refactor(sync): add debug log

include/libs/sync/sync.h

@@ -47,8 +47,9 @@ typedef enum {
 typedef enum {
   TAOS_SYNC_PROPOSE_SUCCESS = 0,
   TAOS_SYNC_PROPOSE_NOT_LEADER = 1,
-  TAOS_SYNC_PROPOSE_OTHER_ERROR = 2,
-  TAOS_SYNC_ONLY_ONE_REPLICA = 3,
+  TAOS_SYNC_ONLY_ONE_REPLICA = 2,
+  TAOS_SYNC_NOT_IN_NEW_CONFIG = 3,
+  TAOS_SYNC_OTHER_ERROR = 100,
 } ESyncProposeCode;
 
 typedef enum {
@@ -199,15 +200,13 @@ bool        syncIsRestoreFinish(int64_t rid);
 int32_t     syncGetSnapshotMeta(int64_t rid, struct SSnapshotMeta* sMeta);
 
 int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg);
-int32_t syncReconfigRaw(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg);
+
+// build SRpcMsg, need to call syncPropose with SRpcMsg
+int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg);
 
 int32_t syncLeaderTransfer(int64_t rid);
 int32_t syncLeaderTransferTo(int64_t rid, SNodeInfo newLeader);
 
-// to be moved to static
-void syncStartNormal(int64_t rid);
-void syncStartStandBy(int64_t rid);
-
 #ifdef __cplusplus
 }
 #endif

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

@@ -382,22 +382,22 @@ void mndStop(SMnode *pMnode) {
 int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
   SMnode    *pMnode = pMsg->info.node;
   SSyncMgmt *pMgmt = &pMnode->syncMgmt;
-  int32_t    code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+  int32_t    code = TAOS_SYNC_OTHER_ERROR;
 
   if (!syncEnvIsStart()) {
     mError("failed to process sync msg:%p type:%s since syncEnv stop", pMsg, TMSG_INFO(pMsg->msgType));
-    return TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    return TAOS_SYNC_OTHER_ERROR;
   }
 
   SSyncNode *pSyncNode = syncNodeAcquire(pMgmt->sync);
   if (pSyncNode == NULL) {
     mError("failed to process sync msg:%p type:%s since syncNode is null", pMsg, TMSG_INFO(pMsg->msgType));
-    return TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    return TAOS_SYNC_OTHER_ERROR;
   }
 
   if (mndAcquireSyncRef(pMnode) != 0) {
     mError("failed to process sync msg:%p type:%s since %s", pMsg, TMSG_INFO(pMsg->msgType), terrstr());
-    return TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    return TAOS_SYNC_OTHER_ERROR;
   }
 
   char  logBuf[512] = {0};
@@ -457,7 +457,7 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
 
     } else {
       mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
-      code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+      code = TAOS_SYNC_OTHER_ERROR;
     }
 
   } else {
@@ -495,7 +495,7 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
       syncAppendEntriesReplyDestroy(pSyncMsg);
     } else {
       mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
-      code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+      code = TAOS_SYNC_OTHER_ERROR;
     }
   }
 

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

@@ -236,7 +236,7 @@ int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) {
     tsem_wait(&pMgmt->syncSem);
   } else if (code == TAOS_SYNC_PROPOSE_NOT_LEADER) {
     terrno = TSDB_CODE_APP_NOT_READY;
-  } else if (code == TAOS_SYNC_PROPOSE_OTHER_ERROR) {
+  } else if (code == TAOS_SYNC_OTHER_ERROR) {
     terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
   } else {
     terrno = TSDB_CODE_APP_ERROR;
@@ -254,13 +254,16 @@ int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) {
 void mndSyncStart(SMnode *pMnode) {
   SSyncMgmt *pMgmt = &pMnode->syncMgmt;
   syncSetMsgCb(pMgmt->sync, &pMnode->msgCb);
+  syncStart(pMgmt->sync);
+  mDebug("mnode sync started, id:%" PRId64 " standby:%d", pMgmt->sync, pMgmt->standby);
 
