other: merge main.

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 69eee2e
         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 5aa25e9
         SOURCE_DIR "${TD_SOURCE_DIR}/tools/taos-tools"
         BINARY_DIR ""
         #BUILD_IN_SOURCE TRUE

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/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,

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];                                                                                                     \
+  int32_t size = backtrace(array, STACKSIZE);                                                                                   \
+  char**  strings = backtrace_symbols(array, size);                                                                       \
+  int32_t offset = 0;                                                                                                     \
+  if (strings != NULL) {                                                                                                  \
+    offset = snprintf(buf, bufSize - 1, "obtained %d stack frames\n", (ignoreNum > 0) ? size - ignoreNum : size);           \
+    for (int32_t i = (ignoreNum > 0) ? ignoreNum : 0; i < size; i++) {                                                    \
+      offset += snprintf(buf + offset, bufSize - 1 - offset, "frame:%d, %s\n", (ignoreNum > 0) ? i - ignoreNum : i, strings[i]);     \
+    }                                                                                                                     \
+  }                                                                                                                       \
+                                                                                                                          \
+  taosMemoryFree(strings);                                                                                                \
+}
+
+#define taosPrintTrace(flags, level, dflag, ignoreNum)                                                          \
   {                                                                                                             \
-    void*   array[100];                                                    \
+    void*   array[STACKSIZE];                                                                                         \
-    int32_t size = backtrace(array, 100);                                  \
+    int32_t size = backtrace(array, STACKSIZE);                                                                       \
     char**  strings = backtrace_symbols(array, size);                                                           \
     if (strings != NULL) {                                                                                      \
-      taosPrintLog(flags, level, dflag, "obtained %d stack frames", size); \
+      taosPrintLog(flags, level, dflag, "obtained %d stack frames", (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]);  \
+        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/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

@@ -481,10 +481,10 @@ 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
-  fi
+  else
-
     [ -f ${script_dir}/cfg/${adapterName}.toml ] &&
       ${csudo}cp -f ${script_dir}/cfg/${adapterName}.toml ${cfg_install_dir}/${adapterName}.toml.new
+  fi
 
   [ -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

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
 
@@ -385,6 +388,150 @@ void destroyRequest(SRequestObj *pRequest) {
   removeRequest(pRequest->self);
 }
 
+void taosClientCrash(int signum, void *sigInfo, void *context) {
+  taosIgnSignal(SIGTERM);
+  taosIgnSignal(SIGHUP);
+  taosIgnSignal(SIGINT);
+  taosIgnSignal(SIGBREAK);
+
+#if !defined(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, lastClusterId, appInfo.startTime)) {
+      taosPrintLog(flags, level, dflag, "failed to generate crash json msg");
+      goto _return;
+    } else {
+      msgLen = strlen(pMsg);  
+    }
+  }
+  
+_return:
+
+  taosLogCrashInfo("taos", pMsg, msgLen, signum, sigInfo);
+
+#ifdef _TD_DARWIN_64
+  exit(signum);
+#elif defined(WINDOWS)
+  exit(signum);
+#endif
+}
+
+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);
+  }
+}
+
+static void tscSetSignalHandle() {
+#if !defined(WINDOWS)
+  taosSetSignal(SIGBUS, taosClientCrash);
+#endif
+  taosSetSignal(SIGABRT, taosClientCrash);
+  taosSetSignal(SIGFPE, taosClientCrash);
+  taosSetSignal(SIGSEGV, taosClientCrash);
+}
+
 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 +539,10 @@ 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) {
@@ -404,6 +555,8 @@ void taos_init_imp(void) {
     return;
   }
 
+  tscSetSignalHandle();
+
   initQueryModuleMsgHandle();
 
   if (taosConvInit() != 0) {
@@ -433,9 +586,8 @@ 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/clientMain.c

@@ -55,6 +55,8 @@ void taos_cleanup(void) {
     return;
   }
 
+  tscStopCrashReport();
+
   int32_t id = clientReqRefPool;
   clientReqRefPool = -1;
   taosCloseRef(id);

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/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/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/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/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");
-  if (selfIndex) {
-    if (selfIndex->type != cJSON_Number) {
-      dError("failed to read %s since selfIndex not found", file);
-      goto _OVER;
-    }
-    pOption->selfIndex = selfIndex->valueint;
-  }
-
-  cJSON *replicas = cJSON_GetObjectItem(root, "replicas");
-  if (replicas) {
-    if (replicas->type != cJSON_Array) {
-      dError("failed to read %s since replicas not found", file);
     goto _OVER;
   }
 
-    int32_t numOfReplicas = cJSON_GetArraySize(replicas);
+  pData[size] = '\0';
-    if (numOfReplicas <= 0) {
+
-      dError("failed to read %s since numOfReplicas:%d invalid", file, numOfReplicas);
+  pJson = tjsonParse(pData);
+  if (pJson == NULL) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
     goto _OVER;
   }
-    pOption->numOfReplicas = numOfReplicas;
 
-    for (int32_t i = 0; i < numOfReplicas; ++i) {
+  if (mmDecodeOption(pJson, pOption) < 0) {
-      SReplica *pReplica = pOption->replicas + i;
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
-
-      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) {
+  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%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

@@ -15,6 +15,7 @@
 
 #define _DEFAULT_SOURCE
 #include "vmInt.h"
+#include "tjson.h"
 
 #define MAX_CONTENT_LEN 2 * 1024 * 1024
 
@@ -144,65 +145,66 @@ _OVER:
   return code;
 }
 
-int32_t vmWriteVnodeListToFile(SVnodeMgmt *pMgmt) {
+static int32_t vmEncodeVnodeList(SJson *pJson, SVnodeObj **ppVnodes, int32_t numOfVnodes) {
-  int32_t code = 0;
+  SJson *vnodes = tjsonCreateArray();
-  char    file[PATH_MAX] = {0};
+  if (vnodes == NULL) return -1;
-  char    realfile[PATH_MAX] = {0};
+  if (tjsonAddItemToObject(pJson, "vnodes", vnodes) < 0) return -1;
-  snprintf(file, sizeof(file), "%s%svnodes.json.bak", pMgmt->path, TD_DIRSEP);
-  snprintf(realfile, sizeof(file), "%s%svnodes.json", pMgmt->path, TD_DIRSEP);
 
-  TdFilePtr pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  if (pFile == NULL) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to write %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;
-  int32_t maxLen = MAX_CONTENT_LEN;
-  char   *content = taosMemoryCalloc(1, maxLen + 1);
-  if (content == NULL) {
-    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");
-  len += snprintf(content + len, maxLen - len, "  \"vnodes\": [\n");
   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");
+  return 0;
-  len += snprintf(content + len, maxLen - len, "}\n");
+}
+
+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 +215,9 @@ _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/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);

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/src/dmEps.c

@@ -15,6 +15,7 @@
 
 #define _DEFAULT_SOURCE
 #include "dmUtil.h"
+#include "tjson.h"
 #include "tmisce.h"
 
 static void dmPrintEps(SDnodeData *pData);
@@ -181,81 +182,73 @@ _OVER:
   return code;
 }
 
+static int32_t dmEncodeEps(SJson *pJson, SDnodeData *pData) {
+  if (tjsonAddDoubleToObject(pJson, "dnodeId", pData->dnodeId) < 0) return -1;
+  if (tjsonAddIntegerToObject(pJson, "dnodeVer", pData->dnodeVer) < 0) return -1;
+  if (tjsonAddIntegerToObject(pJson, "clusterId", pData->clusterId) < 0) return -1;
+  if (tjsonAddDoubleToObject(pJson, "dropped", pData->dropped) < 0) return -1;
+
+  SJson *dnodes = tjsonCreateArray();
+  if (dnodes == NULL) return -1;
+  if (tjsonAddItemToObject(pJson, "dnodes", dnodes) < 0) return -1;
+
+  int32_t numOfEps = (int32_t)taosArrayGetSize(pData->dnodeEps);
+  for (int32_t i = 0; i < numOfEps; ++i) {
+    SDnodeEp *pDnodeEp = taosArrayGet(pData->dnodeEps, i);
+    SJson    *dnode = tjsonCreateObject();
+    if (dnode == NULL) return -1;
+
+    if (tjsonAddDoubleToObject(dnode, "id", pDnodeEp->id) < 0) return -1;
+    if (tjsonAddStringToObject(dnode, "fqdn", pDnodeEp->ep.fqdn) < 0) return -1;
+    if (tjsonAddDoubleToObject(dnode, "port", pDnodeEp->ep.port) < 0) return -1;
+    if (tjsonAddDoubleToObject(dnode, "isMnode", pDnodeEp->isMnode) < 0) return -1;
+    if (tjsonAddItemToArray(dnodes, dnode) < 0) return -1;
+  }
+
+  return 0;
+}
+
 int32_t dmWriteEps(SDnodeData *pData) {
   int32_t   code = -1;
-  char     *content = NULL;
+  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) {
+  if (pFile == NULL) goto _OVER;
-    dError("failed to open %s since %s", file, strerror(errno));
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    goto _OVER;
-  }
 
-  int32_t len = 0;
+  int32_t len = strlen(buffer);
-  int32_t maxLen = 256 * 1024;
+  if (taosWriteFile(pFile, buffer, len) <= 0) goto _OVER;
-  content = taosMemoryCalloc(1, maxLen + 1);
+  if (taosFsyncFile(pFile) < 0) goto _OVER;
-
-  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);
-    len += snprintf(content + len, maxLen - len, "    \"fqdn\": \"%s\",\n", pDnodeEp->ep.fqdn);
-    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) {
-    dError("failed to write %s since %s", file, strerror(errno));
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    goto _OVER;
-  }
-
-  if (taosFsyncFile(pFile) < 0) {
-    dError("failed to fsync %s since %s", file, strerror(errno));
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    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);
+    dInfo("succeed to write dnode file:%s since %s, dnodeVer:%" PRId64, realfile, terrstr(), pData->dnodeVer);
+  }
   return code;
 }
 

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

@@ -15,6 +15,7 @@
 
 #define _DEFAULT_SOURCE
 #include "dmUtil.h"
+#include "tjson.h"
 
 #define MAXLEN 1024
 
@@ -63,56 +64,51 @@ _OVER:
   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/mndDb.c

@@ -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

@@ -308,7 +308,8 @@ 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/mndMnode.c

@@ -785,9 +785,9 @@ 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/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/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/sdb/src/sdbFile.c

@@ -243,7 +243,7 @@ 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;
   }
 
@@ -636,15 +636,20 @@ 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 +658,12 @@ 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 +677,11 @@ 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/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

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/sma/smaCommit.c

@@ -395,6 +395,10 @@ static int32_t tdProcessRSmaAsyncPreCommitImpl(SSma *pSma) {
 static int32_t tdProcessRSmaAsyncCommitImpl(SSma *pSma, SCommitInfo *pInfo) {
   int32_t code = 0;
   SVnode *pVnode = pSma->pVnode;
+  SSmaEnv *pSmaEnv = SMA_RSMA_ENV(pSma);
+  if (!pSmaEnv) {
+    goto _exit;
+  }
 #if 0
   SRSmaStat *pRSmaStat = (SRSmaStat *)SMA_ENV_STAT(pSmaEnv);
 

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/tsdbSnapshot.c

@@ -192,6 +192,7 @@ static int32_t tsdbSnapNextRow(STsdbSnapReader* pReader) {
         int64_t rowVer = pIter->bData.aVersion[pIter->iRow];
 
         if (rowVer >= pReader->sver && rowVer <= pReader->ever) {
+          pIter->rInfo.suid = pIter->bData.suid;
           pIter->rInfo.uid = pIter->bData.uid ? pIter->bData.uid : pIter->bData.aUid[pIter->iRow];
           pIter->rInfo.row = tsdbRowFromBlockData(&pIter->bData, pIter->iRow);
           goto _out;

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;
@@ -720,13 +732,36 @@ int32_t tRowMergerAdd(SRowMerger *pMerger, TSDBROW *pRow, STSchema *pTSchema) {
 
     if (key.version > pMerger->version) {
       if (!COL_VAL_IS_NONE(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);
         }
+      }
     } else if (key.version < pMerger->version) {
       SColVal *tColVal = (SColVal *)taosArrayGet(pMerger->pArray, iCol);
       if (COL_VAL_IS_NONE(tColVal) && !COL_VAL_IS_NONE(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 +800,18 @@ 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 +822,16 @@ _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;
@@ -789,13 +845,48 @@ int32_t tRowMerge(SRowMerger *pMerger, TSDBROW *pRow) {
 
     if (key.version > pMerger->version) {
       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) 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)) {
+        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

@@ -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);
   }
 
+  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/src/ctgAsync.c

@@ -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);
@@ -1098,7 +1109,8 @@ _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);
@@ -1286,7 +1298,8 @@ _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/executor/src/scanoperator.c

@@ -1729,6 +1729,7 @@ 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;

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;
       }
 
@@ -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;
         }

source/libs/function/inc/tpercentile.h

@@ -73,7 +73,7 @@ void tMemBucketDestroy(tMemBucket *pBucket);
 
 int32_t tMemBucketPut(tMemBucket *pBucket, const void *data, size_t size);
 
-double getPercentile(tMemBucket *pMemBucket, double percent);
+int32_t getPercentile(tMemBucket *pMemBucket, double percent, double *result);
 
 #endif  // TDENGINE_TPERCENTILE_H
 

source/libs/function/src/builtinsimpl.c

@@ -1662,9 +1662,9 @@ int32_t percentileFunction(SqlFunctionCtx* pCtx) {
 
 int32_t percentileFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   SVariant* pVal = &pCtx->param[1].param;
-  terrno = 0;
+  int32_t code = 0;
-
   double    v = 0;
+
   GET_TYPED_DATA(v, double, pVal->nType, &pVal->i);
 
   SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx);
@@ -1672,13 +1672,12 @@ int32_t percentileFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
 
   tMemBucket* pMemBucket = ppInfo->pMemBucket;
   if (pMemBucket != NULL && pMemBucket->total > 0) {  // check for null
-    SET_DOUBLE_VAL(&ppInfo->result, getPercentile(pMemBucket, v));
+    code = getPercentile(pMemBucket, v, &ppInfo->result);
   }
 
   tMemBucketDestroy(pMemBucket);
-
+  if (code != TSDB_CODE_SUCCESS) {
-  if (terrno != TSDB_CODE_SUCCESS) {
+    return code;
-    return terrno;
   }
 
   return functionFinalize(pCtx, pBlock);

source/libs/function/src/tpercentile.c

@@ -90,7 +90,7 @@ static void resetPosInfo(SSlotInfo *pInfo) {
   pInfo->data = NULL;
 }
 
-double findOnlyResult(tMemBucket *pMemBucket) {
+int32_t findOnlyResult(tMemBucket *pMemBucket, double *result) {
   ASSERT(pMemBucket->total == 1);
   terrno = 0;
 
@@ -109,19 +109,17 @@ double findOnlyResult(tMemBucket *pMemBucket) {
       int32_t   *pageId = taosArrayGet(list, 0);
       SFilePage *pPage = getBufPage(pMemBucket->pBuffer, *pageId);
       if (pPage == NULL) {
-        terrno = TSDB_CODE_NO_AVAIL_DISK;
+        return TSDB_CODE_NO_AVAIL_DISK;
-        return 0;
       }
-
       ASSERT(pPage->num == 1);
 
-      double v = 0;
+      GET_TYPED_DATA(*result, double, pMemBucket->type, pPage->data);
-      GET_TYPED_DATA(v, double, pMemBucket->type, pPage->data);
+      return TSDB_CODE_SUCCESS;
-      return v;
     }
   }
 
-  return 0;
+  *result = 0.0;
+  return TSDB_CODE_SUCCESS;
 }
 
 int32_t tBucketIntHash(tMemBucket *pBucket, const void *value) {
@@ -443,7 +441,7 @@ static double getIdenticalDataVal(tMemBucket *pMemBucket, int32_t slotIndex) {
   return finalResult;
 }
 
-double getPercentileImpl(tMemBucket *pMemBucket, int32_t count, double fraction) {
+int32_t getPercentileImpl(tMemBucket *pMemBucket, int32_t count, double fraction, double *result) {
   int32_t num = 0;
 
   for (int32_t i = 0; i < pMemBucket->numOfSlots; ++i) {
@@ -476,15 +474,15 @@ double getPercentileImpl(tMemBucket *pMemBucket, int32_t count, double fraction)
 
         ASSERT(minOfNextSlot > maxOfThisSlot);
 
-        double val = (1 - fraction) * maxOfThisSlot + fraction * minOfNextSlot;
+        *result = (1 - fraction) * maxOfThisSlot + fraction * minOfNextSlot;
-        return val;
+        return TSDB_CODE_SUCCESS;
       }
 
       if (pSlot->info.size <= pMemBucket->maxCapacity) {
         // data in buffer and file are merged together to be processed.
         SFilePage *buffer = loadDataFromFilePage(pMemBucket, i);
         if (buffer == NULL) {
-          return -1;
+          return TSDB_CODE_NO_AVAIL_DISK;
         }
         int32_t    currentIdx = count - num;
 
@@ -495,13 +493,14 @@ double getPercentileImpl(tMemBucket *pMemBucket, int32_t count, double fraction)
         GET_TYPED_DATA(td, double, pMemBucket->type, thisVal);
         GET_TYPED_DATA(nd, double, pMemBucket->type, nextVal);
 
-        double val = (1 - fraction) * td + fraction * nd;
+        *result = (1 - fraction) * td + fraction * nd;
         taosMemoryFreeClear(buffer);
 
-        return val;
+        return TSDB_CODE_SUCCESS;
       } else {  // incur a second round bucket split
         if (isIdenticalData(pMemBucket, i)) {
-          return getIdenticalDataVal(pMemBucket, i);
+          *result = getIdenticalDataVal(pMemBucket, i);
+          return TSDB_CODE_SUCCESS;
         }
 
         // try next round
@@ -521,35 +520,37 @@ double getPercentileImpl(tMemBucket *pMemBucket, int32_t count, double fraction)
           int32_t *pageId = taosArrayGet(list, f);
           SFilePage *pg = getBufPage(pMemBucket->pBuffer, *pageId);
           if (pg == NULL) {
-            return -1;
+            return TSDB_CODE_NO_AVAIL_DISK;
           }
 
           int32_t code = tMemBucketPut(pMemBucket, pg->data, (int32_t)pg->num);
           if (code != TSDB_CODE_SUCCESS) {
-            return -1;
+            return code;
           }
           setBufPageDirty(pg, true);
           releaseBufPage(pMemBucket->pBuffer, pg);
         }
 
-        return getPercentileImpl(pMemBucket, count - num, fraction);
+        return getPercentileImpl(pMemBucket, count - num, fraction, result);
       }
     } else {
       num += pSlot->info.size;
     }
   }
 
-  return 0;
+  *result = 0;
+  return TSDB_CODE_SUCCESS;
 }
 
-double getPercentile(tMemBucket *pMemBucket, double percent) {
+int32_t getPercentile(tMemBucket *pMemBucket, double percent, double *result) {
   if (pMemBucket->total == 0) {
-    return 0.0;
+    *result = 0.0;
+    return TSDB_CODE_SUCCESS;
   }
 
