTG_FLAG_UNKNOWN_STB;
-  CTG_ERR_JRET(ctgCopyTbMeta(pCtg, &ctx, &dbCache, &tbCache, pTableMeta, db));
+  code = ctgCopyTbMeta(pCtg, &ctx, &dbCache, &tbCache, pTableMeta, db);
+
+  ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache);
+
+  CTG_RET(code);
 
 _return:
   

source/libs/catalog/src/ctgAsync.c

@@ -483,7 +483,7 @@ int32_t ctgInitTask(SCtgJob* pJob, CTG_TASK_TYPE type, void* param, int32_t* tas
 
 _return:
   CTG_UNLOCK(CTG_WRITE, &pJob->taskLock);
-  
+
   return code;
 }
 
@@ -905,6 +905,14 @@ int32_t ctgCallUserCb(void* param) {
   return TSDB_CODE_SUCCESS;
 }
 
+void ctgUpdateJobErrCode(SCtgJob* pJob, int32_t errCode) {
+  if (!NEED_CLIENT_REFRESH_VG_ERROR(errCode) || errCode == TSDB_CODE_SUCCESS) return;
+
+  atomic_store_32(&pJob->jobResCode, errCode);
+  qDebug("QID:0x%" PRIx64 " ctg job errCode updated to %s", pJob->queryId, tstrerror(errCode));
+  return;
+}
+
 int32_t ctgHandleTaskEnd(SCtgTask* pTask, int32_t rspCode) {
   SCtgJob* pJob = pTask->pJob;
   int32_t  code = 0;
@@ -924,6 +932,8 @@ int32_t ctgHandleTaskEnd(SCtgTask* pTask, int32_t rspCode) {
   if (taskDone < taosArrayGetSize(pJob->pTasks)) {
     qDebug("QID:0x%" PRIx64 " task done: %d, total: %d", pJob->queryId, taskDone,
            (int32_t)taosArrayGetSize(pJob->pTasks));
+
+    ctgUpdateJobErrCode(pJob, rspCode);
     return TSDB_CODE_SUCCESS;
   }
 
@@ -931,7 +941,8 @@ int32_t ctgHandleTaskEnd(SCtgTask* pTask, int32_t rspCode) {
 
 _return:
 
-  pJob->jobResCode = code;
+  ctgUpdateJobErrCode(pJob, rspCode);
+  // pJob->jobResCode = code;
 
   // taosSsleep(2);
   // qDebug("QID:0x%" PRIx64 " ctg after sleep", pJob->queryId);
@@ -988,6 +999,7 @@ int32_t ctgHandleGetTbMetaRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf
             CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, pName, &vgInfo, NULL, tReq));
 
             ctgReleaseVgInfoToCache(pCtg, dbCache);
+            dbCache = NULL;
           } else {
             SBuildUseDBInput input = {0};
 
@@ -1098,7 +1110,8 @@ _return:
   }
 
   if (code) {
-    ctgTaskError("Get table %d.%s.%s meta failed with error %s", pName->acctId, pName->dbname, pName->tname, tstrerror(code));
+    ctgTaskError("Get table %d.%s.%s meta failed with error %s", pName->acctId, pName->dbname, pName->tname,
+                 tstrerror(code));
   }
   if (pTask->res || code) {
     ctgHandleTaskEnd(pTask, code);
@@ -1156,6 +1169,7 @@ int32_t ctgHandleGetTbMetasRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBu
             CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, pName, &vgInfo, NULL, tReq));
 
             ctgReleaseVgInfoToCache(pCtg, dbCache);
+            dbCache = NULL;
           } else {
             SBuildUseDBInput input = {0};
 
@@ -1286,7 +1300,8 @@ _return:
       TSWAP(pTask->res, ctx->pResList);
       taskDone = true;
     }
-    ctgTaskError("Get table %d.%s.%s meta failed with error %s", pName->acctId, pName->dbname, pName->tname, tstrerror(code));
+    ctgTaskError("Get table %d.%s.%s meta failed with error %s", pName->acctId, pName->dbname, pName->tname,
+                 tstrerror(code));
   }
 
   if (pTask->res && taskDone) {

source/libs/catalog/src/ctgCache.c

@@ -2118,7 +2118,7 @@ int32_t ctgOpUpdateEpset(SCtgCacheOperation *operation) {
 
 _return:
 
-  if (dbCache) {
+  if (code == TSDB_CODE_SUCCESS && dbCache) {
     ctgWUnlockVgInfo(dbCache);
   }
 

source/libs/command/src/command.c

@@ -264,10 +264,10 @@ static void setCreateDBResultIntoDataBlock(SSDataBlock* pBlock, char* dbFName, S
   len += sprintf(
       buf2 + VARSTR_HEADER_SIZE,
       "CREATE DATABASE `%s` BUFFER %d CACHESIZE %d CACHEMODEL '%s' COMP %d DURATION %dm "
-      "WAL_FSYNC_PERIOD %d MAXROWS %d MINROWS %d KEEP %dm,%dm,%dm PAGES %d PAGESIZE %d PRECISION '%s' REPLICA %d "
+      "WAL_FSYNC_PERIOD %d MAXROWS %d MINROWS %d STT_TRIGGER %d KEEP %dm,%dm,%dm PAGES %d PAGESIZE %d PRECISION '%s' REPLICA %d "
       "WAL_LEVEL %d VGROUPS %d SINGLE_STABLE %d",
       dbFName, pCfg->buffer, pCfg->cacheSize, cacheModelStr(pCfg->cacheLast), pCfg->compression, pCfg->daysPerFile,
-      pCfg->walFsyncPeriod, pCfg->maxRows, pCfg->minRows, pCfg->daysToKeep0, pCfg->daysToKeep1, pCfg->daysToKeep2,
+      pCfg->walFsyncPeriod, pCfg->maxRows, pCfg->minRows,  pCfg->sstTrigger, pCfg->daysToKeep0, pCfg->daysToKeep1, pCfg->daysToKeep2,
       pCfg->pages, pCfg->pageSize, prec, pCfg->replications, pCfg->walLevel, pCfg->numOfVgroups,
       1 == pCfg->numOfStables);
 