+/*
   if (pMgmt->standby) {
     syncStartStandBy(pMgmt->sync);
   } else {
     syncStart(pMgmt->sync);
   }
-  mDebug("mnode sync started, id:%" PRId64 " standby:%d", pMgmt->sync, pMgmt->standby);
+*/  
 }
 
 void mndSyncStop(SMnode *pMnode) {}

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

@@ -291,7 +291,7 @@ void vnodeUpdateMetaRsp(SVnode *pVnode, STableMetaRsp *pMetaRsp) {
 }
 
 int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
-  int32_t ret = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+  int32_t ret = TAOS_SYNC_OTHER_ERROR;
 
   if (syncEnvIsStart()) {
     SSyncNode *pSyncNode = syncNodeAcquire(pVnode->sync);
@@ -372,13 +372,13 @@ int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
 
     } else {
       vError("==vnodeProcessSyncReq== error msg type:%d", pRpcMsg->msgType);
-      ret = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+      ret = TAOS_SYNC_OTHER_ERROR;
     }
 
     syncNodeRelease(pSyncNode);
   } else {
     vError("==vnodeProcessSyncReq== error syncEnv stop");
-    ret = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    ret = TAOS_SYNC_OTHER_ERROR;
   }
 
   return ret;

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

@@ -290,11 +290,14 @@ int32_t vnodeSyncOpen(SVnode *pVnode, char *path) {
 
 void vnodeSyncStart(SVnode *pVnode) {
   syncSetMsgCb(pVnode->sync, &pVnode->msgCb);
+  syncStart(pVnode->sync);
+  /*
   if (pVnode->config.standby) {
     syncStartStandBy(pVnode->sync);
   } else {
     syncStart(pVnode->sync);
   }
+  */
 }
 
 void vnodeSyncClose(SVnode *pVnode) { syncStop(pVnode->sync); }

source/libs/sync/inc/syncInt.h

@@ -168,6 +168,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo);
 void       syncNodeStart(SSyncNode* pSyncNode);
 void       syncNodeStartStandBy(SSyncNode* pSyncNode);
 void       syncNodeClose(SSyncNode* pSyncNode);
+int32_t    syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak);
 
 // option
 bool syncNodeSnapshotEnable(SSyncNode* pSyncNode);
@@ -232,6 +233,9 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
 bool                 syncNodeInRaftGroup(SSyncNode* ths, SRaftId* pRaftId);
 SSyncSnapshotSender* syncNodeGetSnapshotSender(SSyncNode* ths, SRaftId* pDestId);
 
+void syncStartNormal(int64_t rid);
+void syncStartStandBy(int64_t rid);
+
 // for debug --------------
 void syncNodePrint(SSyncNode* pObj);
 void syncNodePrint2(char* s, SSyncNode* pObj);

source/libs/sync/src/syncMain.c

@@ -149,12 +149,12 @@ void syncStop(int64_t rid) {
 int32_t syncSetStandby(int64_t rid) {
   SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
   if (pSyncNode == NULL) {
-    return -1;
+    return TAOS_SYNC_OTHER_ERROR;
   }
 
   if (pSyncNode->state != TAOS_SYNC_STATE_LEADER) {
     taosReleaseRef(tsNodeRefId, pSyncNode->rid);
-    return -1;
+    return TAOS_SYNC_OTHER_ERROR;
   }
 
   // state change
@@ -173,14 +173,68 @@ int32_t syncSetStandby(int64_t rid) {
   return 0;
 }
 