   // if only one elements exists, return it
   if (pMemBucket->total == 1) {
-    return findOnlyResult(pMemBucket);
+    return findOnlyResult(pMemBucket, result);
   }
 
   percent = fabs(percent);
@@ -559,21 +560,21 @@ double getPercentile(tMemBucket *pMemBucket, double percent) {
     MinMaxEntry *pRange = &pMemBucket->range;
 
     if (IS_SIGNED_NUMERIC_TYPE(pMemBucket->type)) {
-      double v = (double)(fabs(percent - 100) < DBL_EPSILON ? pRange->i64MaxVal : pRange->i64MinVal);
+      *result = (double)(fabs(percent - 100) < DBL_EPSILON ? pRange->i64MaxVal : pRange->i64MinVal);
-      return v;
     } else if (IS_UNSIGNED_NUMERIC_TYPE(pMemBucket->type)) {
-      double v = (double)(fabs(percent - 100) < DBL_EPSILON ? pRange->u64MaxVal : pRange->u64MinVal);
+      *result = (double)(fabs(percent - 100) < DBL_EPSILON ? pRange->u64MaxVal : pRange->u64MinVal);
-      return v;
     } else {
-      return fabs(percent - 100) < DBL_EPSILON ? pRange->dMaxVal : pRange->dMinVal;
+      *result = fabs(percent - 100) < DBL_EPSILON ? pRange->dMaxVal : pRange->dMinVal;
     }
+
+    return TSDB_CODE_SUCCESS;
   }
 
   double percentVal = (percent * (pMemBucket->total - 1)) / ((double)100.0);
 
   // do put data by using buckets
   int32_t orderIdx = (int32_t)percentVal;
-  return getPercentileImpl(pMemBucket, orderIdx, percentVal - orderIdx);
+  return getPercentileImpl(pMemBucket, orderIdx, percentVal - orderIdx, result);
 }
 
 /*

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/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/stream/src/streamMeta.c

@@ -69,8 +69,7 @@ _err:
 }
 
 void streamMetaClose(SStreamMeta* pMeta) {
-  tdbCommit(pMeta->db, pMeta->txn);
+  tdbAbort(pMeta->db, pMeta->txn);
-  tdbPostCommit(pMeta->db, pMeta->txn);
   tdbTbClose(pMeta->pTaskDb);
   tdbTbClose(pMeta->pCheckpointDb);
   tdbClose(pMeta->db);
@@ -88,6 +87,7 @@ void streamMetaClose(SStreamMeta* pMeta) {
     /*streamMetaReleaseTask(pMeta, pTask);*/
   }
   taosHashCleanup(pMeta->pTasks);
+  taosHashCleanup(pMeta->pRecoverStatus);
   taosMemoryFree(pMeta->path);
   taosMemoryFree(pMeta);
 }

source/libs/sync/inc/syncInt.h

@@ -32,11 +32,9 @@ typedef struct SyncRequestVoteReply   SyncRequestVoteReply;
 typedef struct SyncAppendEntries      SyncAppendEntries;
 typedef struct SyncAppendEntriesReply SyncAppendEntriesReply;
 typedef struct SSyncEnv               SSyncEnv;
-typedef struct SRaftStore             SRaftStore;
 typedef struct SVotesGranted          SVotesGranted;
 typedef struct SVotesRespond          SVotesRespond;
 typedef struct SSyncIndexMgr          SSyncIndexMgr;
-typedef struct SRaftCfg               SRaftCfg;
 typedef struct SSyncRespMgr           SSyncRespMgr;
 typedef struct SSyncSnapshotSender    SSyncSnapshotSender;
 typedef struct SSyncSnapshotReceiver  SSyncSnapshotReceiver;
@@ -70,6 +68,11 @@ typedef struct SRaftId {
   SyncGroupId vgId;
 } SRaftId;
 
+typedef struct SRaftStore {
+  SyncTerm currentTerm;
+  SRaftId  voteFor;
+} SRaftStore;
+
 typedef struct SSyncHbTimerData {
   int64_t     syncNodeRid;
   SSyncTimer* pTimer;
@@ -140,7 +143,7 @@ typedef struct SSyncNode {
 
   // tla+ server vars
   ESyncState state;
-  SRaftStore* pRaftStore;
+  SRaftStore raftStore;
 
   // tla+ candidate vars
   SVotesGranted* pVotesGranted;
@@ -229,7 +232,7 @@ int32_t    syncNodeStartStandBy(SSyncNode* pSyncNode);
 void       syncNodeClose(SSyncNode* pSyncNode);
 void       syncNodePreClose(SSyncNode* pSyncNode);
 void       syncNodePostClose(SSyncNode* pSyncNode);
-int32_t    syncNodePropose(SSyncNode* pSyncNode, SRpcMsg* pMsg, bool isWeak, int64_t *seq);
+int32_t    syncNodePropose(SSyncNode* pSyncNode, SRpcMsg* pMsg, bool isWeak, int64_t* seq);
 int32_t    syncNodeRestore(SSyncNode* pSyncNode);
 void       syncHbTimerDataFree(SSyncHbTimerData* pData);
 

source/libs/sync/inc/syncMessage.h

@@ -247,8 +247,8 @@ typedef struct SyncLocalCmd {
   SRaftId  destId;
 
   int32_t   cmd;
-  SyncTerm  sdNewTerm;  // step down new term
+  SyncTerm  currentTerm;  // step down new term
-  SyncIndex fcIndex;    // follower commit index
+  SyncIndex commitIndex;  // follower commit index
 } SyncLocalCmd;
 
 int32_t syncBuildTimeout(SRpcMsg* pMsg, ESyncTimeoutType ttype, uint64_t logicClock, int32_t ms, SSyncNode* pNode);
@@ -258,7 +258,7 @@ int32_t syncBuildRequestVote(SRpcMsg* pMsg, int32_t vgId);
 int32_t syncBuildRequestVoteReply(SRpcMsg* pMsg, int32_t vgId);
 int32_t syncBuildAppendEntries(SRpcMsg* pMsg, int32_t dataLen, int32_t vgId);
 int32_t syncBuildAppendEntriesReply(SRpcMsg* pMsg, int32_t vgId);
-int32_t syncBuildAppendEntriesFromRaftLog(SSyncNode* pNode, SSyncRaftEntry* pEntry, SyncTerm prevLogTerm,
+int32_t syncBuildAppendEntriesFromRaftEntry(SSyncNode* pNode, SSyncRaftEntry* pEntry, SyncTerm prevLogTerm,
                                             SRpcMsg* pRpcMsg);
 int32_t syncBuildHeartbeat(SRpcMsg* pMsg, int32_t vgId);
 int32_t syncBuildHeartbeatReply(SRpcMsg* pMsg, int32_t vgId);

source/libs/sync/inc/syncPipeline.h

@@ -78,14 +78,14 @@ static FORCE_INLINE int32_t syncLogGetNextRetryBackoff(SSyncLogReplMgr* pMgr) {
 SyncTerm syncLogReplMgrGetPrevLogTerm(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index);
 
 int32_t  syncLogReplMgrReplicateOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode);
-int32_t  syncLogBufferReplicateOneTo(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index, SyncTerm* pTerm,
+int32_t  syncLogReplMgrReplicateOneTo(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index, SyncTerm* pTerm,
                                       SRaftId* pDestId, bool* pBarrier);
-int32_t  syncLogReplMgrReplicateAttemptedOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode);
+int32_t  syncLogReplMgrReplicateAttempt(SSyncLogReplMgr* pMgr, SSyncNode* pNode);
-int32_t  syncLogReplMgrReplicateProbeOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index);
+int32_t  syncLogReplMgrReplicateProbe(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index);
 
 int32_t  syncLogReplMgrProcessReply(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg);
-int32_t syncLogReplMgrProcessReplyInRecoveryMode(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg);
+int32_t  syncLogReplMgrProcessReplyAsRecovery(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg);
-int32_t  syncLogReplMgrProcessReplyInNormalMode(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg);
+int32_t  syncLogReplMgrProcessReplyAsNormal(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg);
 
 int32_t syncLogReplMgrProcessHeartbeatReply(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncHeartbeatReply* pMsg);
 int32_t  syncLogReplMgrRetryOnNeed(SSyncLogReplMgr* pMgr, SSyncNode* pNode);
@@ -98,6 +98,7 @@ int32_t         syncLogBufferReInit(SSyncLogBuffer* pBuf, SSyncNode* pNode);
 
 // access
 int64_t syncLogBufferGetEndIndex(SSyncLogBuffer* pBuf);
+SyncTerm syncLogBufferGetLastMatchTerm(SSyncLogBuffer* pBuf);
 int32_t syncLogBufferAppend(SSyncLogBuffer* pBuf, SSyncNode* pNode, SSyncRaftEntry* pEntry);
 int32_t syncLogBufferAccept(SSyncLogBuffer* pBuf, SSyncNode* pNode, SSyncRaftEntry* pEntry, SyncTerm prevTerm);
 int64_t syncLogBufferProceed(SSyncLogBuffer* pBuf, SSyncNode* pNode, SyncTerm* pMatchTerm);

source/libs/sync/inc/syncRaftStore.h

@@ -24,27 +24,16 @@ extern "C" {
 
 #define RAFT_STORE_BLOCK_SIZE 512
 #define RAFT_STORE_PATH_LEN   (TSDB_FILENAME_LEN * 2)
-
 #define EMPTY_RAFT_ID         ((SRaftId){.addr = 0, .vgId = 0})
 
-typedef struct SRaftStore {
+int32_t raftStoreReadFile(SSyncNode *pNode);
-  SyncTerm  currentTerm;
+int32_t raftStoreWriteFile(SSyncNode *pNode);
-  SRaftId   voteFor;
-  TdFilePtr pFile;
-  char      path[RAFT_STORE_PATH_LEN];
-} SRaftStore;
 
-SRaftStore *raftStoreOpen(const char *path);
+bool raftStoreHasVoted(SSyncNode *pNode);
-int32_t     raftStoreClose(SRaftStore *pRaftStore);
+void raftStoreVote(SSyncNode *pNode, SRaftId *pRaftId);
-int32_t     raftStorePersist(SRaftStore *pRaftStore);
+void raftStoreClearVote(SSyncNode *pNode);
-int32_t     raftStoreSerialize(SRaftStore *pRaftStore, char *buf, size_t len);
+void raftStoreNextTerm(SSyncNode *pNode);
-int32_t     raftStoreDeserialize(SRaftStore *pRaftStore, char *buf, size_t len);
+void raftStoreSetTerm(SSyncNode *pNode, SyncTerm term);
-
-bool raftStoreHasVoted(SRaftStore *pRaftStore);
-void raftStoreVote(SRaftStore *pRaftStore, SRaftId *pRaftId);
-void raftStoreClearVote(SRaftStore *pRaftStore);
-void raftStoreNextTerm(SRaftStore *pRaftStore);
-void raftStoreSetTerm(SRaftStore *pRaftStore, SyncTerm term);
 
 #ifdef __cplusplus
 }

source/libs/sync/src/syncAppendEntries.c

@@ -90,6 +90,7 @@
 //
 
 int32_t syncNodeFollowerCommit(SSyncNode* ths, SyncIndex newCommitIndex) {
+  ASSERT(false && "deprecated");
   if (ths->state != TAOS_SYNC_STATE_FOLLOWER) {
     sNTrace(ths, "can not do follower commit");
     return -1;
@@ -127,7 +128,7 @@ int32_t syncNodeFollowerCommit(SSyncNode* ths, SyncIndex newCommitIndex) {
   return 0;
 }
 
-SSyncRaftEntry* syncLogAppendEntriesToRaftEntry(const SyncAppendEntries* pMsg) {
+SSyncRaftEntry* syncBuildRaftEntryFromAppendEntries(const SyncAppendEntries* pMsg) {
   SSyncRaftEntry* pEntry = taosMemoryMalloc(pMsg->dataLen);
   if (pEntry == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
@@ -158,17 +159,17 @@ int32_t syncNodeOnAppendEntries(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   // prepare response msg
   pReply->srcId = ths->myRaftId;
   pReply->destId = pMsg->srcId;
-  pReply->term = ths->pRaftStore->currentTerm;
+  pReply->term = ths->raftStore.currentTerm;
   pReply->success = false;
   pReply->matchIndex = SYNC_INDEX_INVALID;
   pReply->lastSendIndex = pMsg->prevLogIndex + 1;
   pReply->startTime = ths->startTime;
 
-  if (pMsg->term < ths->pRaftStore->currentTerm) {
+  if (pMsg->term < ths->raftStore.currentTerm) {
     goto _SEND_RESPONSE;
   }
 
-  if (pMsg->term > ths->pRaftStore->currentTerm) {
+  if (pMsg->term > ths->raftStore.currentTerm) {
     pReply->term = pMsg->term;
   }
 
@@ -181,7 +182,7 @@ int32_t syncNodeOnAppendEntries(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
     goto _IGNORE;
   }
 
-  SSyncRaftEntry* pEntry = syncLogAppendEntriesToRaftEntry(pMsg);
+  SSyncRaftEntry* pEntry = syncBuildRaftEntryFromAppendEntries(pMsg);
 
   if (pEntry == NULL) {
     sError("vgId:%d, failed to get raft entry from append entries since %s", ths->vgId, terrstr());
@@ -206,12 +207,13 @@ int32_t syncNodeOnAppendEntries(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   accepted = true;
 
 _SEND_RESPONSE:
+  pEntry = NULL;
   pReply->matchIndex = syncLogBufferProceed(ths->pLogBuf, ths, &pReply->lastMatchTerm);
   bool matched = (pReply->matchIndex >= pReply->lastSendIndex);
   if (accepted && matched) {
     pReply->success = true;
     // update commit index only after matching
-    (void)syncNodeUpdateCommitIndex(ths, pMsg->commitIndex);
+    (void)syncNodeUpdateCommitIndex(ths, TMIN(pMsg->commitIndex, pReply->lastSendIndex));
   }
 
   // ack, i.e. send response
@@ -251,19 +253,19 @@ int32_t syncNodeOnAppendEntriesOld(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   SyncAppendEntriesReply* pReply = rpcRsp.pCont;
   pReply->srcId = ths->myRaftId;
   pReply->destId = pMsg->srcId;
-  pReply->term = ths->pRaftStore->currentTerm;
+  pReply->term = ths->raftStore.currentTerm;
   pReply->success = false;
   // pReply->matchIndex = ths->pLogStore->syncLogLastIndex(ths->pLogStore);
   pReply->matchIndex = SYNC_INDEX_INVALID;
   pReply->lastSendIndex = pMsg->prevLogIndex + 1;
   pReply->startTime = ths->startTime;
 
-  if (pMsg->term < ths->pRaftStore->currentTerm) {
+  if (pMsg->term < ths->raftStore.currentTerm) {
     syncLogRecvAppendEntries(ths, pMsg, "reject, small term");
     goto _SEND_RESPONSE;
   }
 
-  if (pMsg->term > ths->pRaftStore->currentTerm) {
+  if (pMsg->term > ths->raftStore.currentTerm) {
     pReply->term = pMsg->term;
   }
 

source/libs/sync/src/syncAppendEntriesReply.c

@@ -50,19 +50,19 @@ int32_t syncNodeOnAppendEntriesReply(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   }
 
   // drop stale response
-  if (pMsg->term < ths->pRaftStore->currentTerm) {
+  if (pMsg->term < ths->raftStore.currentTerm) {
     syncLogRecvAppendEntriesReply(ths, pMsg, "drop stale response");
     return 0;
   }
 
   if (ths->state == TAOS_SYNC_STATE_LEADER) {
-    if (pMsg->term > ths->pRaftStore->currentTerm) {
+    if (pMsg->term > ths->raftStore.currentTerm) {
       syncLogRecvAppendEntriesReply(ths, pMsg, "error term");
       syncNodeStepDown(ths, pMsg->term);
       return -1;
     }
 
-    ASSERT(pMsg->term == ths->pRaftStore->currentTerm);
+    ASSERT(pMsg->term == ths->raftStore.currentTerm);
 
     sTrace("vgId:%d, received append entries reply. srcId:0x%016" PRIx64 ",  term:%" PRId64 ", matchIndex:%" PRId64 "",
            pMsg->vgId, pMsg->srcId.addr, pMsg->term, pMsg->matchIndex);
@@ -100,19 +100,19 @@ int32_t syncNodeOnAppendEntriesReplyOld(SSyncNode* ths, SyncAppendEntriesReply*
   }
 
   // drop stale response
-  if (pMsg->term < ths->pRaftStore->currentTerm) {
+  if (pMsg->term < ths->raftStore.currentTerm) {
     syncLogRecvAppendEntriesReply(ths, pMsg, "drop stale response");
     return 0;
   }
 
   if (ths->state == TAOS_SYNC_STATE_LEADER) {
-    if (pMsg->term > ths->pRaftStore->currentTerm) {
+    if (pMsg->term > ths->raftStore.currentTerm) {
       syncLogRecvAppendEntriesReply(ths, pMsg, "error term");
       syncNodeStepDown(ths, pMsg->term);
       return -1;
     }
 
-    ASSERT(pMsg->term == ths->pRaftStore->currentTerm);
+    ASSERT(pMsg->term == ths->raftStore.currentTerm);
 
     if (pMsg->success) {
       SyncIndex oldMatchIndex = syncIndexMgrGetIndex(ths->pMatchIndex, &(pMsg->srcId));

source/libs/sync/src/syncCommit.c

@@ -44,6 +44,7 @@
 //     /\ UNCHANGED <<messages, serverVars, candidateVars, leaderVars, log>>
 //
 void syncOneReplicaAdvance(SSyncNode* pSyncNode) {
+  ASSERT(false && "deprecated");
   if (pSyncNode == NULL) {
     sError("pSyncNode is NULL");
     return;
@@ -132,7 +133,7 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
         }
       }
       // cannot commit, even if quorum agree. need check term!
-      if (pEntry->term <= pSyncNode->pRaftStore->currentTerm) {
+      if (pEntry->term <= pSyncNode->raftStore.currentTerm) {
         // update commit index
         newCommitIndex = index;
 
@@ -328,7 +329,7 @@ int64_t syncNodeCheckCommitIndex(SSyncNode* ths, SyncIndex indexLikely) {
     SyncIndex commitIndex = indexLikely;
     syncNodeUpdateCommitIndex(ths, commitIndex);
     sTrace("vgId:%d, agreed upon. role:%d, term:%" PRId64 ", index: %" PRId64 "", ths->vgId, ths->state,
-           ths->pRaftStore->currentTerm, commitIndex);
+           ths->raftStore.currentTerm, commitIndex);
   }
   return ths->commitIndex;
 }

source/libs/sync/src/syncElection.c

@@ -48,7 +48,7 @@ static int32_t syncNodeRequestVotePeers(SSyncNode* pNode) {
     SyncRequestVote* pMsg = rpcMsg.pCont;
     pMsg->srcId = pNode->myRaftId;
     pMsg->destId = pNode->peersId[i];
-    pMsg->term = pNode->pRaftStore->currentTerm;
+    pMsg->term = pNode->raftStore.currentTerm;
 
     ret = syncNodeGetLastIndexTerm(pNode, &pMsg->lastLogIndex, &pMsg->lastLogTerm);
     ASSERT(ret == 0);
@@ -75,10 +75,10 @@ int32_t syncNodeElect(SSyncNode* pSyncNode) {
   }
 