source/libs/executor/inc/executorimpl.h

@@ -705,7 +705,8 @@ void doBuildResultDatablock(SOperatorInfo* pOperator, SOptrBasicInfo* pbInfo, SG
 
 bool hasLimitOffsetInfo(SLimitInfo* pLimitInfo);
 void initLimitInfo(const SNode* pLimit, const SNode* pSLimit, SLimitInfo* pLimitInfo);
-void applyLimitOffset(SLimitInfo* pLimitInfo, SSDataBlock* pBlock, SExecTaskInfo* pTaskInfo, SOperatorInfo* pOperator);
+void resetLimitInfoForNextGroup(SLimitInfo* pLimitInfo);
+bool applyLimitOffset(SLimitInfo* pLimitInfo, SSDataBlock* pBlock, SExecTaskInfo* pTaskInfo, SOperatorInfo* pOperator);
 
 void applyAggFunctionOnPartialTuples(SExecTaskInfo* taskInfo, SqlFunctionCtx* pCtx, SColumnInfoData* pTimeWindowData,
                                      int32_t offset, int32_t forwardStep, int32_t numOfTotal, int32_t numOfOutput);

source/libs/executor/src/exchangeoperator.c

@@ -584,7 +584,13 @@ int32_t doExtractResultBlocks(SExchangeInfo* pExchangeInfo, SSourceDataInfo* pDa
   int32_t index = 0;
   int32_t code = 0;
   while (index++ < pRetrieveRsp->numOfBlocks) {
-    SSDataBlock* pb = createOneDataBlock(pExchangeInfo->pDummyBlock, false);
+    SSDataBlock* pb = NULL;
+    if (taosArrayGetSize(pExchangeInfo->pRecycledBlocks) > 0) {
+      pb = *(SSDataBlock**)taosArrayPop(pExchangeInfo->pRecycledBlocks);
+      blockDataCleanup(pb);
+    } else {
+      pb = createOneDataBlock(pExchangeInfo->pDummyBlock, false);
+    }
 
     code = extractDataBlockFromFetchRsp(pb, pStart, NULL, &pStart);
     if (code != 0) {
@@ -732,9 +738,7 @@ int32_t handleLimitOffset(SOperatorInfo* pOperator, SLimitInfo* pLimitInfo, SSDa
     }
 
     // reset the value for a new group data
-    pLimitInfo->numOfOutputRows = 0;
+    resetLimitInfoForNextGroup(pLimitInfo);
-    pLimitInfo->remainOffset = pLimitInfo->limit.offset;
-
     // existing rows that belongs to previous group.
     if (pBlock->info.rows > 0) {
       return PROJECT_RETRIEVE_DONE;
@@ -760,7 +764,12 @@ int32_t handleLimitOffset(SOperatorInfo* pOperator, SLimitInfo* pLimitInfo, SSDa
     int32_t keepRows = (int32_t)(pLimitInfo->limit.limit - pLimitInfo->numOfOutputRows);
     blockDataKeepFirstNRows(pBlock, keepRows);
     if (pLimitInfo->slimit.limit > 0 && pLimitInfo->slimit.limit <= pLimitInfo->numOfOutputGroups) {
-      pOperator->status = OP_EXEC_DONE;
+      setOperatorCompleted(pOperator);
+    } else {
+      // current group limitation is reached, and future blocks of this group need to be discarded.
+      if (pBlock->info.rows == 0) {
+        return PROJECT_RETRIEVE_CONTINUE;
+      }
     }
 
     return PROJECT_RETRIEVE_DONE;

source/libs/executor/src/executil.c

@@ -954,7 +954,7 @@ static int32_t optimizeTbnameInCondImpl(void* metaHandle, int64_t suid, SArray*
           return -1;
         }
       } else {
-        qWarn("failed to get tableIds from by table name: %s, reason: %s", name, tstrerror(terrno));
+//        qWarn("failed to get tableIds from by table name: %s, reason: %s", name, tstrerror(terrno));
         terrno = 0;
       }
     }
@@ -1759,6 +1759,11 @@ void initLimitInfo(const SNode* pLimit, const SNode* pSLimit, SLimitInfo* pLimit
   pLimitInfo->remainGroupOffset = slimit.offset;
 }
 
+void resetLimitInfoForNextGroup(SLimitInfo* pLimitInfo) {
+  pLimitInfo->numOfOutputRows = 0;
+  pLimitInfo->remainOffset = pLimitInfo->limit.offset;
+}
+
 uint64_t tableListGetSize(const STableListInfo* pTableList) {
   ASSERT(taosArrayGetSize(pTableList->pTableList) == taosHashGetSize(pTableList->map));
   return taosArrayGetSize(pTableList->pTableList);

source/libs/executor/src/executor.c

@@ -24,12 +24,16 @@
 static TdThreadOnce initPoolOnce = PTHREAD_ONCE_INIT;
 int32_t             exchangeObjRefPool = -1;
 
-static void initRefPool() { exchangeObjRefPool = taosOpenRef(1024, doDestroyExchangeOperatorInfo); }
 static void cleanupRefPool() {
   int32_t ref = atomic_val_compare_exchange_32(&exchangeObjRefPool, exchangeObjRefPool, 0);
   taosCloseRef(ref);
 }
 
+static void initRefPool() { 
+  exchangeObjRefPool = taosOpenRef(1024, doDestroyExchangeOperatorInfo);   
+  atexit(cleanupRefPool);
+}
+
 static int32_t doSetSMABlock(SOperatorInfo* pOperator, void* input, size_t numOfBlocks, int32_t type, char* id) {
   ASSERT(pOperator != NULL);
   if (pOperator->operatorType != QUERY_NODE_PHYSICAL_PLAN_STREAM_SCAN) {
@@ -442,7 +446,6 @@ int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId,
   SExecTaskInfo** pTask = (SExecTaskInfo**)pTaskInfo;
 
   taosThreadOnce(&initPoolOnce, initRefPool);
-  atexit(cleanupRefPool);
 
   qDebug("start to create subplan task, TID:0x%" PRIx64 " QID:0x%" PRIx64, taskId, pSubplan->id.queryId);
 

source/libs/executor/src/groupoperator.c

@@ -593,8 +593,11 @@ void* getCurrentDataGroupInfo(const SPartitionOperatorInfo* pInfo, SDataGroupInf
 
     int32_t pageId = 0;
     pPage = getNewBufPage(pInfo->pBuf, &pageId);
-    taosArrayPush(p->pPageList, &pageId);
+    if (pPage == NULL) {
+      return pPage;
+    }
 