   // start election
-  raftStoreNextTerm(pSyncNode->pRaftStore);
+  raftStoreNextTerm(pSyncNode);
-  raftStoreClearVote(pSyncNode->pRaftStore);
+  raftStoreClearVote(pSyncNode);
-  voteGrantedReset(pSyncNode->pVotesGranted, pSyncNode->pRaftStore->currentTerm);
+  voteGrantedReset(pSyncNode->pVotesGranted, pSyncNode->raftStore.currentTerm);
-  votesRespondReset(pSyncNode->pVotesRespond, pSyncNode->pRaftStore->currentTerm);
+  votesRespondReset(pSyncNode->pVotesRespond, pSyncNode->raftStore.currentTerm);
 
   syncNodeVoteForSelf(pSyncNode);
   if (voteGrantedMajority(pSyncNode->pVotesGranted)) {

source/libs/sync/src/syncEnv.c

@@ -114,7 +114,7 @@ void syncHbTimerDataRemove(int64_t rid) { taosRemoveRef(gHbDataRefId, rid); }
 
 SSyncHbTimerData *syncHbTimerDataAcquire(int64_t rid) {
   SSyncHbTimerData *pData = taosAcquireRef(gHbDataRefId, rid);
-  if (pData == NULL) {
+  if (pData == NULL && rid > 0) {
     sInfo("failed to acquire hb-timer-data from refId:%" PRId64, rid);
     terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
   }

source/libs/sync/src/syncMain.c

@@ -468,7 +468,7 @@ bool syncNodeIsReadyForRead(SSyncNode* pSyncNode) {
       }
 
       if (code == 0 && pEntry != NULL) {
-        if (pEntry->originalRpcType == TDMT_SYNC_NOOP && pEntry->term == pSyncNode->pRaftStore->currentTerm) {
+        if (pEntry->originalRpcType == TDMT_SYNC_NOOP && pEntry->term == pSyncNode->raftStore.currentTerm) {
           ready = true;
         }
 
@@ -736,7 +736,7 @@ int32_t syncNodePropose(SSyncNode* pSyncNode, SRpcMsg* pMsg, bool isWeak, int64_
     int32_t   code = syncNodeOnClientRequest(pSyncNode, pMsg, &retIndex);
     if (code == 0) {
       pMsg->info.conn.applyIndex = retIndex;
-      pMsg->info.conn.applyTerm = pSyncNode->pRaftStore->currentTerm;
+      pMsg->info.conn.applyTerm = pSyncNode->raftStore.currentTerm;
       sTrace("vgId:%d, propose optimized msg, index:%" PRId64 " type:%s", pSyncNode->vgId, retIndex,
              TMSG_INFO(pMsg->msgType));
       return 1;
@@ -983,8 +983,7 @@ SSyncNode* syncNodeOpen(SSyncInfo* pSyncInfo) {
 
   // init TLA+ server vars
   pSyncNode->state = TAOS_SYNC_STATE_FOLLOWER;
-  pSyncNode->pRaftStore = raftStoreOpen(pSyncNode->raftStorePath);
+  if (raftStoreReadFile(pSyncNode) != 0) {
-  if (pSyncNode->pRaftStore == NULL) {
     sError("vgId:%d, failed to open raft store at path %s", pSyncNode->vgId, pSyncNode->raftStorePath);
     goto _error;
   }
@@ -1030,6 +1029,7 @@ SSyncNode* syncNodeOpen(SSyncInfo* pSyncInfo) {
     }
   }
   pSyncNode->commitIndex = commitIndex;
+  sInfo("vgId:%d, sync node commitIndex initialized as %" PRId64, pSyncNode->vgId, pSyncNode->commitIndex);
 
   if (syncNodeLogStoreRestoreOnNeed(pSyncNode) < 0) {
     goto _error;
@@ -1170,9 +1170,10 @@ int32_t syncNodeRestore(SSyncNode* pSyncNode) {
   }
 
   ASSERT(endIndex == lastVer + 1);
-  commitIndex = TMAX(pSyncNode->commitIndex, commitIndex);
+  pSyncNode->commitIndex = TMAX(pSyncNode->commitIndex, commitIndex);
+  sInfo("vgId:%d, restore sync until commitIndex:%" PRId64, pSyncNode->vgId, pSyncNode->commitIndex);
 
-  if (syncLogBufferCommit(pSyncNode->pLogBuf, pSyncNode, commitIndex) < 0) {
+  if (syncLogBufferCommit(pSyncNode->pLogBuf, pSyncNode, pSyncNode->commitIndex) < 0) {
     return -1;
   }
 
@@ -1182,7 +1183,7 @@ int32_t syncNodeRestore(SSyncNode* pSyncNode) {
 int32_t syncNodeStart(SSyncNode* pSyncNode) {
   // start raft
   if (pSyncNode->replicaNum == 1) {
-    raftStoreNextTerm(pSyncNode->pRaftStore);
+    raftStoreNextTerm(pSyncNode);
     syncNodeBecomeLeader(pSyncNode, "one replica start");
 
     // Raft 3.6.2 Committing entries from previous terms
@@ -1200,7 +1201,7 @@ int32_t syncNodeStart(SSyncNode* pSyncNode) {
 void syncNodeStartOld(SSyncNode* pSyncNode) {
   // start raft
   if (pSyncNode->replicaNum == 1) {
-    raftStoreNextTerm(pSyncNode->pRaftStore);
+    raftStoreNextTerm(pSyncNode);
     syncNodeBecomeLeader(pSyncNode, "one replica start");
 
     // Raft 3.6.2 Committing entries from previous terms
@@ -1286,10 +1287,6 @@ void syncNodeClose(SSyncNode* pSyncNode) {
   if (pSyncNode == NULL) return;
   sNInfo(pSyncNode, "sync close, node:%p", pSyncNode);
 
-  int32_t ret = raftStoreClose(pSyncNode->pRaftStore);
-  ASSERT(ret == 0);
-  pSyncNode->pRaftStore = NULL;
-
   syncNodeLogReplMgrDestroy(pSyncNode);
   syncRespMgrDestroy(pSyncNode->pSyncRespMgr);
   pSyncNode->pSyncRespMgr = NULL;
@@ -1480,16 +1477,21 @@ int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pNode, SRpcMsg
     }
   }
 
+  int32_t code = -1;
   if (pNode->syncSendMSg != NULL && epSet != NULL) {
     syncUtilMsgHtoN(pMsg->pCont);
     pMsg->info.noResp = 1;
-    return pNode->syncSendMSg(epSet, pMsg);
+    code = pNode->syncSendMSg(epSet, pMsg);
-  } else {
+  }
-    sError("vgId:%d, sync send msg by id error, fp:%p epset:%p", pNode->vgId, pNode->syncSendMSg, epSet);
+
+  if (code < 0) {
+    sError("vgId:%d, sync send msg by id error, epset:%p dnode:%d addr:%" PRId64 " err:0x%x", pNode->vgId, epSet,
+           DID(destRaftId), destRaftId->addr, terrno);
     rpcFreeCont(pMsg->pCont);
     terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
-    return -1;
   }
+
+  return code;
 }
 
 inline bool syncNodeInConfig(SSyncNode* pNode, const SSyncCfg* pCfg) {
@@ -1707,39 +1709,39 @@ _END:
 
 // raft state change --------------
 void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term) {
-  if (term > pSyncNode->pRaftStore->currentTerm) {
+  if (term > pSyncNode->raftStore.currentTerm) {
-    raftStoreSetTerm(pSyncNode->pRaftStore, term);
+    raftStoreSetTerm(pSyncNode, term);
     char tmpBuf[64];
     snprintf(tmpBuf, sizeof(tmpBuf), "update term to %" PRId64, term);
     syncNodeBecomeFollower(pSyncNode, tmpBuf);
-    raftStoreClearVote(pSyncNode->pRaftStore);
+    raftStoreClearVote(pSyncNode);
   }
 }
 
 void syncNodeUpdateTermWithoutStepDown(SSyncNode* pSyncNode, SyncTerm term) {
-  if (term > pSyncNode->pRaftStore->currentTerm) {
+  if (term > pSyncNode->raftStore.currentTerm) {
-    raftStoreSetTerm(pSyncNode->pRaftStore, term);
+    raftStoreSetTerm(pSyncNode, term);
   }
 }
 
 void syncNodeStepDown(SSyncNode* pSyncNode, SyncTerm newTerm) {
-  if (pSyncNode->pRaftStore->currentTerm > newTerm) {
+  if (pSyncNode->raftStore.currentTerm > newTerm) {
     sNTrace(pSyncNode, "step down, ignore, new-term:%" PRId64 ", current-term:%" PRId64, newTerm,
-            pSyncNode->pRaftStore->currentTerm);
+            pSyncNode->raftStore.currentTerm);
     return;
   }
 
   do {
     sNTrace(pSyncNode, "step down, new-term:%" PRId64 ", current-term:%" PRId64, newTerm,
-            pSyncNode->pRaftStore->currentTerm);
+            pSyncNode->raftStore.currentTerm);
   } while (0);
 
-  if (pSyncNode->pRaftStore->currentTerm < newTerm) {
+  if (pSyncNode->raftStore.currentTerm < newTerm) {
-    raftStoreSetTerm(pSyncNode->pRaftStore, newTerm);
+    raftStoreSetTerm(pSyncNode, newTerm);
     char tmpBuf[64];
     snprintf(tmpBuf, sizeof(tmpBuf), "step down, update term to %" PRId64, newTerm);
     syncNodeBecomeFollower(pSyncNode, tmpBuf);
-    raftStoreClearVote(pSyncNode->pRaftStore);
+    raftStoreClearVote(pSyncNode);
 
   } else {
     if (pSyncNode->state != TAOS_SYNC_STATE_FOLLOWER) {
@@ -1897,7 +1899,7 @@ void syncNodeCandidate2Leader(SSyncNode* pSyncNode) {
   SyncIndex lastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
   ASSERT(lastIndex >= 0);
   sInfo("vgId:%d, become leader. term: %" PRId64 ", commit index: %" PRId64 ", last index: %" PRId64 "",
-        pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex, lastIndex);
+        pSyncNode->vgId, pSyncNode->raftStore.currentTerm, pSyncNode->commitIndex, lastIndex);
 }
 
 void syncNodeCandidate2LeaderOld(SSyncNode* pSyncNode) {
@@ -1930,7 +1932,7 @@ void syncNodeFollower2Candidate(SSyncNode* pSyncNode) {
   pSyncNode->state = TAOS_SYNC_STATE_CANDIDATE;
   SyncIndex lastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
   sInfo("vgId:%d, become candidate from follower. term: %" PRId64 ", commit index: %" PRId64 ", last index: %" PRId64,
-        pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex, lastIndex);
+        pSyncNode->vgId, pSyncNode->raftStore.currentTerm, pSyncNode->commitIndex, lastIndex);
 
   sNTrace(pSyncNode, "follower to candidate");
 }
@@ -1940,7 +1942,7 @@ void syncNodeLeader2Follower(SSyncNode* pSyncNode) {
   syncNodeBecomeFollower(pSyncNode, "leader to follower");
   SyncIndex lastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
   sInfo("vgId:%d, become follower from leader. term: %" PRId64 ", commit index: %" PRId64 ", last index: %" PRId64,
-        pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex, lastIndex);
+        pSyncNode->vgId, pSyncNode->raftStore.currentTerm, pSyncNode->commitIndex, lastIndex);
 
   sNTrace(pSyncNode, "leader to follower");
 }
@@ -1950,7 +1952,7 @@ void syncNodeCandidate2Follower(SSyncNode* pSyncNode) {
   syncNodeBecomeFollower(pSyncNode, "candidate to follower");
   SyncIndex lastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
   sInfo("vgId:%d, become follower from candidate. term: %" PRId64 ", commit index: %" PRId64 ", last index: %" PRId64,
-        pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex, lastIndex);
+        pSyncNode->vgId, pSyncNode->raftStore.currentTerm, pSyncNode->commitIndex, lastIndex);
 
   sNTrace(pSyncNode, "candidate to follower");
 }
@@ -1958,15 +1960,15 @@ void syncNodeCandidate2Follower(SSyncNode* pSyncNode) {
 // just called by syncNodeVoteForSelf
 // need assert
 void syncNodeVoteForTerm(SSyncNode* pSyncNode, SyncTerm term, SRaftId* pRaftId) {
-  ASSERT(term == pSyncNode->pRaftStore->currentTerm);
+  ASSERT(term == pSyncNode->raftStore.currentTerm);
-  ASSERT(!raftStoreHasVoted(pSyncNode->pRaftStore));
+  ASSERT(!raftStoreHasVoted(pSyncNode));
 
-  raftStoreVote(pSyncNode->pRaftStore, pRaftId);
+  raftStoreVote(pSyncNode, pRaftId);
 }
 
 // simulate get vote from outside
 void syncNodeVoteForSelf(SSyncNode* pSyncNode) {
-  syncNodeVoteForTerm(pSyncNode, pSyncNode->pRaftStore->currentTerm, &pSyncNode->myRaftId);
+  syncNodeVoteForTerm(pSyncNode, pSyncNode->raftStore.currentTerm, &pSyncNode->myRaftId);
 
   SRpcMsg rpcMsg = {0};
   int32_t ret = syncBuildRequestVoteReply(&rpcMsg, pSyncNode->vgId);
@@ -1975,7 +1977,7 @@ void syncNodeVoteForSelf(SSyncNode* pSyncNode) {
   SyncRequestVoteReply* pMsg = rpcMsg.pCont;
   pMsg->srcId = pSyncNode->myRaftId;
   pMsg->destId = pSyncNode->myRaftId;
-  pMsg->term = pSyncNode->pRaftStore->currentTerm;
+  pMsg->term = pSyncNode->raftStore.currentTerm;
   pMsg->voteGranted = true;
 
   voteGrantedVote(pSyncNode->pVotesGranted, pMsg);
@@ -2265,13 +2267,6 @@ static void syncNodeEqPeerHeartbeatTimer(void* param, void* tmrId) {
     return;
   }
 
-  if (pSyncNode->pRaftStore == NULL) {
-    syncNodeRelease(pSyncNode);
-    syncHbTimerDataRelease(pData);
-    sError("vgId:%d, hb timer raft store already stop", pSyncNode->vgId);
-    return;
-  }
-
   // sTrace("vgId:%d, eq peer hb timer", pSyncNode->vgId);
 
   if (pSyncNode->replicaNum > 1) {
@@ -2295,7 +2290,7 @@ static void syncNodeEqPeerHeartbeatTimer(void* param, void* tmrId) {
         SyncHeartbeat* pSyncMsg = rpcMsg.pCont;
         pSyncMsg->srcId = pSyncNode->myRaftId;
         pSyncMsg->destId = pData->destId;
-        pSyncMsg->term = pSyncNode->pRaftStore->currentTerm;
+        pSyncMsg->term = pSyncNode->raftStore.currentTerm;
         pSyncMsg->commitIndex = pSyncNode->commitIndex;
         pSyncMsg->minMatchIndex = syncMinMatchIndex(pSyncNode);
         pSyncMsg->privateTerm = 0;
@@ -2341,7 +2336,7 @@ static int32_t syncNodeEqNoop(SSyncNode* pNode) {
   }
 
   SyncIndex       index = pNode->pLogStore->syncLogWriteIndex(pNode->pLogStore);
-  SyncTerm        term = pNode->pRaftStore->currentTerm;
+  SyncTerm        term = pNode->raftStore.currentTerm;
   SSyncRaftEntry* pEntry = syncEntryBuildNoop(term, index, pNode->vgId);
   if (pEntry == NULL) return -1;
 
@@ -2387,8 +2382,7 @@ int32_t syncNodeAppend(SSyncNode* ths, SSyncRaftEntry* pEntry) {
   if (syncLogBufferAppend(ths->pLogBuf, ths, pEntry) < 0) {
     sError("vgId:%d, failed to enqueue sync log buffer, index:%" PRId64, ths->vgId, pEntry->index);
     terrno = TSDB_CODE_SYN_BUFFER_FULL;
-    (void)syncLogFsmExecute(ths, ths->pFsm, ths->state, ths->pRaftStore->currentTerm, pEntry,
+    (void)syncLogFsmExecute(ths, ths->pFsm, ths->state, ths->raftStore.currentTerm, pEntry, TSDB_CODE_SYN_BUFFER_FULL);
-                            TSDB_CODE_SYN_BUFFER_FULL);
     syncEntryDestroy(pEntry);
     return -1;
   }
@@ -2461,7 +2455,7 @@ bool syncNodeSnapshotRecving(SSyncNode* pSyncNode) {
 
 static int32_t syncNodeAppendNoop(SSyncNode* ths) {
   SyncIndex index = syncLogBufferGetEndIndex(ths->pLogBuf);
-  SyncTerm  term = ths->pRaftStore->currentTerm;
+  SyncTerm  term = ths->raftStore.currentTerm;
 
   SSyncRaftEntry* pEntry = syncEntryBuildNoop(term, index, ths->vgId);
   if (pEntry == NULL) {
@@ -2477,7 +2471,7 @@ static int32_t syncNodeAppendNoopOld(SSyncNode* ths) {
   int32_t ret = 0;
 
   SyncIndex       index = ths->pLogStore->syncLogWriteIndex(ths->pLogStore);
-  SyncTerm        term = ths->pRaftStore->currentTerm;
+  SyncTerm        term = ths->raftStore.currentTerm;
   SSyncRaftEntry* pEntry = syncEntryBuildNoop(term, index, ths->vgId);
   ASSERT(pEntry != NULL);
 
@@ -2519,12 +2513,12 @@ int32_t syncNodeOnHeartbeat(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   SyncHeartbeatReply* pMsgReply = rpcMsg.pCont;
   pMsgReply->destId = pMsg->srcId;
   pMsgReply->srcId = ths->myRaftId;
-  pMsgReply->term = ths->pRaftStore->currentTerm;
+  pMsgReply->term = ths->raftStore.currentTerm;
   pMsgReply->privateTerm = 8864;  // magic number
   pMsgReply->startTime = ths->startTime;
   pMsgReply->timeStamp = tsMs;
 
-  if (pMsg->term == ths->pRaftStore->currentTerm && ths->state != TAOS_SYNC_STATE_LEADER) {
+  if (pMsg->term == ths->raftStore.currentTerm && ths->state != TAOS_SYNC_STATE_LEADER) {
     syncIndexMgrSetRecvTime(ths->pNextIndex, &(pMsg->srcId), tsMs);
 
     syncNodeResetElectTimer(ths);
@@ -2537,8 +2531,9 @@ int32_t syncNodeOnHeartbeat(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
 
       SyncLocalCmd* pSyncMsg = rpcMsgLocalCmd.pCont;
       pSyncMsg->cmd = SYNC_LOCAL_CMD_FOLLOWER_CMT;
-      pSyncMsg->fcIndex = pMsg->commitIndex;
+      pSyncMsg->commitIndex = pMsg->commitIndex;
-      SyncIndex fcIndex = pSyncMsg->fcIndex;
+      pSyncMsg->currentTerm = pMsg->term;
+      SyncIndex fcIndex = pSyncMsg->commitIndex;
 
       if (ths->syncEqMsg != NULL && ths->msgcb != NULL) {
         int32_t code = ths->syncEqMsg(ths->msgcb, &rpcMsgLocalCmd);
@@ -2552,14 +2547,15 @@ int32_t syncNodeOnHeartbeat(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
     }
   }
 