+    taosArrayPush(p->pPageList, &pageId);
     *(int32_t*)pPage = 0;
   } else {
     int32_t* curId = taosArrayGetLast(p->pPageList);
@@ -612,6 +615,11 @@ void* getCurrentDataGroupInfo(const SPartitionOperatorInfo* pInfo, SDataGroupInf
       // add a new page for current group
       int32_t pageId = 0;
       pPage = getNewBufPage(pInfo->pBuf, &pageId);
+      if (pPage == NULL) {
+        qError("failed to get new buffer, code:%s", tstrerror(terrno));
+        return NULL;
+      }
+
       taosArrayPush(p->pPageList, &pageId);
       memset(pPage, 0, getBufPageSize(pInfo->pBuf));
     }

source/libs/executor/src/projectoperator.c

@@ -175,8 +175,7 @@ static int32_t setInfoForNewGroup(SSDataBlock* pBlock, SLimitInfo* pLimitInfo, S
 
     // reset the value for a new group data
     // existing rows that belongs to previous group.
-    pLimitInfo->numOfOutputRows = 0;
+    resetLimitInfoForNextGroup(pLimitInfo);
-    pLimitInfo->remainOffset = pLimitInfo->limit.offset;
   }
 
   return PROJECT_RETRIEVE_DONE;
@@ -200,10 +199,18 @@ static int32_t doIngroupLimitOffset(SLimitInfo* pLimitInfo, uint64_t groupId, SS
   if (pLimitInfo->limit.limit >= 0 && pLimitInfo->numOfOutputRows + pBlock->info.rows >= pLimitInfo->limit.limit) {
     int32_t keepRows = (int32_t)(pLimitInfo->limit.limit - pLimitInfo->numOfOutputRows);
     blockDataKeepFirstNRows(pBlock, keepRows);
+
     // TODO: optimize it later when partition by + limit
+    // all retrieved requirement has been fulfilled, let's finish this
     if ((pLimitInfo->slimit.limit == -1 && pLimitInfo->currentGroupId == 0) ||
         (pLimitInfo->slimit.limit > 0 && pLimitInfo->slimit.limit <= pLimitInfo->numOfOutputGroups)) {
       setOperatorCompleted(pOperator);
+    } else {
+      // Even current group is done, there may be many vgroups remain existed, and we need to continue to retrieve data
+      // from next group. So let's continue this retrieve process
+      if (keepRows == 0) {
+        return PROJECT_RETRIEVE_CONTINUE;
+      }
     }
   }
 
@@ -279,7 +286,6 @@ SSDataBlock* doProjectOperation(SOperatorInfo* pOperator) {
       // for stream interval
       if (pBlock->info.type == STREAM_RETRIEVE || pBlock->info.type == STREAM_DELETE_RESULT ||
           pBlock->info.type == STREAM_DELETE_DATA) {
-        // printDataBlock1(pBlock, "project1");
         return pBlock;
       }
 
@@ -358,7 +364,6 @@ SSDataBlock* doProjectOperation(SOperatorInfo* pOperator) {
     pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
   }
 
-  // printDataBlock1(p, "project");
   return (p->info.rows > 0) ? p : NULL;
 }
 

source/libs/executor/src/scanoperator.c

@@ -173,7 +173,7 @@ static SResultRow* getTableGroupOutputBuf(SOperatorInfo* pOperator, uint64_t gro
   if (NULL == *pPage) {
     return NULL;
   }
-  
+
   return (SResultRow*)((char*)(*pPage) + p1->offset);
 }
 
@@ -257,7 +257,7 @@ static void doSetTagColumnData(STableScanBase* pTableScanInfo, SSDataBlock* pBlo
 }
 
 // todo handle the slimit info
-void applyLimitOffset(SLimitInfo* pLimitInfo, SSDataBlock* pBlock, SExecTaskInfo* pTaskInfo, SOperatorInfo* pOperator) {
+bool applyLimitOffset(SLimitInfo* pLimitInfo, SSDataBlock* pBlock, SExecTaskInfo* pTaskInfo, SOperatorInfo* pOperator) {
   SLimit*     pLimit = &pLimitInfo->limit;
   const char* id = GET_TASKID(pTaskInfo);
 
@@ -266,6 +266,7 @@ void applyLimitOffset(SLimitInfo* pLimitInfo, SSDataBlock* pBlock, SExecTaskInfo
       pLimitInfo->remainOffset -= pBlock->info.rows;
       blockDataEmpty(pBlock);
       qDebug("current block ignore due to offset, current:%" PRId64 ", %s", pLimitInfo->remainOffset, id);
+      return false;
     } else {
       blockDataTrimFirstNRows(pBlock, pLimitInfo->remainOffset);
       pLimitInfo->remainOffset = 0;
@@ -274,13 +275,14 @@ void applyLimitOffset(SLimitInfo* pLimitInfo, SSDataBlock* pBlock, SExecTaskInfo
 
   if (pLimit->limit != -1 && pLimit->limit <= (pLimitInfo->numOfOutputRows + pBlock->info.rows)) {
     // limit the output rows
-    int32_t overflowRows = pLimitInfo->numOfOutputRows + pBlock->info.rows - pLimit->limit;
+    int32_t keep = (int32_t)(pLimit->limit - pLimitInfo->numOfOutputRows);
-    int32_t keep = pBlock->info.rows - overflowRows;
 
     blockDataKeepFirstNRows(pBlock, keep);
     qDebug("output limit %" PRId64 " has reached, %s", pLimit->limit, id);
-    pOperator->status = OP_EXEC_DONE;
+    return true;
   }
+
+  return false;
 }
 
 static int32_t loadDataBlock(SOperatorInfo* pOperator, STableScanBase* pTableScanInfo, SSDataBlock* pBlock,
@@ -391,7 +393,10 @@ static int32_t loadDataBlock(SOperatorInfo* pOperator, STableScanBase* pTableSca
     }
   }
 
-  applyLimitOffset(&pTableScanInfo->limitInfo, pBlock, pTaskInfo, pOperator);
+  bool limitReached = applyLimitOffset(&pTableScanInfo->limitInfo, pBlock, pTaskInfo, pOperator);
+  if (limitReached) { // set operator flag is done
+    setOperatorCompleted(pOperator);
+  }
 
   pCost->totalRows += pBlock->info.rows;
   pTableScanInfo->limitInfo.numOfOutputRows = pCost->totalRows;
@@ -768,8 +773,7 @@ static SSDataBlock* doTableScan(SOperatorInfo* pOperator) {
 
     // reset value for the next group data output
     pOperator->status = OP_OPENED;
-    pInfo->base.limitInfo.numOfOutputRows = 0;
+    resetLimitInfoForNextGroup(&pInfo->base.limitInfo);
-    pInfo->base.limitInfo.remainOffset = pInfo->base.limitInfo.limit.offset;
 
     int32_t        num = 0;
     STableKeyInfo* pList = NULL;
@@ -1729,6 +1733,7 @@ static SSDataBlock* doStreamScan(SOperatorInfo* pOperator) {
     /*resetTableScanInfo(pTSInfo, pWin);*/
     tsdbReaderClose(pTSInfo->base.dataReader);
     pTSInfo->base.dataReader = NULL;
+    pInfo->pTableScanOp->status = OP_OPENED;
 
     pTSInfo->scanTimes = 0;
     pTSInfo->currentGroupId = -1;
@@ -2684,9 +2689,12 @@ int32_t stopGroupTableMergeScan(SOperatorInfo* pOperator) {
   taosArrayDestroy(pInfo->queryConds);
   pInfo->queryConds = NULL;
 
+  resetLimitInfoForNextGroup(&pInfo->limitInfo);
   return TSDB_CODE_SUCCESS;
 }
 
+// all data produced by this function only belongs to one group
+// slimit/soffset does not need to be concerned here, since this function only deal with data within one group.
 SSDataBlock* getSortedTableMergeScanBlockData(SSortHandle* pHandle, SSDataBlock* pResBlock, int32_t capacity,
                                               SOperatorInfo* pOperator) {
   STableMergeScanInfo* pInfo = pOperator->info;
@@ -2706,10 +2714,12 @@ SSDataBlock* getSortedTableMergeScanBlockData(SSortHandle* pHandle, SSDataBlock*
     }
   }
 
-  qDebug("%s get sorted row blocks, rows:%d", GET_TASKID(pTaskInfo), pResBlock->info.rows);
   applyLimitOffset(&pInfo->limitInfo, pResBlock, pTaskInfo, pOperator);
   pInfo->limitInfo.numOfOutputRows += pResBlock->info.rows;
 
+  qDebug("%s get sorted row block, rows:%d, limit:%"PRId64, GET_TASKID(pTaskInfo), pResBlock->info.rows,
+      pInfo->limitInfo.numOfOutputRows);
+
   return (pResBlock->info.rows > 0) ? pResBlock : NULL;
 }
 
@@ -2748,11 +2758,13 @@ SSDataBlock* doTableMergeScan(SOperatorInfo* pOperator) {
       pOperator->resultInfo.totalRows += pBlock->info.rows;
       return pBlock;
     } else {
+      // Data of this group are all dumped, let's try the next group
       stopGroupTableMergeScan(pOperator);
       if (pInfo->tableEndIndex >= tableListSize - 1) {
         setOperatorCompleted(pOperator);
         break;
       }
+
       pInfo->tableStartIndex = pInfo->tableEndIndex + 1;
       pInfo->groupId = tableListGetInfo(pTaskInfo->pTableInfoList, pInfo->tableStartIndex)->groupId;
       startGroupTableMergeScan(pOperator);
@@ -3221,7 +3233,9 @@ static void buildVnodeGroupedNtbTableCount(STableCountScanOperatorInfo* pInfo, S
   uint64_t groupId = calcGroupId(fullStbName, strlen(fullStbName));
   pRes->info.id.groupId = groupId;
   int64_t ntbNum = metaGetNtbNum(pInfo->readHandle.meta);
-  fillTableCountScanDataBlock(pSupp, dbName, "", ntbNum, pRes);
+  if (ntbNum != 0) {
+    fillTableCountScanDataBlock(pSupp, dbName, "", ntbNum, pRes);
+  }
 }
 
 static void buildVnodeGroupedStbTableCount(STableCountScanOperatorInfo* pInfo, STableCountScanSupp* pSupp,

source/libs/executor/src/sortoperator.c

@@ -680,11 +680,13 @@ SSDataBlock* getMultiwaySortedBlockData(SSortHandle* pHandle, SSDataBlock* pData
       break;
     }
 
+    bool limitReached = applyLimitOffset(&pInfo->limitInfo, p, pTaskInfo, pOperator);
+    if (limitReached) {
+      resetLimitInfoForNextGroup(&pInfo->limitInfo);
+    }
+
     if (p->info.rows > 0) {
-      applyLimitOffset(&pInfo->limitInfo, p, pTaskInfo, pOperator);
+      break;
-      if (p->info.rows > 0) {
-        break;
-      }
     }
   }
 
@@ -698,7 +700,6 @@ SSDataBlock* getMultiwaySortedBlockData(SSortHandle* pHandle, SSDataBlock* pData
       colDataAssign(pDst, pSrc, p->info.rows, &pDataBlock->info);
     }
 
-    pInfo->limitInfo.numOfOutputRows += p->info.rows;
     pDataBlock->info.rows = p->info.rows;
     pDataBlock->info.id.groupId = pInfo->groupId;
     pDataBlock->info.dataLoad = 1;

source/libs/executor/src/sysscanoperator.c

@@ -491,6 +491,7 @@ static SSDataBlock* sysTableScanUserTags(SOperatorInfo* pOperator) {
     pInfo->pCur = metaOpenTbCursor(pInfo->readHandle.meta);
   }
 
+  bool blockFull = false;
   while ((ret = metaTbCursorNext(pInfo->pCur)) == 0) {
     if (pInfo->pCur->mr.me.type != TSDB_CHILD_TABLE) {
       continue;
@@ -512,17 +513,24 @@ static SSDataBlock* sysTableScanUserTags(SOperatorInfo* pOperator) {
       T_LONG_JMP(pTaskInfo->env, terrno);
     }
 