-  if (pMsg->term >= ths->pRaftStore->currentTerm && ths->state != TAOS_SYNC_STATE_FOLLOWER) {
+  if (pMsg->term >= ths->raftStore.currentTerm && ths->state != TAOS_SYNC_STATE_FOLLOWER) {
     // syncNodeStepDown(ths, pMsg->term);
     SRpcMsg rpcMsgLocalCmd = {0};
     (void)syncBuildLocalCmd(&rpcMsgLocalCmd, ths->vgId);
 
     SyncLocalCmd* pSyncMsg = rpcMsgLocalCmd.pCont;
     pSyncMsg->cmd = SYNC_LOCAL_CMD_STEP_DOWN;
-    pSyncMsg->sdNewTerm = pMsg->term;
+    pSyncMsg->currentTerm = pMsg->term;
+    pSyncMsg->commitIndex = pMsg->commitIndex;
 
     if (ths->syncEqMsg != NULL && ths->msgcb != NULL) {
       int32_t code = ths->syncEqMsg(ths->msgcb, &rpcMsgLocalCmd);
@@ -2567,7 +2563,7 @@ int32_t syncNodeOnHeartbeat(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
         sError("vgId:%d, sync enqueue step-down msg error, code:%d", ths->vgId, code);
         rpcFreeCont(rpcMsgLocalCmd.pCont);
       } else {
-        sTrace("vgId:%d, sync enqueue step-down msg, new-term: %" PRId64, ths->vgId, pSyncMsg->sdNewTerm);
+        sTrace("vgId:%d, sync enqueue step-down msg, new-term: %" PRId64, ths->vgId, pSyncMsg->currentTerm);
       }
     }
   }
@@ -2625,10 +2621,13 @@ int32_t syncNodeOnLocalCmd(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   syncLogRecvLocalCmd(ths, pMsg, "");
 
   if (pMsg->cmd == SYNC_LOCAL_CMD_STEP_DOWN) {
-    syncNodeStepDown(ths, pMsg->sdNewTerm);
+    syncNodeStepDown(ths, pMsg->currentTerm);
 
   } else if (pMsg->cmd == SYNC_LOCAL_CMD_FOLLOWER_CMT) {
-    (void)syncNodeUpdateCommitIndex(ths, pMsg->fcIndex);
+    SyncTerm matchTerm = syncLogBufferGetLastMatchTerm(ths->pLogBuf);
+    if (pMsg->currentTerm == matchTerm) {
+      (void)syncNodeUpdateCommitIndex(ths, pMsg->commitIndex);
+    }
     if (syncLogBufferCommit(ths->pLogBuf, ths, ths->commitIndex) < 0) {
       sError("vgId:%d, failed to commit raft log since %s. commit index: %" PRId64 "", ths->vgId, terrstr(),
              ths->commitIndex);
@@ -2641,14 +2640,15 @@ int32_t syncNodeOnLocalCmd(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
 }
 
 int32_t syncNodeOnLocalCmdOld(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
+  ASSERT(false && "deprecated");
   SyncLocalCmd* pMsg = pRpcMsg->pCont;
   syncLogRecvLocalCmd(ths, pMsg, "");
 
   if (pMsg->cmd == SYNC_LOCAL_CMD_STEP_DOWN) {
-    syncNodeStepDown(ths, pMsg->sdNewTerm);
+    syncNodeStepDown(ths, pMsg->currentTerm);
 
   } else if (pMsg->cmd == SYNC_LOCAL_CMD_FOLLOWER_CMT) {
-    syncNodeFollowerCommit(ths, pMsg->fcIndex);
+    syncNodeFollowerCommit(ths, pMsg->commitIndex);
 
   } else {
     sError("error local cmd");
@@ -2674,7 +2674,7 @@ int32_t syncNodeOnClientRequest(SSyncNode* ths, SRpcMsg* pMsg, SyncIndex* pRetIn
   int32_t code = 0;
 
   SyncIndex       index = syncLogBufferGetEndIndex(ths->pLogBuf);
-  SyncTerm        term = ths->pRaftStore->currentTerm;
+  SyncTerm        term = ths->raftStore.currentTerm;
   SSyncRaftEntry* pEntry = NULL;
   if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
     pEntry = syncEntryBuildFromClientRequest(pMsg->pCont, term, index);
@@ -2708,7 +2708,7 @@ int32_t syncNodeOnClientRequestOld(SSyncNode* ths, SRpcMsg* pMsg, SyncIndex* pRe
   int32_t code = 0;
 
   SyncIndex       index = ths->pLogStore->syncLogWriteIndex(ths->pLogStore);
-  SyncTerm        term = ths->pRaftStore->currentTerm;
+  SyncTerm        term = ths->raftStore.currentTerm;
   SSyncRaftEntry* pEntry;
 
   if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
@@ -2742,7 +2742,7 @@ int32_t syncNodeOnClientRequestOld(SSyncNode* ths, SRpcMsg* pMsg, SyncIndex* pRe
             .state = ths->state,
             .seqNum = pEntry->seqNum,
             .term = pEntry->term,
-            .currentTerm = ths->pRaftStore->currentTerm,
+            .currentTerm = ths->raftStore.currentTerm,
             .flag = 0,
         };
         ths->pFsm->FpCommitCb(ths->pFsm, pMsg, &cbMeta);
@@ -2820,7 +2820,7 @@ int32_t syncDoLeaderTransfer(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* p
     return 0;
   }
 
-  if (pEntry->term < ths->pRaftStore->currentTerm) {
+  if (pEntry->term < ths->raftStore.currentTerm) {
     sNTrace(ths, "little term:%" PRId64 ", can not do leader transfer", pEntry->term);
     return 0;
   }
@@ -2858,7 +2858,7 @@ int32_t syncDoLeaderTransfer(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* p
   if (ths->pFsm->FpLeaderTransferCb != NULL) {
     SFsmCbMeta cbMeta = {
         .code = 0,
-        .currentTerm = ths->pRaftStore->currentTerm,
+        .currentTerm = ths->raftStore.currentTerm,
         .flag = 0,
         .index = pEntry->index,
         .lastConfigIndex = syncNodeGetSnapshotConfigIndex(ths, pEntry->index),
@@ -2974,7 +2974,7 @@ int32_t syncNodeDoCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endInde
                 .state = ths->state,
                 .seqNum = pEntry->seqNum,
                 .term = pEntry->term,
-                .currentTerm = ths->pRaftStore->currentTerm,
+                .currentTerm = ths->raftStore.currentTerm,
                 .flag = flag,
             };
 

source/libs/sync/src/syncMessage.c

@@ -154,7 +154,7 @@ int32_t syncBuildAppendEntriesReply(SRpcMsg* pMsg, int32_t vgId) {
   return 0;
 }
 
-int32_t syncBuildAppendEntriesFromRaftLog(SSyncNode* pNode, SSyncRaftEntry* pEntry, SyncTerm prevLogTerm,
+int32_t syncBuildAppendEntriesFromRaftEntry(SSyncNode* pNode, SSyncRaftEntry* pEntry, SyncTerm prevLogTerm,
                                             SRpcMsg* pRpcMsg) {
   uint32_t dataLen = pEntry->bytes;
   uint32_t bytes = sizeof(SyncAppendEntries) + dataLen;
@@ -176,7 +176,7 @@ int32_t syncBuildAppendEntriesFromRaftLog(SSyncNode* pNode, SSyncRaftEntry* pEnt
   pMsg->prevLogTerm = prevLogTerm;
   pMsg->vgId = pNode->vgId;
   pMsg->srcId = pNode->myRaftId;
-  pMsg->term = pNode->pRaftStore->currentTerm;
+  pMsg->term = pNode->raftStore.currentTerm;
   pMsg->commitIndex = pNode->commitIndex;
   pMsg->privateTerm = 0;
   return 0;

source/libs/sync/src/syncPipeline.c

@@ -31,6 +31,10 @@ static bool syncIsMsgBlock(tmsg_t type) {
          (type == TDMT_VND_UPDATE_TAG_VAL) || (type == TDMT_VND_ALTER_CONFIRM);
 }
 
+FORCE_INLINE static int64_t syncGetRetryMaxWaitMs() {
+  return SYNC_LOG_REPL_RETRY_WAIT_MS * (1 << SYNC_MAX_RETRY_BACKOFF);
+}
+
 int64_t syncLogBufferGetEndIndex(SSyncLogBuffer* pBuf) {
   taosThreadMutexLock(&pBuf->mutex);
   int64_t index = pBuf->endIndex;
@@ -264,20 +268,27 @@ int32_t syncLogBufferReInit(SSyncLogBuffer* pBuf, SSyncNode* pNode) {
   return ret;
 }
 
-FORCE_INLINE SyncTerm syncLogBufferGetLastMatchTerm(SSyncLogBuffer* pBuf) {
+FORCE_INLINE SyncTerm syncLogBufferGetLastMatchTermWithoutLock(SSyncLogBuffer* pBuf) {
   SyncIndex       index = pBuf->matchIndex;
   SSyncRaftEntry* pEntry = pBuf->entries[(index + pBuf->size) % pBuf->size].pItem;
   ASSERT(pEntry != NULL);
   return pEntry->term;
 }
 
+SyncTerm syncLogBufferGetLastMatchTerm(SSyncLogBuffer* pBuf) {
+  taosThreadMutexLock(&pBuf->mutex);
+  SyncTerm term = syncLogBufferGetLastMatchTermWithoutLock(pBuf);
+  taosThreadMutexUnlock(&pBuf->mutex);
+  return term;
+}
+
 int32_t syncLogBufferAccept(SSyncLogBuffer* pBuf, SSyncNode* pNode, SSyncRaftEntry* pEntry, SyncTerm prevTerm) {
   taosThreadMutexLock(&pBuf->mutex);
   syncLogBufferValidate(pBuf);
   int32_t         ret = -1;
   SyncIndex       index = pEntry->index;
   SyncIndex       prevIndex = pEntry->index - 1;
-  SyncTerm  lastMatchTerm = syncLogBufferGetLastMatchTerm(pBuf);
+  SyncTerm        lastMatchTerm = syncLogBufferGetLastMatchTermWithoutLock(pBuf);
   SSyncRaftEntry* pExist = NULL;
   bool            inBuf = true;
 
@@ -329,6 +340,8 @@ int32_t syncLogBufferAccept(SSyncLogBuffer* pBuf, SSyncNode* pNode, SSyncRaftEnt
   }
 
   // update
+  ASSERT(pBuf->startIndex < index);
+  ASSERT(index - pBuf->startIndex < pBuf->size);
   ASSERT(pBuf->entries[index % pBuf->size].pItem == NULL);
   SSyncLogBufEntry tmp = {.pItem = pEntry, .prevLogIndex = prevIndex, .prevLogTerm = prevTerm};
   pEntry = NULL;
@@ -456,6 +469,11 @@ int32_t syncLogFsmExecute(SSyncNode* pNode, SSyncFSM* pFsm, ESyncState role, Syn
            pNode->vgId, pEntry->index, pEntry->term, TMSG_INFO(pEntry->originalRpcType), applyCode);
   }
 
+  if (pEntry->originalRpcType == TDMT_VND_COMMIT) {
+    sInfo("vgId:%d, fsm execute vnode commit. index: %" PRId64 ", term: %" PRId64 "", pNode->vgId, pEntry->index,
+          pEntry->term);
+  }
+
   SRpcMsg rpcMsg = {.code = applyCode};
   syncEntry2OriginalRpc(pEntry, &rpcMsg);
 
@@ -491,7 +509,7 @@ int32_t syncLogBufferCommit(SSyncLogBuffer* pBuf, SSyncNode* pNode, int64_t comm
   SSyncLogStore*  pLogStore = pNode->pLogStore;
   SSyncFSM*       pFsm = pNode->pFsm;
   ESyncState      role = pNode->state;
-  SyncTerm        term = pNode->pRaftStore->currentTerm;
+  SyncTerm        term = pNode->raftStore.currentTerm;
   SyncGroupId     vgId = pNode->vgId;
   int32_t         ret = -1;
   int64_t         upperIndex = TMIN(commitIndex, pBuf->matchIndex);
@@ -552,7 +570,8 @@ int32_t syncLogBufferCommit(SSyncLogBuffer* pBuf, SSyncNode* pNode, int64_t comm
   ret = 0;
 _out:
   // mark as restored if needed
-  if (!pNode->restoreFinish && pBuf->commitIndex >= pNode->commitIndex) {
+  if (!pNode->restoreFinish && pBuf->commitIndex >= pNode->commitIndex && pEntry != NULL &&
+      pNode->raftStore.currentTerm <= pEntry->term) {
     pNode->pFsm->FpRestoreFinishCb(pNode->pFsm);
     pNode->restoreFinish = true;
     sInfo("vgId:%d, restore finished. log buffer: [%" PRId64 " %" PRId64 " %" PRId64 ", %" PRId64 ")", pNode->vgId,
@@ -613,11 +632,17 @@ int32_t syncLogReplMgrRetryOnNeed(SSyncLogReplMgr* pMgr, SSyncNode* pNode) {
     }
 
     if (pMgr->states[pos].acked) {
+      if (pMgr->matchIndex < index && pMgr->states[pos].timeMs + (syncGetRetryMaxWaitMs() << 3) < nowMs) {
+        syncLogReplMgrReset(pMgr);
+        sWarn("vgId:%d, reset sync log repl mgr since stagnation. index: %" PRId64 ", peer: %" PRIx64, pNode->vgId,
+              index, pDestId->addr);
+        goto _out;
+      }
       continue;
     }
 
     bool barrier = false;
-    if (syncLogBufferReplicateOneTo(pMgr, pNode, index, &term, pDestId, &barrier) < 0) {
+    if (syncLogReplMgrReplicateOneTo(pMgr, pNode, index, &term, pDestId, &barrier) < 0) {
       sError("vgId:%d, failed to replicate sync log entry since %s. index: %" PRId64 ", dest: %" PRIx64 "", pNode->vgId,
              terrstr(), index, pDestId->addr);
       goto _out;
@@ -639,16 +664,17 @@ int32_t syncLogReplMgrRetryOnNeed(SSyncLogReplMgr* pMgr, SSyncNode* pNode) {
 _out:
   if (retried) {
     pMgr->retryBackoff = syncLogGetNextRetryBackoff(pMgr);
-    sInfo("vgId:%d, resent %d sync log entries. dest: %" PRIx64 ", indexes: %" PRId64 " ..., terms: ... %" PRId64
+    SSyncLogBuffer* pBuf = pNode->pLogBuf;
-          ", retryWaitMs: %" PRId64 ", repl mgr: [%" PRId64 " %" PRId64 ", %" PRId64 ")",
+    sInfo("vgId:%d, resend %d sync log entries. dest: %" PRIx64 ", indexes: %" PRId64 " ..., terms: ... %" PRId64
+          ", retryWaitMs: %" PRId64 ", mgr: [%" PRId64 " %" PRId64 ", %" PRId64 "), buffer: [%" PRId64 " %" PRId64
+          " %" PRId64 ", %" PRId64 ")",
           pNode->vgId, count, pDestId->addr, firstIndex, term, retryWaitMs, pMgr->startIndex, pMgr->matchIndex,
-          pMgr->endIndex);
+          pMgr->endIndex, pBuf->startIndex, pBuf->commitIndex, pBuf->matchIndex, pBuf->endIndex);
   }
   return ret;
 }
 
-int32_t syncLogReplMgrProcessReplyInRecoveryMode(SSyncLogReplMgr* pMgr, SSyncNode* pNode,
+int32_t syncLogReplMgrProcessReplyAsRecovery(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg) {
-                                                 SyncAppendEntriesReply* pMsg) {
   SSyncLogBuffer* pBuf = pNode->pLogBuf;
   SRaftId         destId = pMsg->srcId;
   ASSERT(pMgr->restored == false);
@@ -723,7 +749,7 @@ int32_t syncLogReplMgrProcessReplyInRecoveryMode(SSyncLogReplMgr* pMgr, SSyncNod
 
   // attempt to replicate the raft log at index
   (void)syncLogReplMgrReset(pMgr);
-  return syncLogReplMgrReplicateProbeOnce(pMgr, pNode, index);
+  return syncLogReplMgrReplicateProbe(pMgr, pNode, index);
 }
 
 int32_t syncLogReplMgrProcessHeartbeatReply(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncHeartbeatReply* pMsg) {
@@ -751,9 +777,9 @@ int32_t syncLogReplMgrProcessReply(SSyncLogReplMgr* pMgr, SSyncNode* pNode, Sync
   }
 
   if (pMgr->restored) {
-    (void)syncLogReplMgrProcessReplyInNormalMode(pMgr, pNode, pMsg);
+    (void)syncLogReplMgrProcessReplyAsNormal(pMgr, pNode, pMsg);
   } else {
-    (void)syncLogReplMgrProcessReplyInRecoveryMode(pMgr, pNode, pMsg);
+    (void)syncLogReplMgrProcessReplyAsRecovery(pMgr, pNode, pMsg);
   }
   taosThreadMutexUnlock(&pBuf->mutex);
   return 0;
@@ -761,17 +787,17 @@ int32_t syncLogReplMgrProcessReply(SSyncLogReplMgr* pMgr, SSyncNode* pNode, Sync
 
 int32_t syncLogReplMgrReplicateOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode) {
   if (pMgr->restored) {
-    (void)syncLogReplMgrReplicateAttemptedOnce(pMgr, pNode);
+    (void)syncLogReplMgrReplicateAttempt(pMgr, pNode);
   } else {
-    (void)syncLogReplMgrReplicateProbeOnce(pMgr, pNode, pNode->pLogBuf->matchIndex);
+    (void)syncLogReplMgrReplicateProbe(pMgr, pNode, pNode->pLogBuf->matchIndex);
   }
   return 0;
 }
 
-int32_t syncLogReplMgrReplicateProbeOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index) {
+int32_t syncLogReplMgrReplicateProbe(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index) {
   ASSERT(!pMgr->restored);
   ASSERT(pMgr->startIndex >= 0);
-  int64_t retryMaxWaitMs = SYNC_LOG_REPL_RETRY_WAIT_MS * (1 << SYNC_MAX_RETRY_BACKOFF);
+  int64_t retryMaxWaitMs = syncGetRetryMaxWaitMs();
   int64_t nowMs = taosGetMonoTimestampMs();
 
   if (pMgr->endIndex > pMgr->startIndex &&
@@ -783,7 +809,7 @@ int32_t syncLogReplMgrReplicateProbeOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode
   SRaftId* pDestId = &pNode->replicasId[pMgr->peerId];
   bool     barrier = false;
   SyncTerm term = -1;
-  if (syncLogBufferReplicateOneTo(pMgr, pNode, index, &term, pDestId, &barrier) < 0) {
+  if (syncLogReplMgrReplicateOneTo(pMgr, pNode, index, &term, pDestId, &barrier) < 0) {
     sError("vgId:%d, failed to replicate log entry since %s. index: %" PRId64 ", dest: 0x%016" PRIx64 "", pNode->vgId,
            terrstr(), index, pDestId->addr);
     return -1;
@@ -799,15 +825,14 @@ int32_t syncLogReplMgrReplicateProbeOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode
   pMgr->endIndex = index + 1;
 