-    sysTableUserTagsFillOneTableTags(pInfo, &smrSuperTable, &pInfo->pCur->mr, dbname, tableName, &numOfRows, dataBlock);
+    if ((smrSuperTable.me.stbEntry.schemaTag.nCols + numOfRows) > pOperator->resultInfo.capacity) {
-
+      metaTbCursorPrev(pInfo->pCur);
+      blockFull = true;
+    } else {
+      sysTableUserTagsFillOneTableTags(pInfo, &smrSuperTable, &pInfo->pCur->mr, dbname, tableName, &numOfRows, dataBlock);
+    }
+    
     metaReaderClear(&smrSuperTable);
 
-    if (numOfRows >= pOperator->resultInfo.capacity) {
+    if (blockFull || numOfRows >= pOperator->resultInfo.capacity) {
       relocateAndFilterSysTagsScanResult(pInfo, numOfRows, dataBlock, pOperator->exprSupp.pFilterInfo);
       numOfRows = 0;
 
       if (pInfo->pRes->info.rows > 0) {
         break;
       }
+
+      blockFull = false;
     }
   }
 
@@ -1918,6 +1926,13 @@ static SSDataBlock* doBlockInfoScan(SOperatorInfo* pOperator) {
   colDataAppend(pColInfo, 0, p, false);
   taosMemoryFree(p);
 
+  // make the valgrind happy that all memory buffer has been initialized already.
+  if (slotId != 0) {
+    SColumnInfoData* p1 = taosArrayGet(pBlock->pDataBlock, 0);
+    int64_t v = 0;
+    colDataAppendInt64(p1, 0, &v);
+  }
+
   pBlock->info.rows = 1;
   pOperator->status = OP_EXEC_DONE;
   return pBlock;

source/libs/executor/src/tsort.c

@@ -108,7 +108,7 @@ static int32_t sortComparCleanup(SMsortComparParam* cmpParam) {
   return TSDB_CODE_SUCCESS;
 }
 
-void tsortClearOrderdSource(SArray *pOrderedSource) {
+void tsortClearOrderdSource(SArray* pOrderedSource) {
   for (size_t i = 0; i < taosArrayGetSize(pOrderedSource); i++) {
     SSortSource** pSource = taosArrayGet(pOrderedSource, i);
     if (NULL == *pSource) {
@@ -121,6 +121,12 @@ void tsortClearOrderdSource(SArray *pOrderedSource) {
     if ((*pSource)->param && !(*pSource)->onlyRef) {
       taosMemoryFree((*pSource)->param);
     }
+
+    if (!(*pSource)->onlyRef && (*pSource)->src.pBlock) {
+      blockDataDestroy((*pSource)->src.pBlock);
+      (*pSource)->src.pBlock = NULL;
+    }
+
     taosMemoryFreeClear(*pSource);
   }
 
@@ -245,7 +251,8 @@ static int32_t sortComparInit(SMsortComparParam* pParam, SArray* pSources, int32
   if (pHandle->pBuf == NULL) {
     if (!osTempSpaceAvailable()) {
       code = TSDB_CODE_NO_AVAIL_DISK;
-      qError("Sort compare init failed since %s, %s", terrstr(code), pHandle->idStr);
+      terrno = code;
+      qError("Sort compare init failed since %s, %s", tstrerror(code), pHandle->idStr);
       return code;
     }
 
@@ -253,6 +260,7 @@ static int32_t sortComparInit(SMsortComparParam* pParam, SArray* pSources, int32
                               "sortComparInit", tsTempDir);
     dBufSetPrintInfo(pHandle->pBuf);
     if (code != TSDB_CODE_SUCCESS) {
+      terrno = code;
       return code;
     }
   }
@@ -276,6 +284,7 @@ static int32_t sortComparInit(SMsortComparParam* pParam, SArray* pSources, int32
       
       code = blockDataFromBuf(pSource->src.pBlock, pPage);
       if (code != TSDB_CODE_SUCCESS) {
+        terrno = code;
         return code;
       }
 
@@ -629,9 +638,9 @@ static int32_t createInitialSources(SSortHandle* pHandle) {
 
   if (pHandle->type == SORT_SINGLESOURCE_SORT) {
     SSortSource** pSource = taosArrayGet(pHandle->pOrderedSource, 0);
-    SSortSource* source = *pSource;
+    SSortSource*  source = *pSource;
     *pSource = NULL;
-    
+
     tsortClearOrderdSource(pHandle->pOrderedSource);
 
     while (1) {
@@ -659,6 +668,10 @@ static int32_t createInitialSources(SSortHandle* pHandle) {
         if (source->param && !source->onlyRef) {
           taosMemoryFree(source->param);
         }
+        if (!source->onlyRef && source->src.pBlock) {
+          blockDataDestroy(source->src.pBlock);
+          source->src.pBlock = NULL;
+        }
         taosMemoryFree(source);
         return code;
       }
@@ -672,6 +685,10 @@ static int32_t createInitialSources(SSortHandle* pHandle) {
           if (source->param && !source->onlyRef) {
             taosMemoryFree(source->param);
           }
+          if (!source->onlyRef && source->src.pBlock) {
+            blockDataDestroy(source->src.pBlock);
+            source->src.pBlock = NULL;
+          }
           taosMemoryFree(source);
           return code;
         }
@@ -849,8 +866,8 @@ SSortExecInfo tsortGetSortExecInfo(SSortHandle* pHandle) {
   SSortExecInfo info = {0};
 
   if (pHandle == NULL) {
-    info.sortMethod = SORT_QSORT_T; // by default
+    info.sortMethod = SORT_QSORT_T;  // by default
-    info.sortBuffer = 2 * 1048576;  // 2mb by default
+    info.sortBuffer = 2 * 1048576;   // 2mb by default
   } else {
     info.sortBuffer = pHandle->pageSize * pHandle->numOfPages;
     info.sortMethod = pHandle->inMemSort ? SORT_QSORT_T : SORT_SPILLED_MERGE_SORT_T;

source/libs/function/src/builtinsimpl.c

@@ -1662,7 +1662,9 @@ int32_t percentileFunction(SqlFunctionCtx* pCtx) {
 
 int32_t percentileFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   SVariant* pVal = &pCtx->param[1].param;
+  int32_t code = 0;
   double    v = 0;
+
   GET_TYPED_DATA(v, double, pVal->nType, &pVal->i);
 
   SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx);
@@ -1670,14 +1672,14 @@ int32_t percentileFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
 
   tMemBucket* pMemBucket = ppInfo->pMemBucket;
   if (pMemBucket != NULL && pMemBucket->total > 0) {  // check for null
-    int32_t code = getPercentile(pMemBucket, v, &ppInfo->result);
+    code = getPercentile(pMemBucket, v, &ppInfo->result);
-    if (code != TSDB_CODE_SUCCESS) {
-      tMemBucketDestroy(pMemBucket);
-      return code;
-    }
   }
 
   tMemBucketDestroy(pMemBucket);
+  if (code != TSDB_CODE_SUCCESS) {
+    return code;
+  }
+
   return functionFinalize(pCtx, pBlock);
 }
 
@@ -2644,7 +2646,7 @@ static int32_t doHandleDiff(SDiffInfo* pDiffInfo, int32_t type, const char* pv,
       int32_t v = *(int32_t*)pv;
       int64_t delta = factor * (v - pDiffInfo->prev.i64);  // direct previous may be null
       if (delta < 0 && pDiffInfo->ignoreNegative) {
-        colDataSetNull_f(pOutput->nullbitmap, pos);
+        colDataSetNull_f_s(pOutput, pos);
       } else {
         colDataAppendInt64(pOutput, pos, &delta);
       }
@@ -2657,7 +2659,7 @@ static int32_t doHandleDiff(SDiffInfo* pDiffInfo, int32_t type, const char* pv,
       int8_t  v = *(int8_t*)pv;
       int64_t delta = factor * (v - pDiffInfo->prev.i64);  // direct previous may be null
       if (delta < 0 && pDiffInfo->ignoreNegative) {
-        colDataSetNull_f(pOutput->nullbitmap, pos);
+        colDataSetNull_f_s(pOutput, pos);
       } else {
         colDataAppendInt64(pOutput, pos, &delta);
       }
@@ -2668,7 +2670,7 @@ static int32_t doHandleDiff(SDiffInfo* pDiffInfo, int32_t type, const char* pv,
       int16_t v = *(int16_t*)pv;
       int64_t delta = factor * (v - pDiffInfo->prev.i64);  // direct previous may be null
       if (delta < 0 && pDiffInfo->ignoreNegative) {
-        colDataSetNull_f(pOutput->nullbitmap, pos);
+        colDataSetNull_f_s(pOutput, pos);
       } else {
         colDataAppendInt64(pOutput, pos, &delta);
       }
@@ -2680,7 +2682,7 @@ static int32_t doHandleDiff(SDiffInfo* pDiffInfo, int32_t type, const char* pv,
       int64_t v = *(int64_t*)pv;
       int64_t delta = factor * (v - pDiffInfo->prev.i64);  // direct previous may be null
       if (delta < 0 && pDiffInfo->ignoreNegative) {
-        colDataSetNull_f(pOutput->nullbitmap, pos);
+        colDataSetNull_f_s(pOutput, pos);
       } else {
         colDataAppendInt64(pOutput, pos, &delta);
       }
@@ -2691,7 +2693,7 @@ static int32_t doHandleDiff(SDiffInfo* pDiffInfo, int32_t type, const char* pv,
       float  v = *(float*)pv;
       double delta = factor * (v - pDiffInfo->prev.d64);                               // direct previous may be null
       if ((delta < 0 && pDiffInfo->ignoreNegative) || isinf(delta) || isnan(delta)) {  // check for overflow
-        colDataSetNull_f(pOutput->nullbitmap, pos);
+        colDataSetNull_f_s(pOutput, pos);
       } else {
         colDataAppendDouble(pOutput, pos, &delta);
       }
@@ -2702,7 +2704,7 @@ static int32_t doHandleDiff(SDiffInfo* pDiffInfo, int32_t type, const char* pv,
       double v = *(double*)pv;
       double delta = factor * (v - pDiffInfo->prev.d64);                               // direct previous may be null
       if ((delta < 0 && pDiffInfo->ignoreNegative) || isinf(delta) || isnan(delta)) {  // check for overflow
-        colDataSetNull_f(pOutput->nullbitmap, pos);
+        colDataSetNull_f_s(pOutput, pos);
       } else {
         colDataAppendDouble(pOutput, pos, &delta);
       }
@@ -2737,7 +2739,7 @@ int32_t diffFunction(SqlFunctionCtx* pCtx) {
 
       if (colDataIsNull_f(pInputCol->nullbitmap, i)) {
         if (pDiffInfo->includeNull) {
-          colDataSetNull_f(pOutput->nullbitmap, pos);
+          colDataSetNull_f_s(pOutput, pos);
 
           numOfElems += 1;
         }
@@ -2775,8 +2777,7 @@ int32_t diffFunction(SqlFunctionCtx* pCtx) {
 
       if (colDataIsNull_f(pInputCol->nullbitmap, i)) {
         if (pDiffInfo->includeNull) {
-          colDataSetNull_f(pOutput->nullbitmap, pos);
+          colDataSetNull_f_s(pOutput, pos);
-
           numOfElems += 1;
         }
         continue;
@@ -3060,14 +3061,12 @@ static int32_t doSaveTupleData(SSerializeDataHandle* pHandle, const void* pBuf,
     if (pHandle->currentPage == -1) {
       pPage = getNewBufPage(pHandle->pBuf, &pHandle->currentPage);
       if (pPage == NULL) {
-        terrno = TSDB_CODE_NO_AVAIL_DISK;
         return terrno;
       }
       pPage->num = sizeof(SFilePage);
     } else {
       pPage = getBufPage(pHandle->pBuf, pHandle->currentPage);
       if (pPage == NULL) {
-        terrno = TSDB_CODE_NO_AVAIL_DISK;
         return terrno;
       }
       if (pPage->num + length > getBufPageSize(pHandle->pBuf)) {
@@ -3075,7 +3074,6 @@ static int32_t doSaveTupleData(SSerializeDataHandle* pHandle, const void* pBuf,
         releaseBufPage(pHandle->pBuf, pPage);
         pPage = getNewBufPage(pHandle->pBuf, &pHandle->currentPage);
         if (pPage == NULL) {
-          terrno = TSDB_CODE_NO_AVAIL_DISK;
           return terrno;
         }
         pPage->num = sizeof(SFilePage);
@@ -3122,7 +3120,6 @@ static int32_t doUpdateTupleData(SSerializeDataHandle* pHandle, const void* pBuf
   if (pHandle->pBuf != NULL) {
     SFilePage* pPage = getBufPage(pHandle->pBuf, pPos->pageId);
     if (pPage == NULL) {
-      terrno = TSDB_CODE_NO_AVAIL_DISK;
       return terrno;
     }
     memcpy(pPage->data + pPos->offset, pBuf, length);

source/libs/function/src/tpercentile.c

@@ -43,8 +43,8 @@ static SFilePage *loadDataFromFilePage(tMemBucket *pMemBucket, int32_t slotIdx)
     if (pg == NULL) {
       return NULL;
     }
-    memcpy(buffer->data + offset, pg->data, (size_t)(pg->num * pMemBucket->bytes));
 
+    memcpy(buffer->data + offset, pg->data, (size_t)(pg->num * pMemBucket->bytes));
     offset += (int32_t)(pg->num * pMemBucket->bytes);
   }
 
@@ -92,6 +92,7 @@ static void resetPosInfo(SSlotInfo *pInfo) {
 
 int32_t findOnlyResult(tMemBucket *pMemBucket, double *result) {
   ASSERT(pMemBucket->total == 1);
+  terrno = 0;
 
   for (int32_t i = 0; i < pMemBucket->numOfSlots; ++i) {
     tMemBucketSlot *pSlot = &pMemBucket->pSlots[i];
@@ -108,7 +109,7 @@ int32_t findOnlyResult(tMemBucket *pMemBucket, double *result) {
       int32_t   *pageId = taosArrayGet(list, 0);
       SFilePage *pPage = getBufPage(pMemBucket->pBuffer, *pageId);
       if (pPage == NULL) {
-        return TSDB_CODE_NO_AVAIL_DISK;
+        return terrno;
       }
       ASSERT(pPage->num == 1);
 
@@ -275,7 +276,7 @@ tMemBucket *tMemBucketCreate(int16_t nElemSize, int16_t dataType, double minval,
     return NULL;
   }
 
-  int32_t ret = createDiskbasedBuf(&pBucket->pBuffer, pBucket->bufPageSize, pBucket->bufPageSize * 512, "1", tsTempDir);
+  int32_t ret = createDiskbasedBuf(&pBucket->pBuffer, pBucket->bufPageSize, pBucket->bufPageSize * 1024, "1", tsTempDir);
   if (ret != 0) {
     tMemBucketDestroy(pBucket);
     return NULL;
@@ -387,7 +388,7 @@ int32_t tMemBucketPut(tMemBucket *pBucket, const void *data, size_t size) {
 
       pSlot->info.data = getNewBufPage(pBucket->pBuffer, &pageId);
       if (pSlot->info.data == NULL) {
-        return TSDB_CODE_NO_AVAIL_DISK;
+        return terrno;
       }
       pSlot->info.pageId = pageId;
       taosArrayPush(pPageIdList, &pageId);
@@ -481,8 +482,9 @@ int32_t getPercentileImpl(tMemBucket *pMemBucket, int32_t count, double fraction
         // data in buffer and file are merged together to be processed.
         SFilePage *buffer = loadDataFromFilePage(pMemBucket, i);
         if (buffer == NULL) {
-          return TSDB_CODE_NO_AVAIL_DISK;
+          return terrno;
         }
+
         int32_t    currentIdx = count - num;
 
         char *thisVal = buffer->data + pMemBucket->bytes * currentIdx;
@@ -519,7 +521,7 @@ int32_t getPercentileImpl(tMemBucket *pMemBucket, int32_t count, double fraction
           int32_t *pageId = taosArrayGet(list, f);
           SFilePage *pg = getBufPage(pMemBucket->pBuffer, *pageId);
           if (pg == NULL) {
-            return TSDB_CODE_NO_AVAIL_DISK;
+            return terrno;
           }
 
           int32_t code = tMemBucketPut(pMemBucket, pg->data, (int32_t)pg->num);

source/libs/monitor/src/monMain.c

@@ -20,6 +20,7 @@
 #include "ttime.h"
 
 static SMonitor tsMonitor = {0};
+static char* tsMonUri = "/report";
 
 void monRecordLog(int64_t ts, ELogLevel level, const char *content) {
   taosThreadMutexLock(&tsMonitor.lock);
@@ -550,7 +551,7 @@ void monSendReport() {
   // uDebugL("report cont:%s\n", pCont);
   if (pCont != NULL) {
     EHttpCompFlag flag = tsMonitor.cfg.comp ? HTTP_GZIP : HTTP_FLAT;
-    if (taosSendHttpReport(tsMonitor.cfg.server, tsMonitor.cfg.port, pCont, strlen(pCont), flag) != 0) {
+    if (taosSendHttpReport(tsMonitor.cfg.server, tsMonUri, tsMonitor.cfg.port, pCont, strlen(pCont), flag) != 0) {
       uError("failed to send monitor msg");
     }
     taosMemoryFree(pCont);

source/libs/parser/src/parInsertStmt.c

@@ -425,6 +425,27 @@ int32_t qCloneStmtDataBlock(void** pDst, void* pSrc) {
     pBlock->pTableMeta = pNewMeta;
   }
 
+  if (pBlock->boundColumnInfo.boundColumns) {
+    int32_t size = pBlock->boundColumnInfo.numOfCols * sizeof(col_id_t);
+    void* tmp = taosMemoryMalloc(size);
+    memcpy(tmp, pBlock->boundColumnInfo.boundColumns, size);
+    pBlock->boundColumnInfo.boundColumns = tmp;
+  }
+
+  if (pBlock->boundColumnInfo.cols) {
+    int32_t size = pBlock->boundColumnInfo.numOfCols * sizeof(SBoundColumn);
+    void* tmp = taosMemoryMalloc(size);
+    memcpy(tmp, pBlock->boundColumnInfo.cols, size);
+    pBlock->boundColumnInfo.cols = tmp;
+  }
+
+  if (pBlock->boundColumnInfo.colIdxInfo) {
+    int32_t size = pBlock->boundColumnInfo.numOfBound * sizeof(SBoundIdxInfo);
+    void* tmp = taosMemoryMalloc(size);
+    memcpy(tmp, pBlock->boundColumnInfo.colIdxInfo, size);
+    pBlock->boundColumnInfo.colIdxInfo = tmp;
+  }
+
   return qResetStmtDataBlock(*pDst, false);
 }
 