   SSyncLogBuffer* pBuf = pNode->pLogBuf;
-  sTrace("vgId:%d, attempted to probe the %d'th peer with msg of index:%" PRId64 " term: %" PRId64
+  sInfo("vgId:%d, probe peer:%" PRIx64 " with msg of index:%" PRId64 " term: %" PRId64 ". mgr (rs:%d): [%" PRId64
-         ". pMgr(rs:%d): [%" PRId64 " %" PRId64 ", %" PRId64 "), pBuf: [%" PRId64 " %" PRId64 " %" PRId64 ", %" PRId64
+        " %" PRId64 ", %" PRId64 "), buffer: [%" PRId64 " %" PRId64 " %" PRId64 ", %" PRId64 ")",
-         ")",
+        pNode->vgId, pDestId->addr, index, term, pMgr->restored, pMgr->startIndex, pMgr->matchIndex, pMgr->endIndex,
-         pNode->vgId, pMgr->peerId, index, term, pMgr->restored, pMgr->startIndex, pMgr->matchIndex, pMgr->endIndex,
         pBuf->startIndex, pBuf->commitIndex, pBuf->matchIndex, pBuf->endIndex);
   return 0;
 }
 
-int32_t syncLogReplMgrReplicateAttemptedOnce(SSyncLogReplMgr* pMgr, SSyncNode* pNode) {
+int32_t syncLogReplMgrReplicateAttempt(SSyncLogReplMgr* pMgr, SSyncNode* pNode) {
   ASSERT(pMgr->restored);
 
   SRaftId*  pDestId = &pNode->replicasId[pMgr->peerId];
@@ -815,9 +840,11 @@ int32_t syncLogReplMgrReplicateAttemptedOnce(SSyncLogReplMgr* pMgr, SSyncNode* p
   int32_t   count = 0;
   int64_t   nowMs = taosGetMonoTimestampMs();
   int64_t   limit = pMgr->size >> 1;
+  SyncTerm  term = -1;
+  SyncIndex firstIndex = -1;
 
   for (SyncIndex index = pMgr->endIndex; index <= pNode->pLogBuf->matchIndex; index++) {
-    if (batchSize < count++ || limit <= index - pMgr->startIndex) {
+    if (batchSize < count || limit <= index - pMgr->startIndex) {
       break;
     }
     if (pMgr->startIndex + 1 < index && pMgr->states[(index - 1) % pMgr->size].barrier) {
@@ -827,7 +854,7 @@ int32_t syncLogReplMgrReplicateAttemptedOnce(SSyncLogReplMgr* pMgr, SSyncNode* p
     SRaftId* pDestId = &pNode->replicasId[pMgr->peerId];
     bool     barrier = false;
     SyncTerm term = -1;
-    if (syncLogBufferReplicateOneTo(pMgr, pNode, index, &term, pDestId, &barrier) < 0) {
+    if (syncLogReplMgrReplicateOneTo(pMgr, pNode, index, &term, pDestId, &barrier) < 0) {
       sError("vgId:%d, failed to replicate log entry since %s. index: %" PRId64 ", dest: 0x%016" PRIx64 "", pNode->vgId,
              terrstr(), index, pDestId->addr);
       return -1;
@@ -837,6 +864,9 @@ int32_t syncLogReplMgrReplicateAttemptedOnce(SSyncLogReplMgr* pMgr, SSyncNode* p
     pMgr->states[pos].term = term;
     pMgr->states[pos].acked = false;
 
+    if (firstIndex == -1) firstIndex = index;
+    count++;
+
     pMgr->endIndex = index + 1;
     if (barrier) {
       sInfo("vgId:%d, replicated sync barrier to dest: %" PRIx64 ". index: %" PRId64 ", term: %" PRId64
@@ -850,14 +880,15 @@ int32_t syncLogReplMgrReplicateAttemptedOnce(SSyncLogReplMgr* pMgr, SSyncNode* p
   syncLogReplMgrRetryOnNeed(pMgr, pNode);
 
   SSyncLogBuffer* pBuf = pNode->pLogBuf;
-  sTrace("vgId:%d, attempted to replicate %d msgs to the %d'th peer. pMgr(rs:%d): [%" PRId64 " %" PRId64 ", %" PRId64
+  sTrace("vgId:%d, replicated %d msgs to peer: %" PRIx64 ". indexes: %" PRId64 "..., terms: ...%" PRId64
-         "), pBuf: [%" PRId64 " %" PRId64 " %" PRId64 ", %" PRId64 ")",
+         ", mgr: (rs:%d) [%" PRId64 " %" PRId64 ", %" PRId64 "), buffer: [%" PRId64 " %" PRId64 " %" PRId64 ", %" PRId64
-         pNode->vgId, count, pMgr->peerId, pMgr->restored, pMgr->startIndex, pMgr->matchIndex, pMgr->endIndex,
+         ")",
-         pBuf->startIndex, pBuf->commitIndex, pBuf->matchIndex, pBuf->endIndex);
+         pNode->vgId, count, pDestId->addr, firstIndex, term, pMgr->restored, pMgr->startIndex, pMgr->matchIndex,
+         pMgr->endIndex, pBuf->startIndex, pBuf->commitIndex, pBuf->matchIndex, pBuf->endIndex);
   return 0;
 }
 
-int32_t syncLogReplMgrProcessReplyInNormalMode(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg) {
+int32_t syncLogReplMgrProcessReplyAsNormal(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncAppendEntriesReply* pMsg) {
   ASSERT(pMgr->restored == true);
   if (pMgr->startIndex <= pMsg->lastSendIndex && pMsg->lastSendIndex < pMgr->endIndex) {
     if (pMgr->startIndex < pMgr->matchIndex && pMgr->retryBackoff > 0) {
@@ -876,7 +907,7 @@ int32_t syncLogReplMgrProcessReplyInNormalMode(SSyncLogReplMgr* pMgr, SSyncNode*
     pMgr->startIndex = pMgr->matchIndex;
   }
 
-  return syncLogReplMgrReplicateAttemptedOnce(pMgr, pNode);
+  return syncLogReplMgrReplicateAttempt(pMgr, pNode);
 }
 
 SSyncLogReplMgr* syncLogReplMgrCreate() {
@@ -985,6 +1016,10 @@ void syncLogBufferDestroy(SSyncLogBuffer* pBuf) {
 int32_t syncLogBufferRollback(SSyncLogBuffer* pBuf, SSyncNode* pNode, SyncIndex toIndex) {
   ASSERT(pBuf->commitIndex < toIndex && toIndex <= pBuf->endIndex);
 
+  if (toIndex == pBuf->endIndex) {
+    return 0;
+  }
+
   sInfo("vgId:%d, rollback sync log buffer. toindex: %" PRId64 ", buffer: [%" PRId64 " %" PRId64 " %" PRId64
         ", %" PRId64 ")",
         pNode->vgId, toIndex, pBuf->startIndex, pBuf->commitIndex, pBuf->matchIndex, pBuf->endIndex);
@@ -1066,12 +1101,11 @@ SSyncRaftEntry* syncLogBufferGetOneEntry(SSyncLogBuffer* pBuf, SSyncNode* pNode,
   return pEntry;
 }
 
-int32_t syncLogBufferReplicateOneTo(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index, SyncTerm* pTerm,
+int32_t syncLogReplMgrReplicateOneTo(SSyncLogReplMgr* pMgr, SSyncNode* pNode, SyncIndex index, SyncTerm* pTerm,
                                      SRaftId* pDestId, bool* pBarrier) {
   SSyncRaftEntry* pEntry = NULL;
   SRpcMsg         msgOut = {0};
   bool            inBuf = false;
-  int32_t         ret = -1;
   SyncTerm        prevLogTerm = -1;
   SSyncLogBuffer* pBuf = pNode->pLogBuf;
 
@@ -1097,14 +1131,13 @@ int32_t syncLogBufferReplicateOneTo(SSyncLogReplMgr* pMgr, SSyncNode* pNode, Syn
   }
   if (pTerm) *pTerm = pEntry->term;
 
-  int32_t code = syncBuildAppendEntriesFromRaftLog(pNode, pEntry, prevLogTerm, &msgOut);
+  int32_t code = syncBuildAppendEntriesFromRaftEntry(pNode, pEntry, prevLogTerm, &msgOut);
   if (code < 0) {
     sError("vgId:%d, failed to get append entries for index:%" PRId64 "", pNode->vgId, index);
     goto _err;
   }
 
   (void)syncNodeSendAppendEntries(pNode, pDestId, &msgOut);
-  ret = 0;
 
   sTrace("vgId:%d, replicate one msg index: %" PRId64 " term: %" PRId64 " prevterm: %" PRId64 " to dest: 0x%016" PRIx64,
          pNode->vgId, pEntry->index, pEntry->term, prevLogTerm, pDestId->addr);

source/libs/sync/src/syncRaftCfg.c

@@ -71,31 +71,23 @@ int32_t syncWriteCfgFile(SSyncNode *pNode) {
   char        file[PATH_MAX] = {0};
   snprintf(file, sizeof(file), "%s.bak", realfile);
 
-  pFile = taosOpenFile(file, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  if (pFile == NULL) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    sError("vgId:%d, failed to open sync cfg file:%s since %s", pNode->vgId, realfile, terrstr());
-    goto _OVER;
-  }
-
   terrno = TSDB_CODE_OUT_OF_MEMORY;
   pJson = tjsonCreateObject();
   if (pJson == NULL) goto _OVER;
   if (tjsonAddObject(pJson, "RaftCfg", syncEncodeRaftCfg, pCfg) < 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);
 
-  if (taosRenameFile(file, realfile) != 0) {
+  taosCloseFile(&pFile);
-    terrno = TAOS_SYSTEM_ERROR(errno);
+  if (taosRenameFile(file, realfile) != 0) goto _OVER;
-    sError("vgId:%d, failed to rename sync cfg file:%s to %s since %s", pNode->vgId, file, realfile, terrstr());
-    goto _OVER;
-  }
 
   code = 0;
   sInfo("vgId:%d, succeed to write sync cfg file:%s, len:%d", pNode->vgId, realfile, len);
@@ -106,6 +98,7 @@ _OVER:
   if (pFile != NULL) taosCloseFile(&pFile);
 
   if (code != 0) {
+    if (terrno == 0) terrno = TAOS_SYSTEM_ERROR(errno);
     sError("vgId:%d, failed to write sync cfg file:%s since %s", pNode->vgId, realfile, terrstr());
   }
   return code;

source/libs/sync/src/syncRaftLog.c

@@ -219,6 +219,10 @@ static int32_t raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntr
 
   ASSERT(pEntry->index == index);
 
+  if (pEntry->originalRpcType == TDMT_VND_COMMIT) {
+    walFsync(pWal, true);
+  }
+
   sNTrace(pData->pSyncNode, "write index:%" PRId64 ", type:%s, origin type:%s, elapsed:%" PRId64, pEntry->index,
           TMSG_INFO(pEntry->msgType), TMSG_INFO(pEntry->originalRpcType), tsElapsed);
   return 0;
@@ -312,29 +316,6 @@ static int32_t raftLogTruncate(struct SSyncLogStore* pLogStore, SyncIndex fromIn
   SSyncLogStoreData* pData = pLogStore->data;
   SWal*              pWal = pData->pWal;
 
-  // need not truncate
-  SyncIndex wallastVer = walGetLastVer(pWal);
-  if (fromIndex > wallastVer) {
-    return 0;
-  }
-
-  // need not truncate
-  SyncIndex walCommitVer = walGetCommittedVer(pWal);
-  if (fromIndex <= walCommitVer) {
-    return 0;
-  }
-
-  // delete from cache
-  for (SyncIndex index = fromIndex; index <= wallastVer; ++index) {
-    SLRUCache* pCache = pData->pSyncNode->pLogStore->pCache;
-    LRUHandle* h = taosLRUCacheLookup(pCache, &index, sizeof(index));
-    if (h) {
-      sNTrace(pData->pSyncNode, "cache delete index:%" PRId64, index);
-
-      taosLRUCacheRelease(pData->pSyncNode->pLogStore->pCache, h, true);
-    }
-  }
-
   int32_t code = walRollback(pWal, fromIndex);
   if (code != 0) {
     int32_t     err = terrno;

source/libs/sync/src/syncRaftStore.c

@@ -16,156 +16,161 @@
 #define _DEFAULT_SOURCE
 #include "syncRaftStore.h"
 #include "syncUtil.h"
+#include "tjson.h"
 
-// private function
+static int32_t raftStoreDecode(const SJson *pJson, SRaftStore *pStore) {
-static int32_t raftStoreInit(SRaftStore *pRaftStore);
+  int32_t code = 0;
-static bool    raftStoreFileExist(char *path);
 
-// public function
+  tjsonGetNumberValue(pJson, "current_term", pStore->currentTerm, code);
-SRaftStore *raftStoreOpen(const char *path) {
+  if (code < 0) return -1;
-  int32_t ret;
+  tjsonGetNumberValue(pJson, "vote_for_addr", pStore->voteFor.addr, code);
+  if (code < 0) return -1;
+  tjsonGetInt32ValueFromDouble(pJson, "vote_for_vgid", pStore->voteFor.vgId, code);
+  if (code < 0) return -1;
 
-  SRaftStore *pRaftStore = taosMemoryCalloc(1, sizeof(SRaftStore));
+  return 0;
-  if (pRaftStore == NULL) {
+}
+
+int32_t raftStoreReadFile(SSyncNode *pNode) {
+  int32_t     code = -1;
+  TdFilePtr   pFile = NULL;
+  char       *pData = NULL;
+  SJson      *pJson = NULL;
+  const char *file = pNode->raftStorePath;
+  SRaftStore *pStore = &pNode->raftStore;
+
+  if (taosStatFile(file, NULL, NULL) < 0) {
+    sInfo("vgId:%d, raft store file:%s not exist, use default value", pNode->vgId, file);
+    pStore->currentTerm = 0;
+    pStore->voteFor.addr = 0;
+    pStore->voteFor.vgId = 0;
+    return raftStoreWriteFile(pNode);
+  }
+
+  pFile = taosOpenFile(file, TD_FILE_READ);
+  if (pFile == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    sError("vgId:%d, failed to open raft store file:%s since %s", pNode->vgId, file, terrstr());
+    goto _OVER;
+  }
+
+  int64_t size = 0;
+  if (taosFStatFile(pFile, &size, NULL) < 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    sError("vgId:%d, failed to fstat raft store file:%s since %s", pNode->vgId, file, terrstr());
+    goto _OVER;
+  }
+
+  pData = taosMemoryMalloc(size + 1);
+  if (pData == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
-    return NULL;
+    goto _OVER;
   }
 
-  snprintf(pRaftStore->path, sizeof(pRaftStore->path), "%s", path);
+  if (taosReadFile(pFile, pData, size) != size) {
-  if (!raftStoreFileExist(pRaftStore->path)) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
-    ret = raftStoreInit(pRaftStore);
+    sError("vgId:%d, failed to read raft store file:%s since %s", pNode->vgId, file, terrstr());
-    ASSERT(ret == 0);
+    goto _OVER;
   }
 
-  char storeBuf[RAFT_STORE_BLOCK_SIZE] = {0};
+  pData[size] = '\0';
-  pRaftStore->pFile = taosOpenFile(path, TD_FILE_READ | TD_FILE_WRITE);
-  ASSERT(pRaftStore->pFile != NULL);
 
-  int len = taosReadFile(pRaftStore->pFile, storeBuf, RAFT_STORE_BLOCK_SIZE);
+  pJson = tjsonParse(pData);
-  ASSERT(len > 0);
+  if (pJson == NULL) {
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
+    goto _OVER;
+  }
 
-  ret = raftStoreDeserialize(pRaftStore, storeBuf, len);
+  if (raftStoreDecode(pJson, pStore) < 0) {
-  ASSERT(ret == 0);
+    terrno = TSDB_CODE_INVALID_JSON_FORMAT;
+    goto _OVER;
+  }
 
-  return pRaftStore;
+  code = 0;
+  sInfo("vgId:%d, succceed to read raft store file %s", pNode->vgId, file);
+
+_OVER:
+  if (pData != NULL) taosMemoryFree(pData);
+  if (pJson != NULL) cJSON_Delete(pJson);
+  if (pFile != NULL) taosCloseFile(&pFile);
+
+  if (code != 0) {
+    sError("vgId:%d, failed to read raft store file:%s since %s", pNode->vgId, file, terrstr());
+  }
+  return code;
 }
 
-static int32_t raftStoreInit(SRaftStore *pRaftStore) {
+static int32_t raftStoreEncode(SJson *pJson, SRaftStore *pStore) {
-  ASSERT(pRaftStore != NULL);
+  if (tjsonAddIntegerToObject(pJson, "current_term", pStore->currentTerm) < 0) return -1;
-
+  if (tjsonAddIntegerToObject(pJson, "vote_for_addr", pStore->voteFor.addr) < 0) return -1;
-  pRaftStore->pFile = taosOpenFile(pRaftStore->path, TD_FILE_CREATE | TD_FILE_WRITE);
+  if (tjsonAddDoubleToObject(pJson, "vote_for_vgid", pStore->voteFor.vgId) < 0) return -1;
-  ASSERT(pRaftStore->pFile != NULL);
-
-  pRaftStore->currentTerm = 0;
-  pRaftStore->voteFor.addr = 0;
-  pRaftStore->voteFor.vgId = 0;
-
-  int32_t ret = raftStorePersist(pRaftStore);
-  ASSERT(ret == 0);
-
-  taosCloseFile(&pRaftStore->pFile);
   return 0;
 }
 
-int32_t raftStoreClose(SRaftStore *pRaftStore) {
+int32_t raftStoreWriteFile(SSyncNode *pNode) {
-  if (pRaftStore == NULL) return 0;
+  int32_t     code = -1;
+  char       *buffer = NULL;
+  SJson      *pJson = NULL;
+  TdFilePtr   pFile = NULL;
+  const char *realfile = pNode->raftStorePath;
+  SRaftStore *pStore = &pNode->raftStore;
+  char        file[PATH_MAX] = {0};
+  snprintf(file, sizeof(file), "%s.bak", realfile);
 
-  taosCloseFile(&pRaftStore->pFile);
+  terrno = TSDB_CODE_OUT_OF_MEMORY;
-  taosMemoryFree(pRaftStore);
+  pJson = tjsonCreateObject();
-  pRaftStore = NULL;
+  if (pJson == NULL) goto _OVER;
-  return 0;
+  if (raftStoreEncode(pJson, pStore) != 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);
+  if (taosRenameFile(file, realfile) != 0) goto _OVER;
+
+  code = 0;
+  sInfo("vgId:%d, succeed to write raft store file:%s, len:%d", pNode->vgId, realfile, len);
+
+_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);
+    sError("vgId:%d, failed to write raft store file:%s since %s", pNode->vgId, realfile, terrstr());
+  }
+  return code;
 }
 
-int32_t raftStorePersist(SRaftStore *pRaftStore) {
+bool raftStoreHasVoted(SSyncNode *pNode) {
-  ASSERT(pRaftStore != NULL);
+  bool b = syncUtilEmptyId(&pNode->raftStore.voteFor);
-
-  int32_t ret;
-  char    storeBuf[RAFT_STORE_BLOCK_SIZE] = {0};
-  ret = raftStoreSerialize(pRaftStore, storeBuf, sizeof(storeBuf));
-  ASSERT(ret == 0);
-
-  taosLSeekFile(pRaftStore->pFile, 0, SEEK_SET);
-
-  ret = taosWriteFile(pRaftStore->pFile, storeBuf, sizeof(storeBuf));
-  ASSERT(ret == RAFT_STORE_BLOCK_SIZE);
-
-  taosFsyncFile(pRaftStore->pFile);
-  return 0;
-}
-
-static bool raftStoreFileExist(char *path) {
-  bool b = taosStatFile(path, NULL, NULL) >= 0;
-  return b;
-}
-
-int32_t raftStoreSerialize(SRaftStore *pRaftStore, char *buf, size_t len) {
-  ASSERT(pRaftStore != NULL);
-
-  cJSON *pRoot = cJSON_CreateObject();
-
-  char u64Buf[128] = {0};
-  snprintf(u64Buf, sizeof(u64Buf), "%" PRIu64 "", pRaftStore->currentTerm);
-  cJSON_AddStringToObject(pRoot, "current_term", u64Buf);
-
-  snprintf(u64Buf, sizeof(u64Buf), "%" PRIu64 "", pRaftStore->voteFor.addr);
-  cJSON_AddStringToObject(pRoot, "vote_for_addr", u64Buf);
-
-  cJSON_AddNumberToObject(pRoot, "vote_for_vgid", pRaftStore->voteFor.vgId);
-
-  char *serialized = cJSON_Print(pRoot);
-  int   len2 = strlen(serialized);
-  ASSERT(len2 < len);
-  memset(buf, 0, len);
-  snprintf(buf, len, "%s", serialized);
-  taosMemoryFree(serialized);
-
-  cJSON_Delete(pRoot);
-  return 0;
-}
-
-int32_t raftStoreDeserialize(SRaftStore *pRaftStore, char *buf, size_t len) {
-  ASSERT(pRaftStore != NULL);
-
-  ASSERT(len > 0 && len <= RAFT_STORE_BLOCK_SIZE);
-  cJSON *pRoot = cJSON_Parse(buf);
-
-  cJSON *pCurrentTerm = cJSON_GetObjectItem(pRoot, "current_term");
-  ASSERT(cJSON_IsString(pCurrentTerm));
-  sscanf(pCurrentTerm->valuestring, "%" PRIu64 "", &(pRaftStore->currentTerm));
-
-  cJSON *pVoteForAddr = cJSON_GetObjectItem(pRoot, "vote_for_addr");
-  ASSERT(cJSON_IsString(pVoteForAddr));
-  sscanf(pVoteForAddr->valuestring, "%" PRIu64 "", &(pRaftStore->voteFor.addr));
-
-  cJSON *pVoteForVgid = cJSON_GetObjectItem(pRoot, "vote_for_vgid");
-  pRaftStore->voteFor.vgId = pVoteForVgid->valueint;
-
-  cJSON_Delete(pRoot);
-  return 0;
-}
-
-bool raftStoreHasVoted(SRaftStore *pRaftStore) {
-  bool b = syncUtilEmptyId(&(pRaftStore->voteFor));
   return (!b);
 }
 
-void raftStoreVote(SRaftStore *pRaftStore, SRaftId *pRaftId) {
+void raftStoreVote(SSyncNode *pNode, SRaftId *pRaftId) {
-  ASSERT(!syncUtilEmptyId(pRaftId));
+  pNode->raftStore.voteFor = *pRaftId;
-  pRaftStore->voteFor = *pRaftId;
+  (void)raftStoreWriteFile(pNode);
-  raftStorePersist(pRaftStore);
 }
 
-void raftStoreClearVote(SRaftStore *pRaftStore) {
+void raftStoreClearVote(SSyncNode *pNode) {
-  pRaftStore->voteFor = EMPTY_RAFT_ID;
+  pNode->raftStore.voteFor = EMPTY_RAFT_ID;
-  raftStorePersist(pRaftStore);
+  (void)raftStoreWriteFile(pNode);
 }
 
-void raftStoreNextTerm(SRaftStore *pRaftStore) {
+void raftStoreNextTerm(SSyncNode *pNode) {
-  ++(pRaftStore->currentTerm);
+  pNode->raftStore.currentTerm++;
-  raftStorePersist(pRaftStore);
+  (void)raftStoreWriteFile(pNode);
 }
 
-void raftStoreSetTerm(SRaftStore *pRaftStore, SyncTerm term) {
+void raftStoreSetTerm(SSyncNode *pNode, SyncTerm term) {
-  pRaftStore->currentTerm = term;
+  pNode->raftStore.currentTerm = term;
-  raftStorePersist(pRaftStore);
+  (void)raftStoreWriteFile(pNode);
 }

source/libs/sync/src/syncReplication.c

@@ -122,7 +122,7 @@ int32_t syncNodeReplicateOne(SSyncNode* pSyncNode, SRaftId* pDestId, bool snapsh
   ASSERT(pMsg != NULL);
   pMsg->srcId = pSyncNode->myRaftId;
   pMsg->destId = *pDestId;
-  pMsg->term = pSyncNode->pRaftStore->currentTerm;
+  pMsg->term = pSyncNode->raftStore.currentTerm;
   pMsg->prevLogIndex = preLogIndex;
   pMsg->prevLogTerm = preLogTerm;
   pMsg->commitIndex = pSyncNode->commitIndex;
@@ -245,7 +245,7 @@ int32_t syncNodeHeartbeatPeers(SSyncNode* pSyncNode) {
     SyncHeartbeat* pSyncMsg = rpcMsg.pCont;
     pSyncMsg->srcId = pSyncNode->myRaftId;
     pSyncMsg->destId = pSyncNode->peersId[i];
-    pSyncMsg->term = pSyncNode->pRaftStore->currentTerm;
+    pSyncMsg->term = pSyncNode->raftStore.currentTerm;
     pSyncMsg->commitIndex = pSyncNode->commitIndex;
     pSyncMsg->minMatchIndex = syncMinMatchIndex(pSyncNode);
     pSyncMsg->privateTerm = 0;

source/libs/sync/src/syncRequestVote.c

@@ -44,21 +44,12 @@
 //       /\ UNCHANGED <<state, currentTerm, candidateVars, leaderVars, logVars>>
 //
 
-static bool syncNodeOnRequestVoteLogOK(SSyncNode* pSyncNode, SyncRequestVote* pMsg) {
+static bool syncNodeOnRequestVoteLogOK(SSyncNode* ths, SyncRequestVote* pMsg) {
-  SyncTerm  myLastTerm = syncNodeGetLastTerm(pSyncNode);
+  SyncTerm  myLastTerm = syncNodeGetLastTerm(ths);
-  SyncIndex myLastIndex = syncNodeGetLastIndex(pSyncNode);
+  SyncIndex myLastIndex = syncNodeGetLastIndex(ths);
-
-  if (pMsg->lastLogIndex < pSyncNode->commitIndex) {
-    sNTrace(pSyncNode,
-            "logok:0, {my-lterm:%" PRIu64 ", my-lindex:%" PRId64 ", recv-lterm:%" PRIu64 ", recv-lindex:%" PRId64
-            ", recv-term:%" PRIu64 "}",
-            myLastTerm, myLastIndex, pMsg->lastLogTerm, pMsg->lastLogIndex, pMsg->term);
-
-    return false;
-  }
 
   if (myLastTerm == SYNC_TERM_INVALID) {
-    sNTrace(pSyncNode,
+    sNTrace(ths,
             "logok:0, {my-lterm:%" PRIu64 ", my-lindex:%" PRId64 ", recv-lterm:%" PRIu64 ", recv-lindex:%" PRId64
             ", recv-term:%" PRIu64 "}",
             myLastTerm, myLastIndex, pMsg->lastLogTerm, pMsg->lastLogIndex, pMsg->term);
@@ -66,22 +57,29 @@ static bool syncNodeOnRequestVoteLogOK(SSyncNode* pSyncNode, SyncRequestVote* pM
   }
 
   if (pMsg->lastLogTerm > myLastTerm) {
-    sNTrace(pSyncNode,
+    sNTrace(ths,
             "logok:1, {my-lterm:%" PRIu64 ", my-lindex:%" PRId64 ", recv-lterm:%" PRIu64 ", recv-lindex:%" PRId64
             ", recv-term:%" PRIu64 "}",
             myLastTerm, myLastIndex, pMsg->lastLogTerm, pMsg->lastLogIndex, pMsg->term);
+
+    if (pMsg->lastLogIndex < ths->commitIndex) {
+      sNWarn(ths,
+             "logok:1, commit rollback required. {my-lterm:%" PRIu64 ", my-lindex:%" PRId64 ", recv-lterm:%" PRIu64
+             ", recv-lindex:%" PRId64 ", recv-term:%" PRIu64 "}",
+             myLastTerm, myLastIndex, pMsg->lastLogTerm, pMsg->lastLogIndex, pMsg->term);
+    }
     return true;
   }
 
   if (pMsg->lastLogTerm == myLastTerm && pMsg->lastLogIndex >= myLastIndex) {
-    sNTrace(pSyncNode,
+    sNTrace(ths,
             "logok:1, {my-lterm:%" PRIu64 ", my-lindex:%" PRId64 ", recv-lterm:%" PRIu64 ", recv-lindex:%" PRId64
             ", recv-term:%" PRIu64 "}",
             myLastTerm, myLastIndex, pMsg->lastLogTerm, pMsg->lastLogIndex, pMsg->term);
     return true;
   }
 
-  sNTrace(pSyncNode,
+  sNTrace(ths,
           "logok:0, {my-lterm:%" PRIu64 ", my-lindex:%" PRId64 ", recv-lterm:%" PRIu64 ", recv-lindex:%" PRId64
           ", recv-term:%" PRIu64 "}",
           myLastTerm, myLastIndex, pMsg->lastLogTerm, pMsg->lastLogIndex, pMsg->term);
@@ -93,7 +91,7 @@ int32_t syncNodeOnRequestVote(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   SyncRequestVote* pMsg = pRpcMsg->pCont;
 
   // if already drop replica, do not process
-  if (!syncNodeInRaftGroup(ths, &(pMsg->srcId))) {
+  if (!syncNodeInRaftGroup(ths, &pMsg->srcId)) {
     syncLogRecvRequestVote(ths, pMsg, -1, "not in my config");
     return -1;
   }
@@ -101,21 +99,21 @@ int32_t syncNodeOnRequestVote(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   bool logOK = syncNodeOnRequestVoteLogOK(ths, pMsg);
 
   // maybe update term
-  if (pMsg->term > ths->pRaftStore->currentTerm) {
+  if (pMsg->term > ths->raftStore.currentTerm) {
     syncNodeStepDown(ths, pMsg->term);
     // syncNodeUpdateTerm(ths, pMsg->term);
   }
-  ASSERT(pMsg->term <= ths->pRaftStore->currentTerm);
+  ASSERT(pMsg->term <= ths->raftStore.currentTerm);
 
-  bool grant = (pMsg->term == ths->pRaftStore->currentTerm) && logOK &&
+  bool grant = (pMsg->term == ths->raftStore.currentTerm) && logOK &&
-               ((!raftStoreHasVoted(ths->pRaftStore)) || (syncUtilSameId(&(ths->pRaftStore->voteFor), &(pMsg->srcId))));
+               ((!raftStoreHasVoted(ths)) || (syncUtilSameId(&ths->raftStore.voteFor, &pMsg->srcId)));
   if (grant) {
     // maybe has already voted for pMsg->srcId
     // vote again, no harm
-    raftStoreVote(ths->pRaftStore, &(pMsg->srcId));
+    raftStoreVote(ths, &(pMsg->srcId));
 
     // candidate ?
-    syncNodeStepDown(ths, ths->pRaftStore->currentTerm);
+    syncNodeStepDown(ths, ths->raftStore.currentTerm);
 
     // forbid elect for this round
     syncNodeResetElectTimer(ths);
@@ -129,7 +127,7 @@ int32_t syncNodeOnRequestVote(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   SyncRequestVoteReply* pReply = rpcMsg.pCont;
   pReply->srcId = ths->myRaftId;
   pReply->destId = pMsg->srcId;
-  pReply->term = ths->pRaftStore->currentTerm;
+  pReply->term = ths->raftStore.currentTerm;
   pReply->voteGranted = grant;
 
   // trace log

source/libs/sync/src/syncRequestVoteReply.c

@@ -49,25 +49,25 @@ int32_t syncNodeOnRequestVoteReply(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
   }
 
   // drop stale response
-  if (pMsg->term < ths->pRaftStore->currentTerm) {
+  if (pMsg->term < ths->raftStore.currentTerm) {
     syncLogRecvRequestVoteReply(ths, pMsg, "drop stale response");
     return -1;
   }
 
-  // ASSERT(!(pMsg->term > ths->pRaftStore->currentTerm));
+  // ASSERT(!(pMsg->term > ths->raftStore.currentTerm));
   //  no need this code, because if I receive reply.term, then I must have sent for that term.
-  //   if (pMsg->term > ths->pRaftStore->currentTerm) {
+  //   if (pMsg->term > ths->raftStore.currentTerm) {
   //     syncNodeUpdateTerm(ths, pMsg->term);
   //   }
 
-  if (pMsg->term > ths->pRaftStore->currentTerm) {
+  if (pMsg->term > ths->raftStore.currentTerm) {
     syncLogRecvRequestVoteReply(ths, pMsg, "error term");
     syncNodeStepDown(ths, pMsg->term);
     return -1;
   }
 
   syncLogRecvRequestVoteReply(ths, pMsg, "");
-  ASSERT(pMsg->term == ths->pRaftStore->currentTerm);
+  ASSERT(pMsg->term == ths->raftStore.currentTerm);
 
   // This tallies votes even when the current state is not Candidate,
   // but they won't be looked at, so it doesn't matter.

source/libs/sync/src/syncRespMgr.c

@@ -143,7 +143,7 @@ static void syncRespCleanByTTL(SSyncRespMgr *pObj, int64_t ttl, bool rsp) {
           .state = pNode->state,
           .seqNum = *pSeqNum,
           .term = SYNC_TERM_INVALID,
-          .currentTerm = pNode->pRaftStore->currentTerm,
+          .currentTerm = pNode->raftStore.currentTerm,
           .flag = 0,
       };
 

source/libs/sync/src/syncSnapshot.c

@@ -43,7 +43,7 @@ SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaI
   pSender->sendingMS = SYNC_SNAPSHOT_RETRY_MS;
   pSender->pSyncNode = pSyncNode;
   pSender->replicaIndex = replicaIndex;
-  pSender->term = pSyncNode->pRaftStore->currentTerm;
+  pSender->term = pSyncNode->raftStore.currentTerm;
   pSender->startTime = 0;
   pSender->endTime = 0;
   pSender->pSyncNode->pFsm->FpGetSnapshotInfo(pSender->pSyncNode->pFsm, &pSender->snapshot);
@@ -90,7 +90,7 @@ int32_t snapshotSenderStart(SSyncSnapshotSender *pSender) {
 
   memset(&pSender->lastConfig, 0, sizeof(pSender->lastConfig));
   pSender->sendingMS = 0;
-  pSender->term = pSender->pSyncNode->pRaftStore->currentTerm;
+  pSender->term = pSender->pSyncNode->raftStore.currentTerm;
   pSender->startTime = taosGetTimestampMs();
   pSender->lastSendTime = pSender->startTime;
   pSender->finish = false;
@@ -105,7 +105,7 @@ int32_t snapshotSenderStart(SSyncSnapshotSender *pSender) {
   SyncSnapshotSend *pMsg = rpcMsg.pCont;
   pMsg->srcId = pSender->pSyncNode->myRaftId;
   pMsg->destId = pSender->pSyncNode->replicasId[pSender->replicaIndex];
-  pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
+  pMsg->term = pSender->pSyncNode->raftStore.currentTerm;
   pMsg->beginIndex = pSender->snapshotParam.start;
   pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
   pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
@@ -185,7 +185,7 @@ static int32_t snapshotSend(SSyncSnapshotSender *pSender) {
   SyncSnapshotSend *pMsg = rpcMsg.pCont;
   pMsg->srcId = pSender->pSyncNode->myRaftId;
   pMsg->destId = pSender->pSyncNode->replicasId[pSender->replicaIndex];
-  pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
+  pMsg->term = pSender->pSyncNode->raftStore.currentTerm;
   pMsg->beginIndex = pSender->snapshotParam.start;
   pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
   pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
@@ -226,7 +226,7 @@ int32_t snapshotReSend(SSyncSnapshotSender *pSender) {
   SyncSnapshotSend *pMsg = rpcMsg.pCont;
   pMsg->srcId = pSender->pSyncNode->myRaftId;
   pMsg->destId = pSender->pSyncNode->replicasId[pSender->replicaIndex];
-  pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
+  pMsg->term = pSender->pSyncNode->raftStore.currentTerm;
   pMsg->beginIndex = pSender->snapshotParam.start;
   pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
   pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
@@ -314,7 +314,7 @@ SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, SRaftId from
   pReceiver->pWriter = NULL;
   pReceiver->pSyncNode = pSyncNode;
   pReceiver->fromId = fromId;
-  pReceiver->term = pSyncNode->pRaftStore->currentTerm;
+  pReceiver->term = pSyncNode->raftStore.currentTerm;
   pReceiver->snapshot.data = NULL;
   pReceiver->snapshot.lastApplyIndex = SYNC_INDEX_INVALID;
   pReceiver->snapshot.lastApplyTerm = 0;
@@ -380,7 +380,7 @@ void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncSnapshotSend *p
 
   pReceiver->start = true;
   pReceiver->ack = SYNC_SNAPSHOT_SEQ_PRE_SNAPSHOT;
-  pReceiver->term = pReceiver->pSyncNode->pRaftStore->currentTerm;
+  pReceiver->term = pReceiver->pSyncNode->raftStore.currentTerm;
   pReceiver->fromId = pPreMsg->srcId;
   pReceiver->startTime = pPreMsg->startTime;
 
@@ -437,9 +437,9 @@ static int32_t snapshotReceiverFinish(SSyncSnapshotReceiver *pReceiver, SyncSnap
     }
 
     // maybe update term
-    if (pReceiver->snapshot.lastApplyTerm > pReceiver->pSyncNode->pRaftStore->currentTerm) {
+    if (pReceiver->snapshot.lastApplyTerm > pReceiver->pSyncNode->raftStore.currentTerm) {
-      pReceiver->pSyncNode->pRaftStore->currentTerm = pReceiver->snapshot.lastApplyTerm;
+      pReceiver->pSyncNode->raftStore.currentTerm = pReceiver->snapshot.lastApplyTerm;
-      raftStorePersist(pReceiver->pSyncNode->pRaftStore);
+      (void)raftStoreWriteFile(pReceiver->pSyncNode);
     }
 