@@ -437,7 +458,7 @@ int32_t qRebuildStmtDataBlock(void** pDst, void* pSrc, uint64_t uid, int32_t vgI
   STableDataBlocks* pBlock = (STableDataBlocks*)*pDst;
   pBlock->pData = taosMemoryMalloc(pBlock->nAllocSize);
   if (NULL == pBlock->pData) {
-    qFreeStmtDataBlock(pBlock);
+    qDestroyStmtDataBlock(pBlock);
     return TSDB_CODE_OUT_OF_MEMORY;
   }
 

source/libs/planner/src/planLogicCreater.c

@@ -1334,6 +1334,7 @@ static int32_t createSetOpLogicNode(SLogicPlanContext* pCxt, SSetOperator* pSetO
   }
 
   if (TSDB_CODE_SUCCESS == code) {
+    pSetOp->precision = pSetOperator->precision;
     *pLogicNode = (SLogicNode*)pSetOp;
   } else {
     nodesDestroyNode((SNode*)pSetOp);

source/libs/planner/src/planOptimizer.c

@@ -1080,29 +1080,29 @@ static bool sortPriKeyOptMayBeOptimized(SLogicNode* pNode) {
     return false;
   }
   SSortLogicNode* pSort = (SSortLogicNode*)pNode;
-  if (pSort->groupSort || !sortPriKeyOptIsPriKeyOrderBy(pSort->pSortKeys) || 1 != LIST_LENGTH(pSort->node.pChildren)) {
+  if (!sortPriKeyOptIsPriKeyOrderBy(pSort->pSortKeys) || 1 != LIST_LENGTH(pSort->node.pChildren)) {
     return false;
   }
   return true;
 }
 
-static int32_t sortPriKeyOptGetSequencingNodesImpl(SLogicNode* pNode, bool* pNotOptimize,
+static int32_t sortPriKeyOptGetSequencingNodesImpl(SLogicNode* pNode, bool groupSort, bool* pNotOptimize,
                                                    SNodeList** pSequencingNodes) {
   switch (nodeType(pNode)) {
     case QUERY_NODE_LOGIC_PLAN_SCAN: {
       SScanLogicNode* pScan = (SScanLogicNode*)pNode;
-      if (NULL != pScan->pGroupTags || TSDB_SYSTEM_TABLE == pScan->tableType) {
+      if ((!groupSort && NULL != pScan->pGroupTags) || TSDB_SYSTEM_TABLE == pScan->tableType) {
         *pNotOptimize = true;
         return TSDB_CODE_SUCCESS;
       }
       return nodesListMakeAppend(pSequencingNodes, (SNode*)pNode);
     }
     case QUERY_NODE_LOGIC_PLAN_JOIN: {
-      int32_t code = sortPriKeyOptGetSequencingNodesImpl((SLogicNode*)nodesListGetNode(pNode->pChildren, 0),
+      int32_t code = sortPriKeyOptGetSequencingNodesImpl((SLogicNode*)nodesListGetNode(pNode->pChildren, 0), groupSort,
                                                          pNotOptimize, pSequencingNodes);
       if (TSDB_CODE_SUCCESS == code) {
-        code = sortPriKeyOptGetSequencingNodesImpl((SLogicNode*)nodesListGetNode(pNode->pChildren, 1), pNotOptimize,
+        code = sortPriKeyOptGetSequencingNodesImpl((SLogicNode*)nodesListGetNode(pNode->pChildren, 1), groupSort,
-                                                   pSequencingNodes);
+                                                   pNotOptimize, pSequencingNodes);
       }
       return code;
     }
@@ -1121,13 +1121,13 @@ static int32_t sortPriKeyOptGetSequencingNodesImpl(SLogicNode* pNode, bool* pNot
     return TSDB_CODE_SUCCESS;
   }
 
-  return sortPriKeyOptGetSequencingNodesImpl((SLogicNode*)nodesListGetNode(pNode->pChildren, 0), pNotOptimize,
+  return sortPriKeyOptGetSequencingNodesImpl((SLogicNode*)nodesListGetNode(pNode->pChildren, 0), groupSort,
-                                             pSequencingNodes);
+                                             pNotOptimize, pSequencingNodes);
 }
 
-static int32_t sortPriKeyOptGetSequencingNodes(SLogicNode* pNode, SNodeList** pSequencingNodes) {
+static int32_t sortPriKeyOptGetSequencingNodes(SLogicNode* pNode, bool groupSort, SNodeList** pSequencingNodes) {
   bool    notOptimize = false;
-  int32_t code = sortPriKeyOptGetSequencingNodesImpl(pNode, &notOptimize, pSequencingNodes);
+  int32_t code = sortPriKeyOptGetSequencingNodesImpl(pNode, groupSort, &notOptimize, pSequencingNodes);
   if (TSDB_CODE_SUCCESS != code || notOptimize) {
     NODES_CLEAR_LIST(*pSequencingNodes);
   }
@@ -1175,8 +1175,8 @@ static int32_t sortPriKeyOptApply(SOptimizeContext* pCxt, SLogicSubplan* pLogicS
 
 static int32_t sortPrimaryKeyOptimizeImpl(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan, SSortLogicNode* pSort) {
   SNodeList* pSequencingNodes = NULL;
-  int32_t    code =
+  int32_t    code = sortPriKeyOptGetSequencingNodes((SLogicNode*)nodesListGetNode(pSort->node.pChildren, 0),
-      sortPriKeyOptGetSequencingNodes((SLogicNode*)nodesListGetNode(pSort->node.pChildren, 0), &pSequencingNodes);
+                                                    pSort->groupSort, &pSequencingNodes);
   if (TSDB_CODE_SUCCESS == code && NULL != pSequencingNodes) {
     code = sortPriKeyOptApply(pCxt, pLogicSubplan, pSort, pSequencingNodes);
   }