     // stop writer, apply data
@@ -592,7 +592,7 @@ _SEND_REPLY:
   SyncSnapshotRsp *pRspMsg = rpcMsg.pCont;
   pRspMsg->srcId = pSyncNode->myRaftId;
   pRspMsg->destId = pMsg->srcId;
-  pRspMsg->term = pSyncNode->pRaftStore->currentTerm;
+  pRspMsg->term = pSyncNode->raftStore.currentTerm;
   pRspMsg->lastIndex = pMsg->lastIndex;
   pRspMsg->lastTerm = pMsg->lastTerm;
   pRspMsg->startTime = pReceiver->startTime;
@@ -648,7 +648,7 @@ _SEND_REPLY:
   SyncSnapshotRsp *pRspMsg = rpcMsg.pCont;
   pRspMsg->srcId = pSyncNode->myRaftId;
   pRspMsg->destId = pMsg->srcId;
-  pRspMsg->term = pSyncNode->pRaftStore->currentTerm;
+  pRspMsg->term = pSyncNode->raftStore.currentTerm;
   pRspMsg->lastIndex = pMsg->lastIndex;
   pRspMsg->lastTerm = pMsg->lastTerm;
   pRspMsg->startTime = pReceiver->startTime;
@@ -698,7 +698,7 @@ static int32_t syncNodeOnSnapshotReceive(SSyncNode *pSyncNode, SyncSnapshotSend
   SyncSnapshotRsp *pRspMsg = rpcMsg.pCont;
   pRspMsg->srcId = pSyncNode->myRaftId;
   pRspMsg->destId = pMsg->srcId;
-  pRspMsg->term = pSyncNode->pRaftStore->currentTerm;
+  pRspMsg->term = pSyncNode->raftStore.currentTerm;
   pRspMsg->lastIndex = pMsg->lastIndex;
   pRspMsg->lastTerm = pMsg->lastTerm;
   pRspMsg->startTime = pReceiver->startTime;
@@ -745,7 +745,7 @@ static int32_t syncNodeOnSnapshotEnd(SSyncNode *pSyncNode, SyncSnapshotSend *pMs
   SyncSnapshotRsp *pRspMsg = rpcMsg.pCont;
   pRspMsg->srcId = pSyncNode->myRaftId;
   pRspMsg->destId = pMsg->srcId;
-  pRspMsg->term = pSyncNode->pRaftStore->currentTerm;
+  pRspMsg->term = pSyncNode->raftStore.currentTerm;
   pRspMsg->lastIndex = pMsg->lastIndex;
   pRspMsg->lastTerm = pMsg->lastTerm;
   pRspMsg->startTime = pReceiver->startTime;
@@ -794,13 +794,13 @@ int32_t syncNodeOnSnapshot(SSyncNode *pSyncNode, const SRpcMsg *pRpcMsg) {
     return -1;
   }
 
-  if (pMsg->term < pSyncNode->pRaftStore->currentTerm) {
+  if (pMsg->term < pSyncNode->raftStore.currentTerm) {
     syncLogRecvSyncSnapshotSend(pSyncNode, pMsg, "reject since small term");
     terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
     return -1;
   }
 
-  if (pMsg->term > pSyncNode->pRaftStore->currentTerm) {
+  if (pMsg->term > pSyncNode->raftStore.currentTerm) {
     syncNodeStepDown(pSyncNode, pMsg->term);
   }
   syncNodeResetElectTimer(pSyncNode);
@@ -808,7 +808,7 @@ int32_t syncNodeOnSnapshot(SSyncNode *pSyncNode, const SRpcMsg *pRpcMsg) {
   // state, term, seq/ack
   int32_t code = 0;
   if (pSyncNode->state == TAOS_SYNC_STATE_FOLLOWER) {
-    if (pMsg->term == pSyncNode->pRaftStore->currentTerm) {
+    if (pMsg->term == pSyncNode->raftStore.currentTerm) {
       if (pMsg->seq == SYNC_SNAPSHOT_SEQ_PRE_SNAPSHOT) {
         syncLogRecvSyncSnapshotSend(pSyncNode, pMsg, "process seq pre-snapshot");
         code = syncNodeOnSnapshotPre(pSyncNode, pMsg);
@@ -892,7 +892,7 @@ static int32_t syncNodeOnSnapshotPreRsp(SSyncNode *pSyncNode, SSyncSnapshotSende
   SyncSnapshotSend *pSendMsg = rpcMsg.pCont;
   pSendMsg->srcId = pSender->pSyncNode->myRaftId;
   pSendMsg->destId = pSender->pSyncNode->replicasId[pSender->replicaIndex];
-  pSendMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
+  pSendMsg->term = pSender->pSyncNode->raftStore.currentTerm;
   pSendMsg->beginIndex = pSender->snapshotParam.start;
   pSendMsg->lastIndex = pSender->snapshot.lastApplyIndex;
   pSendMsg->lastTerm = pSender->snapshot.lastApplyTerm;
@@ -951,10 +951,10 @@ int32_t syncNodeOnSnapshotRsp(SSyncNode *pSyncNode, const SRpcMsg *pRpcMsg) {
     goto _ERROR;
   }
 
-  if (pMsg->term != pSyncNode->pRaftStore->currentTerm) {
+  if (pMsg->term != pSyncNode->raftStore.currentTerm) {
     syncLogRecvSyncSnapshotRsp(pSyncNode, pMsg, "snapshot sender and receiver term not match");
     sSError(pSender, "snapshot sender term not equal, msg term:%" PRId64 " currentTerm:%" PRId64, pMsg->term,
-            pSyncNode->pRaftStore->currentTerm);
+            pSyncNode->raftStore.currentTerm);
     terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
     goto _ERROR;
   }

source/libs/sync/src/syncUtil.c

@@ -17,20 +17,20 @@
 #include "syncUtil.h"
 #include "syncIndexMgr.h"
 #include "syncMessage.h"
+#include "syncPipeline.h"
 #include "syncRaftCfg.h"
 #include "syncRaftStore.h"
 #include "syncSnapshot.h"
 
 void syncCfg2SimpleStr(const SSyncCfg* pCfg, char* buf, int32_t bufLen) {
-  int32_t len = snprintf(buf, bufLen, "{r-num:%d, my:%d, ", pCfg->replicaNum, pCfg->myIndex);
+  int32_t len = snprintf(buf, bufLen, "{num:%d, as:%d, [", pCfg->replicaNum, pCfg->myIndex);
-
   for (int32_t i = 0; i < pCfg->replicaNum; ++i) {
+    len += snprintf(buf + len, bufLen - len, "%s:%d", pCfg->nodeInfo[i].nodeFqdn, pCfg->nodeInfo[i].nodePort);
     if (i < pCfg->replicaNum - 1) {
-      len += snprintf(buf + len, bufLen - len, "%s:%d, ", pCfg->nodeInfo[i].nodeFqdn, pCfg->nodeInfo[i].nodePort);
+      len += snprintf(buf + len, bufLen - len, "%s", ", ");
-    } else {
-      len += snprintf(buf + len, bufLen - len, "%s:%d}", pCfg->nodeInfo[i].nodeFqdn, pCfg->nodeInfo[i].nodePort);
     }
   }
+  len += snprintf(buf + len, bufLen - len, "%s", "]}");
 }
 
 void syncUtilNodeInfo2EpSet(const SNodeInfo* pInfo, SEpSet* pEpSet) {
@@ -89,32 +89,55 @@ bool syncUtilUserRollback(tmsg_t msgType) { return msgType != TDMT_SYNC_NOOP &&
 
 // for leader
 static void syncHearbeatReplyTime2Str(SSyncNode* pSyncNode, char* buf, int32_t bufLen) {
-  int32_t len = 5;
+  int32_t len = 0;
-
+  len += snprintf(buf + len, bufLen - len, "%s", "{");
   for (int32_t i = 0; i < pSyncNode->replicaNum; ++i) {
     int64_t tsMs = syncIndexMgrGetRecvTime(pSyncNode->pMatchIndex, &(pSyncNode->replicasId[i]));
-
+    len += snprintf(buf + len, bufLen - len, "%d:%" PRId64, i, tsMs);
     if (i < pSyncNode->replicaNum - 1) {
-      len += snprintf(buf + len, bufLen - len, "%d:%" PRId64 ",", i, tsMs);
+      len += snprintf(buf + len, bufLen - len, "%s", ",");
-    } else {
-      len += snprintf(buf + len, bufLen - len, "%d:%" PRId64 "}", i, tsMs);
     }
   }
+  len += snprintf(buf + len, bufLen - len, "%s", "}");
 }
 
 // for follower
 static void syncHearbeatTime2Str(SSyncNode* pSyncNode, char* buf, int32_t bufLen) {
-  int32_t len = 4;
+  int32_t len = 0;
-
+  len += snprintf(buf + len, bufLen - len, "%s", "{");
   for (int32_t i = 0; i < pSyncNode->replicaNum; ++i) {
     int64_t tsMs = syncIndexMgrGetRecvTime(pSyncNode->pNextIndex, &(pSyncNode->replicasId[i]));
-
+    len += snprintf(buf + len, bufLen - len, "%d:%" PRId64, i, tsMs);
     if (i < pSyncNode->replicaNum - 1) {
-      len += snprintf(buf + len, bufLen - len, "%d:%" PRId64 ",", i, tsMs);
+      len += snprintf(buf + len, bufLen - len, "%s", ",");
-    } else {
-      len += snprintf(buf + len, bufLen - len, "%d:%" PRId64 "}", i, tsMs);
     }
   }
+  len += snprintf(buf + len, bufLen - len, "%s", "}");
+}
+
+static void syncLogBufferStates2Str(SSyncNode* pSyncNode, char* buf, int32_t bufLen) {
+  SSyncLogBuffer* pBuf = pSyncNode->pLogBuf;
+  if (pBuf == NULL) {
+    return;
+  }
+  int len = 0;
+  len += snprintf(buf + len, bufLen - len, "[%" PRId64 " %" PRId64 " %" PRId64 ", %" PRId64 ")", pBuf->startIndex,
+                  pBuf->commitIndex, pBuf->matchIndex, pBuf->endIndex);
+}
+
+static void syncLogReplMgrStates2Str(SSyncNode* pSyncNode, char* buf, int32_t bufLen) {
+  int len = 0;
+  len += snprintf(buf + len, bufLen - len, "%s", "{");
+  for (int32_t i = 0; i < pSyncNode->replicaNum; i++) {
+    SSyncLogReplMgr* pMgr = pSyncNode->logReplMgrs[i];
+    if (pMgr == NULL) break;
+    len += snprintf(buf + len, bufLen - len, "%d:%d [%" PRId64 " %" PRId64 ", %" PRId64 ")", i, pMgr->restored,
+                    pMgr->startIndex, pMgr->matchIndex, pMgr->endIndex);
+    if (i + 1 < pSyncNode->replicaNum) {
+      len += snprintf(buf + len, bufLen - len, "%s", ", ");
+    }
+  }
+  len += snprintf(buf + len, bufLen - len, "%s", "}");
 }
 
 static void syncPeerState2Str(SSyncNode* pSyncNode, char* buf, int32_t bufLen) {
@@ -135,8 +158,8 @@ static void syncPeerState2Str(SSyncNode* pSyncNode, char* buf, int32_t bufLen) {
 }
 
 void syncPrintNodeLog(const char* flags, ELogLevel level, int32_t dflag, SSyncNode* pNode, const char* format, ...) {
-  if (pNode == NULL || pNode->pRaftStore == NULL || pNode->pLogStore == NULL) return;
+  if (pNode == NULL || pNode->pLogStore == NULL) return;
-  int64_t currentTerm = pNode->pRaftStore->currentTerm;
+  int64_t currentTerm = pNode->raftStore.currentTerm;
 
   // save error code, otherwise it will be overwritten
   int32_t errCode = terrno;
@@ -156,16 +179,19 @@ void syncPrintNodeLog(const char* flags, ELogLevel level, int32_t dflag, SSyncNo
   int32_t cacheHit = pNode->pLogStore->cacheHit;
   int32_t cacheMiss = pNode->pLogStore->cacheMiss;
 
-  char cfgStr[1024];
+  char cfgStr[1024] = "";
   syncCfg2SimpleStr(&pNode->raftCfg.cfg, cfgStr, sizeof(cfgStr));
 
-  char peerStr[1024] = "{";
+  char replMgrStatesStr[1024] = "";
-  syncPeerState2Str(pNode, peerStr, sizeof(peerStr));
+  syncLogReplMgrStates2Str(pNode, replMgrStatesStr, sizeof(replMgrStatesStr));
 
-  char hbrTimeStr[256] = "hbr:{";
+  char bufferStatesStr[256] = "";
+  syncLogBufferStates2Str(pNode, bufferStatesStr, sizeof(bufferStatesStr));
+
+  char hbrTimeStr[256] = "";
   syncHearbeatReplyTime2Str(pNode, hbrTimeStr, sizeof(hbrTimeStr));
 
-  char hbTimeStr[256] = "hb:{";
+  char hbTimeStr[256] = "";
   syncHearbeatTime2Str(pNode, hbTimeStr, sizeof(hbTimeStr));
 
   char    eventLog[512];  // {0};
@@ -186,23 +212,23 @@ void syncPrintNodeLog(const char* flags, ELogLevel level, int32_t dflag, SSyncNo
     taosPrintLog(flags, level, dflag,
                  "vgId:%d, %s, sync:%s, term:%" PRIu64 ", commit-index:%" PRId64 ", first-ver:%" PRId64
                  ", last-ver:%" PRId64 ", min:%" PRId64 ", snap:%" PRId64 ", snap-term:%" PRIu64
-                 ", elect-times:%d, as-leader-times:%d, cfg-ch-times:%d, hit:%d, mis:%d, hb-slow:%d, hbr-slow:%d, "
+                 ", elect-times:%d, as-leader-times:%d, cfg-ch-times:%d, hb-slow:%d, hbr-slow:%d, "
                  "aq-items:%d, snaping:%" PRId64 ", replicas:%d, last-cfg:%" PRId64
-                 ", chging:%d, restore:%d, quorum:%d, elect-lc-timer:%" PRId64 ", hb:%" PRId64 ", %s, %s, %s, %s",
+                 ", chging:%d, restore:%d, quorum:%d, elect-lc-timer:%" PRId64 ", hb:%" PRId64
+                 ", buffer:%s, repl-mgrs:%s, members:%s, hb:%s, hb-reply:%s",
                  pNode->vgId, eventLog, syncStr(pNode->state), currentTerm, pNode->commitIndex, logBeginIndex,
                  logLastIndex, pNode->minMatchIndex, snapshot.lastApplyIndex, snapshot.lastApplyTerm, pNode->electNum,
-                 pNode->becomeLeaderNum, pNode->configChangeNum, cacheHit, cacheMiss, pNode->hbSlowNum,
+                 pNode->becomeLeaderNum, pNode->configChangeNum, pNode->hbSlowNum, pNode->hbrSlowNum, aqItems,
-                 pNode->hbrSlowNum, aqItems, pNode->snapshottingIndex, pNode->replicaNum,
+                 pNode->snapshottingIndex, pNode->replicaNum, pNode->raftCfg.lastConfigIndex, pNode->changing,
-                 pNode->raftCfg.lastConfigIndex, pNode->changing, pNode->restoreFinish, syncNodeDynamicQuorum(pNode),
+                 pNode->restoreFinish, syncNodeDynamicQuorum(pNode), pNode->electTimerLogicClock, pNode->heartbeatTimerLogicClockUser,
-                 pNode->electTimerLogicClock, pNode->heartbeatTimerLogicClockUser, peerStr, cfgStr, hbTimeStr,
+                 bufferStatesStr, replMgrStatesStr, cfgStr, hbTimeStr, hbrTimeStr);
-                 hbrTimeStr);
   }
 }
 
 void syncPrintSnapshotSenderLog(const char* flags, ELogLevel level, int32_t dflag, SSyncSnapshotSender* pSender,
                                 const char* format, ...) {
   SSyncNode* pNode = pSender->pSyncNode;
-  if (pNode == NULL || pNode->pRaftStore == NULL || pNode->pLogStore == NULL) return;
+  if (pNode == NULL || pNode->pLogStore == NULL) return;
 
   SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
   if (pNode->pFsm != NULL && pNode->pFsm->FpGetSnapshotInfo != NULL) {
@@ -216,7 +242,7 @@ void syncPrintSnapshotSenderLog(const char* flags, ELogLevel level, int32_t dfla
     logBeginIndex = pNode->pLogStore->syncLogBeginIndex(pNode->pLogStore);
   }
 
-  char cfgStr[1024];
+  char cfgStr[1024] = "";
   syncCfg2SimpleStr(&pNode->raftCfg.cfg, cfgStr, sizeof(cfgStr));
 
   char peerStr[1024] = "{";
@@ -238,7 +264,7 @@ void syncPrintSnapshotSenderLog(const char* flags, ELogLevel level, int32_t dfla
                pNode->vgId, eventLog, syncStr(pNode->state), pSender, pSender->snapshotParam.start,
                pSender->snapshotParam.end, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
                pSender->snapshot.lastConfigIndex, pSender->seq, pSender->ack, pSender->finish, pSender->replicaIndex,
-               DID(&pNode->replicasId[pSender->replicaIndex]), pNode->pRaftStore->currentTerm, pNode->commitIndex,
+               DID(&pNode->replicasId[pSender->replicaIndex]), pNode->raftStore.currentTerm, pNode->commitIndex,
                logBeginIndex, logLastIndex, pNode->minMatchIndex, snapshot.lastApplyIndex, snapshot.lastApplyTerm,
                pNode->raftCfg.isStandBy, pNode->raftCfg.snapshotStrategy, pNode->raftCfg.batchSize, pNode->replicaNum,
                pNode->raftCfg.lastConfigIndex, pNode->changing, pNode->restoreFinish, syncNodeDynamicQuorum(pNode),
@@ -248,7 +274,7 @@ void syncPrintSnapshotSenderLog(const char* flags, ELogLevel level, int32_t dfla
 void syncPrintSnapshotReceiverLog(const char* flags, ELogLevel level, int32_t dflag, SSyncSnapshotReceiver* pReceiver,
                                   const char* format, ...) {
   SSyncNode* pNode = pReceiver->pSyncNode;
-  if (pNode == NULL || pNode->pRaftStore == NULL || pNode->pLogStore == NULL) return;
+  if (pNode == NULL || pNode->pLogStore == NULL) return;
 
   SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
   if (pNode->pFsm != NULL && pNode->pFsm->FpGetSnapshotInfo != NULL) {
@@ -262,7 +288,7 @@ void syncPrintSnapshotReceiverLog(const char* flags, ELogLevel level, int32_t df
     logBeginIndex = pNode->pLogStore->syncLogBeginIndex(pNode->pLogStore);
   }
 
-  char cfgStr[1024];
+  char cfgStr[1024] = "";
   syncCfg2SimpleStr(&pNode->raftCfg.cfg, cfgStr, sizeof(cfgStr));
 
   char peerStr[1024] = "{";
@@ -285,7 +311,7 @@ void syncPrintSnapshotReceiverLog(const char* flags, ELogLevel level, int32_t df
                pNode->vgId, eventLog, syncStr(pNode->state), pReceiver, pReceiver->start, pReceiver->ack,
                pReceiver->term, pReceiver->startTime, DID(&pReceiver->fromId), pReceiver->snapshotParam.start,
                pReceiver->snapshotParam.end, pReceiver->snapshot.lastApplyIndex, pReceiver->snapshot.lastApplyTerm,
-               pReceiver->snapshot.lastConfigIndex, pNode->pRaftStore->currentTerm, pNode->commitIndex, logBeginIndex,
+               pReceiver->snapshot.lastConfigIndex, pNode->raftStore.currentTerm, pNode->commitIndex, logBeginIndex,
                logLastIndex, pNode->minMatchIndex, snapshot.lastApplyIndex, snapshot.lastApplyTerm,
                pNode->raftCfg.isStandBy, pNode->raftCfg.snapshotStrategy, pNode->raftCfg.batchSize, pNode->replicaNum,
                pNode->raftCfg.lastConfigIndex, pNode->changing, pNode->restoreFinish, syncNodeDynamicQuorum(pNode),
@@ -304,7 +330,7 @@ void syncLogRecvTimer(SSyncNode* pSyncNode, const SyncTimeout* pMsg, const char*
 
 void syncLogRecvLocalCmd(SSyncNode* pSyncNode, const SyncLocalCmd* pMsg, const char* s) {
   sNTrace(pSyncNode, "recv sync-local-cmd {cmd:%d-%s, sd-new-term:%" PRId64 ", fc-index:%" PRId64 "}, %s", pMsg->cmd,
-          syncLocalCmdGetStr(pMsg->cmd), pMsg->sdNewTerm, pMsg->fcIndex, s);
+          syncLocalCmdGetStr(pMsg->cmd), pMsg->currentTerm, pMsg->commitIndex, s);
 }
 
 void syncLogSendAppendEntriesReply(SSyncNode* pSyncNode, const SyncAppendEntriesReply* pMsg, const char* s) {

source/libs/sync/test/syncLocalCmdTest.cpp

@@ -16,8 +16,8 @@ SyncLocalCmd *createMsg() {
   pMsg->srcId.vgId = 100;
   pMsg->destId.addr = syncUtilAddr2U64("127.0.0.1", 5678);
   pMsg->destId.vgId = 100;
-  pMsg->sdNewTerm = 123;
+  // pMsg->sdNewTerm = 123;
-  pMsg->fcIndex = 456;
+  // pMsg->fcIndex = 456;
   pMsg->cmd = SYNC_LOCAL_CMD_STEP_DOWN;
 
   return pMsg;

source/libs/sync/test/syncRaftStoreTest.cpp

@@ -33,35 +33,35 @@ int main() {
 
   initRaftId();
 
-  SRaftStore* pRaftStore = raftStoreOpen("./test_raft_store.json");
+  // SRaftStore* pRaftStore = raftStoreOpen("./test_raft_store.json");
-  assert(pRaftStore != NULL);
+  // assert(pRaftStore != NULL);
-  raftStoreLog2((char*)"==raftStoreOpen==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreOpen==", pRaftStore);
 
-  raftStoreSetTerm(pRaftStore, 100);
+  // raftStoreSetTerm(pRaftStore, 100);
-  raftStoreLog2((char*)"==raftStoreSetTerm==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreSetTerm==", pRaftStore);
 
-  raftStoreVote(pRaftStore, &ids[0]);
+  // raftStoreVote(pRaftStore, &ids[0]);
-  raftStoreLog2((char*)"==raftStoreVote==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreVote==", pRaftStore);
 
-  raftStoreClearVote(pRaftStore);
+  // raftStoreClearVote(pRaftStore);
-  raftStoreLog2((char*)"==raftStoreClearVote==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreClearVote==", pRaftStore);
 
-  raftStoreVote(pRaftStore, &ids[1]);
+  // raftStoreVote(pRaftStore, &ids[1]);
-  raftStoreLog2((char*)"==raftStoreVote==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreVote==", pRaftStore);
 
-  raftStoreNextTerm(pRaftStore);
+  // raftStoreNextTerm(pRaftStore);
-  raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
 
-  raftStoreNextTerm(pRaftStore);
+  // raftStoreNextTerm(pRaftStore);
-  raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
 
-  raftStoreNextTerm(pRaftStore);
+  // raftStoreNextTerm(pRaftStore);
-  raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
 
-  raftStoreNextTerm(pRaftStore);
+  // raftStoreNextTerm(pRaftStore);
-  raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
+  // raftStoreLog2((char*)"==raftStoreNextTerm==", pRaftStore);
 
-  raftStoreClose(pRaftStore);
+  // raftStoreClose(pRaftStore);
 
   return 0;
 }

source/libs/sync/test/syncSnapshotReceiverTest.cpp

@@ -29,7 +29,7 @@ int32_t SnapshotDoWrite(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_
 
 SSyncSnapshotReceiver* createReceiver() {
   SSyncNode* pSyncNode = (SSyncNode*)taosMemoryMalloc(sizeof(*pSyncNode));
-  pSyncNode->pRaftStore = (SRaftStore*)taosMemoryMalloc(sizeof(*(pSyncNode->pRaftStore)));
+  // pSyncNode->pRaftStore = (SRaftStore*)taosMemoryMalloc(sizeof(*(pSyncNode->pRaftStore)));
   pSyncNode->pFsm = (SSyncFSM*)taosMemoryMalloc(sizeof(*(pSyncNode->pFsm)));
 
 #if 0 

source/libs/sync/test/syncSnapshotSenderTest.cpp

@@ -29,7 +29,7 @@ int32_t SnapshotDoWrite(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_
 
 SSyncSnapshotSender* createSender() {
   SSyncNode* pSyncNode = (SSyncNode*)taosMemoryMalloc(sizeof(*pSyncNode));
-  pSyncNode->pRaftStore = (SRaftStore*)taosMemoryMalloc(sizeof(*(pSyncNode->pRaftStore)));
+  // pSyncNode->pRaftStore = (SRaftStore*)taosMemoryMalloc(sizeof(*(pSyncNode->pRaftStore)));
   pSyncNode->pFsm = (SSyncFSM*)taosMemoryMalloc(sizeof(*(pSyncNode->pFsm)));
 
 #if 0 

source/libs/sync/test/sync_test_lib/src/syncMainDebug.c

@@ -80,7 +80,7 @@ cJSON* syncNode2Json(const SSyncNode* pSyncNode) {
     // tla+ server vars
     cJSON_AddNumberToObject(pRoot, "state", pSyncNode->state);
     cJSON_AddStringToObject(pRoot, "state_str", syncStr(pSyncNode->state));
-    cJSON_AddItemToObject(pRoot, "pRaftStore", raftStore2Json(pSyncNode->pRaftStore));
+    // cJSON_AddItemToObject(pRoot, "pRaftStore", raftStore2Json(&pSyncNode.raftStore));
 
     // tla+ candidate vars
     cJSON_AddItemToObject(pRoot, "pVotesGranted", voteGranted2Json(pSyncNode->pVotesGranted));
@@ -199,7 +199,7 @@ inline char* syncNode2SimpleStr(const SSyncNode* pSyncNode) {
            ", sby:%d, "
            "r-num:%d, "
            "lcfg:%" PRId64 ", chging:%d, rsto:%d",
-           pSyncNode->vgId, syncStr(pSyncNode->state), pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex,
+           pSyncNode->vgId, syncStr(pSyncNode->state), pSyncNode->raftStore.currentTerm, pSyncNode->commitIndex,
            logBeginIndex, logLastIndex, snapshot.lastApplyIndex, pSyncNode->raftCfg.isStandBy, pSyncNode->replicaNum,
            pSyncNode->raftCfg.lastConfigIndex, pSyncNode->changing, pSyncNode->restoreFinish);
 

source/libs/sync/test/sync_test_lib/src/syncMessageDebug.c

@@ -2858,11 +2858,11 @@ cJSON* syncLocalCmd2Json(const SyncLocalCmd* pMsg) {
 
     cJSON_AddNumberToObject(pRoot, "cmd", pMsg->cmd);
 
-    snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pMsg->sdNewTerm);
+    // snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pMsg->sdNewTerm);
-    cJSON_AddStringToObject(pRoot, "sd-new-term", u64buf);
+    // cJSON_AddStringToObject(pRoot, "sd-new-term", u64buf);
 
-    snprintf(u64buf, sizeof(u64buf), "%" PRId64, pMsg->fcIndex);
+    // snprintf(u64buf, sizeof(u64buf), "%" PRId64, pMsg->fcIndex);
-    cJSON_AddStringToObject(pRoot, "fc-index", u64buf);
+    // cJSON_AddStringToObject(pRoot, "fc-index", u64buf);
   }
 
   cJSON* pJson = cJSON_CreateObject();

source/libs/sync/test/sync_test_lib/src/syncRaftStoreDebug.c

@@ -41,8 +41,8 @@ cJSON *raftStore2Json(SRaftStore *pRaftStore) {
     cJSON_AddNumberToObject(pVoteFor, "vgId", pRaftStore->voteFor.vgId);
     cJSON_AddItemToObject(pRoot, "voteFor", pVoteFor);
 
-    int hasVoted = raftStoreHasVoted(pRaftStore);
+    // int hasVoted = raftStoreHasVoted(pRaftStore);
-    cJSON_AddNumberToObject(pRoot, "hasVoted", hasVoted);
+    // cJSON_AddNumberToObject(pRoot, "hasVoted", hasVoted);
   }
 
   cJSON *pJson = cJSON_CreateObject();

source/libs/sync/test/sync_test_lib/src/syncSnapshotDebug.c

@@ -137,7 +137,7 @@ int32_t syncNodeOnPreSnapshot(SSyncNode *ths, SyncPreSnapshot *pMsg) {
   SyncPreSnapshotReply *pMsgReply = syncPreSnapshotReplyBuild(ths->vgId);
   pMsgReply->srcId = ths->myRaftId;
   pMsgReply->destId = pMsg->srcId;
-  pMsgReply->term = ths->pRaftStore->currentTerm;
+  pMsgReply->term = ths->raftStore.currentTerm;
 
   SSyncLogStoreData *pData = ths->pLogStore->data;
   SWal              *pWal = pData->pWal;

source/libs/tdb/inc/tdb.h

@@ -74,7 +74,12 @@ int32_t tdbTbcUpsert(TBC *pTbc, const void *pKey, int nKey, const void *pData, i
 
 int32_t tdbTxnOpen(TXN *pTxn, int64_t txnid, void *(*xMalloc)(void *, size_t), void (*xFree)(void *, void *),
                    void *xArg, int flags);
-int32_t tdbTxnClose(TXN *pTxn);
+int32_t tdbTxnCloseImpl(TXN *pTxn);
+#define tdbTxnClose(pTxn)  \
+  do {                     \
+    tdbTxnCloseImpl(pTxn); \
+    (pTxn) = NULL;         \
+  } while (0)
 
 // other
 void tdbFree(void *);

source/libs/tdb/src/db/tdbTxn.c

@@ -28,13 +28,18 @@ int tdbTxnOpen(TXN *pTxn, int64_t txnid, void *(*xMalloc)(void *, size_t), void
   return 0;
 }
 
-int tdbTxnClose(TXN *pTxn) {
+int tdbTxnCloseImpl(TXN *pTxn) {
   if (pTxn) {
     if (pTxn->jPageSet) {
       hashset_destroy(pTxn->jPageSet);
       pTxn->jPageSet = NULL;
     }
 
+    if (pTxn->jfd) {
+      tdbOsClose(pTxn->jfd);
+      ASSERT(pTxn->jfd == NULL);
+    }
+
     tdbOsFree(pTxn);
   }
 

source/libs/transport/inc/transComm.h

@@ -101,13 +101,6 @@ typedef void* queue[2];
 #define TRANS_PACKET_LIMIT 1024 * 1024 * 512
 
 #define TRANS_MAGIC_NUM           0x5f375a86
-
-#define TRANS_NOVALID_PACKET(src) ((src) != TRANS_MAGIC_NUM ? 1 : 0)
-
-#define TRANS_PACKET_LIMIT 1024 * 1024 * 512
-
-#define TRANS_MAGIC_NUM 0x5f375a86
-
 #define TRANS_NOVALID_PACKET(src) ((src) != TRANS_MAGIC_NUM ? 1 : 0)
 
 typedef SRpcMsg      STransMsg;

source/libs/transport/src/thttp.c

@@ -35,6 +35,7 @@ typedef struct SHttpModule {
 typedef struct SHttpMsg {
   queue         q;
   char*         server;
+  char*         uri;
   int32_t       port;
   char*         cont;
   int32_t       len;
@@ -63,26 +64,26 @@ static void    httpHandleReq(SHttpMsg* msg);
 static void    httpHandleQuit(SHttpMsg* msg);
 static int32_t httpSendQuit();
 
-static int32_t taosSendHttpReportImpl(const char* server, uint16_t port, char* pCont, int32_t contLen,
+static int32_t taosSendHttpReportImpl(const char* server, const char* uri, uint16_t port, char* pCont, int32_t contLen,
                                       EHttpCompFlag flag);
 
-static int32_t taosBuildHttpHeader(const char* server, int32_t contLen, char* pHead, int32_t headLen,
+static int32_t taosBuildHttpHeader(const char* server, const char* uri, int32_t contLen, char* pHead, int32_t headLen,
                                    EHttpCompFlag flag) {
   if (flag == HTTP_FLAT) {
     return snprintf(pHead, headLen,
-                    "POST /report HTTP/1.1\n"
+                    "POST %s HTTP/1.1\n"
                     "Host: %s\n"
                     "Content-Type: application/json\n"
                     "Content-Length: %d\n\n",
-                    server, contLen);
+                    uri, server, contLen);
   } else if (flag == HTTP_GZIP) {
     return snprintf(pHead, headLen,
-                    "POST /report HTTP/1.1\n"
+                    "POST %s HTTP/1.1\n"
                     "Host: %s\n"
                     "Content-Type: application/json\n"
                     "Content-Encoding: gzip\n"
                     "Content-Length: %d\n\n",
-                    server, contLen);
+                    uri, server, contLen);
   } else {
     terrno = TSDB_CODE_INVALID_CFG;
     return -1;
@@ -181,6 +182,7 @@ static void httpDestroyMsg(SHttpMsg* msg) {
   if (msg == NULL) return;
 
   taosMemoryFree(msg->server);
+  taosMemoryFree(msg->uri);
   taosMemoryFree(msg->cont);
   taosMemoryFree(msg);
 }
@@ -293,10 +295,11 @@ int32_t httpSendQuit() {
   return 0;
 }
 
-static int32_t taosSendHttpReportImpl(const char* server, uint16_t port, char* pCont, int32_t contLen,
+static int32_t taosSendHttpReportImpl(const char* server, const char* uri, uint16_t port, char* pCont, int32_t contLen,
                                       EHttpCompFlag flag) {
   SHttpMsg* msg = taosMemoryMalloc(sizeof(SHttpMsg));
   msg->server = strdup(server);
+  msg->uri  = strdup(uri);
   msg->port = port;
   msg->cont = taosMemoryMalloc(contLen);
   memcpy(msg->cont, pCont, contLen);
@@ -309,12 +312,10 @@ static int32_t taosSendHttpReportImpl(const char* server, uint16_t port, char* p
     httpDestroyMsg(msg);
     tError("http-report already released");
     return -1;
-  } else {
-    msg->http = load;
-    transAsyncSend(load->asyncPool, &(msg->q));
   }
   
-  return 0;
+  msg->http = load;
+  return transAsyncSend(load->asyncPool, &(msg->q));
 }
 
 static void httpDestroyClientCb(uv_handle_t* handle) {
@@ -360,7 +361,7 @@ static void httpHandleReq(SHttpMsg* msg) {
 
   int32_t len = 2048;
   char*   header = taosMemoryCalloc(1, len);
-  int32_t headLen = taosBuildHttpHeader(msg->server, msg->len, header, len, msg->flag);
+  int32_t headLen = taosBuildHttpHeader(msg->server, msg->uri, msg->len, header, len, msg->flag);
   if (headLen < 0) {
     taosMemoryFree(header);
     goto END;
@@ -380,6 +381,7 @@ static void httpHandleReq(SHttpMsg* msg) {
   cli->port = msg->port;
   cli->dest = dest;
 
+  taosMemoryFree(msg->uri);
   taosMemoryFree(msg);
 
   uv_tcp_init(http->loop, &cli->tcp);
@@ -406,9 +408,9 @@ END:
   httpDestroyMsg(msg);
 }
 
-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) {
   taosThreadOnce(&transHttpInit, transHttpEnvInit);
-  return taosSendHttpReportImpl(server, port, pCont, contLen, flag);
+  return taosSendHttpReportImpl(server, uri, port, pCont, contLen, flag);
 }
 
 static void transHttpEnvInit() {

source/libs/transport/src/transCli.c

@@ -1126,7 +1126,7 @@ void cliHandleReq(SCliMsg* pMsg, SCliThrd* pThrd) {
 
     int ret = uv_tcp_connect(&conn->connReq, (uv_tcp_t*)(conn->stream), (const struct sockaddr*)&addr, cliConnCb);
     if (ret != 0) {
-      tGTrace("%s conn %p failed to connect to %s:%d, reason:%s", pTransInst->label, conn, conn->ip, conn->port,
+      tGError("%s conn %p failed to connect to %s:%d, reason:%s", pTransInst->label, conn, conn->ip, conn->port,
               uv_err_name(ret));
 
       uv_timer_stop(conn->timer);

source/libs/wal/src/walMeta.c

@@ -292,19 +292,10 @@ void walAlignVersions(SWal* pWal) {
     }
     pWal->vers.lastVer = pWal->vers.snapshotVer;
   }
-  if (pWal->vers.commitVer < pWal->vers.snapshotVer) {
+  // reset commitVer and appliedVer
-    wWarn("vgId:%d, commitVer:%" PRId64 " is less than snapshotVer:%" PRId64 ". align with it.", pWal->cfg.vgId,
-          pWal->vers.commitVer, pWal->vers.snapshotVer);
   pWal->vers.commitVer = pWal->vers.snapshotVer;
-  }
-  if (pWal->vers.appliedVer < pWal->vers.snapshotVer) {
-    wWarn("vgId:%d, appliedVer:%" PRId64 " is less than snapshotVer:%" PRId64 ". align with it.", pWal->cfg.vgId,
-          pWal->vers.appliedVer, pWal->vers.snapshotVer);
   pWal->vers.appliedVer = pWal->vers.snapshotVer;
-  }
+  wInfo("vgId:%d, reset commitVer to %" PRId64, pWal->cfg.vgId, pWal->vers.commitVer);
-
-  pWal->vers.commitVer = TMIN(pWal->vers.lastVer, pWal->vers.commitVer);
-  pWal->vers.appliedVer = TMIN(pWal->vers.commitVer, pWal->vers.appliedVer);
 }
 
 bool walLogEntriesComplete(const SWal* pWal) {

source/libs/wal/src/walWrite.c

@@ -105,7 +105,7 @@ int32_t walRollback(SWal *pWal, int64_t ver) {
   wInfo("vgId:%d, wal rollback for version %" PRId64, pWal->cfg.vgId, ver);
   int64_t code;
   char    fnameStr[WAL_FILE_LEN];
-  if (ver > pWal->vers.lastVer || ver < pWal->vers.commitVer || ver <= pWal->vers.snapshotVer) {
+  if (ver > pWal->vers.lastVer || ver <= pWal->vers.commitVer || ver <= pWal->vers.snapshotVer) {
     terrno = TSDB_CODE_WAL_INVALID_VER;
     taosThreadMutexUnlock(&pWal->mutex);
     return -1;
@@ -635,6 +635,7 @@ int32_t walWrite(SWal *pWal, int64_t index, tmsg_t msgType, const void *body, in
 }
 
 void walFsync(SWal *pWal, bool forceFsync) {
+  taosThreadMutexLock(&pWal->mutex);
   if (forceFsync || (pWal->cfg.level == TAOS_WAL_FSYNC && pWal->cfg.fsyncPeriod == 0)) {
     wTrace("vgId:%d, fileId:%" PRId64 ".idx, do fsync", pWal->cfg.vgId, walGetCurFileFirstVer(pWal));
     if (taosFsyncFile(pWal->pIdxFile) < 0) {
@@ -647,4 +648,5 @@ void walFsync(SWal *pWal, bool forceFsync) {
              strerror(errno));
     }
   }
+  taosThreadMutexUnlock(&pWal->mutex);
 }