-int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) {
+int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg) {
+  SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
+  if (pSyncNode == NULL) {
+    return TAOS_SYNC_OTHER_ERROR;
+  }
+  ASSERT(rid == pSyncNode->rid);
+
   int32_t ret = 0;
-  char*   newconfig = syncCfg2Str((SSyncCfg*)pSyncCfg);
+  bool    IamInNew = false;
+  for (int i = 0; i < pNewCfg->replicaNum; ++i) {
+    SRaftId newId;
+    newId.addr = syncUtilAddr2U64((pNewCfg->nodeInfo)[i].nodeFqdn, (pNewCfg->nodeInfo)[i].nodePort);
+    newId.vgId = pSyncNode->vgId;
+    if (syncUtilSameId(&(pSyncNode->myRaftId), &newId)) {
+      IamInNew = true;
+    }
+  }
+  if (!IamInNew) {
+    taosReleaseRef(tsNodeRefId, pSyncNode->rid);
+    return TAOS_SYNC_NOT_IN_NEW_CONFIG;
+  }
+
+  char* newconfig = syncCfg2Str((SSyncCfg*)pNewCfg);
+  pRpcMsg->msgType = TDMT_SYNC_CONFIG_CHANGE;
+  pRpcMsg->info.noResp = 1;
+  pRpcMsg->contLen = strlen(newconfig) + 1;
+  pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
+  snprintf(pRpcMsg->pCont, pRpcMsg->contLen, "%s", newconfig);
+  taosMemoryFree(newconfig);
+
+  taosReleaseRef(tsNodeRefId, pSyncNode->rid);
+  return ret;
+}
 
+int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg) {
+  SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
+  if (pSyncNode == NULL) {
+    return TAOS_SYNC_OTHER_ERROR;
+  }
+  ASSERT(rid == pSyncNode->rid);
+
+  bool IamInNew = false;
+  for (int i = 0; i < pNewCfg->replicaNum; ++i) {
+    SRaftId newId;
+    newId.addr = syncUtilAddr2U64((pNewCfg->nodeInfo)[i].nodeFqdn, (pNewCfg->nodeInfo)[i].nodePort);
+    newId.vgId = pSyncNode->vgId;
+    if (syncUtilSameId(&(pSyncNode->myRaftId), &newId)) {
+      IamInNew = true;
+    }
+  }
+  if (!IamInNew) {
+    taosReleaseRef(tsNodeRefId, pSyncNode->rid);
+    return TAOS_SYNC_NOT_IN_NEW_CONFIG;
+  }
+
+  char* newconfig = syncCfg2Str((SSyncCfg*)pNewCfg);
   if (gRaftDetailLog) {
     sInfo("==syncReconfig== newconfig:%s", newconfig);
   }
 
+  int32_t ret = 0;
+
   SRpcMsg rpcMsg = {0};
   rpcMsg.msgType = TDMT_SYNC_CONFIG_CHANGE;
   rpcMsg.info.noResp = 1;
@@ -188,27 +242,42 @@ int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) {
   rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
   snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", newconfig);
   taosMemoryFree(newconfig);
-  ret = syncPropose(rid, &rpcMsg, false);
+  ret = syncNodePropose(pSyncNode, &rpcMsg, false);
+
+  taosReleaseRef(tsNodeRefId, pSyncNode->rid);
   return ret;
 }
 
 int32_t syncLeaderTransfer(int64_t rid) {
-  int32_t ret = 0;
+  SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
+  if (pSyncNode == NULL) {
+    return TAOS_SYNC_OTHER_ERROR;
+  }
+  ASSERT(rid == pSyncNode->rid);
 
+  if (pSyncNode->peersNum > 0) {
+    taosReleaseRef(tsNodeRefId, pSyncNode->rid);
+    return TAOS_SYNC_OTHER_ERROR;
+  }
+
+  SNodeInfo newLeader = (pSyncNode->peersNodeInfo)[0];
+  taosReleaseRef(tsNodeRefId, pSyncNode->rid);
+
+  int32_t ret = syncLeaderTransferTo(rid, newLeader);
   return ret;
 }
 
 int32_t syncLeaderTransferTo(int64_t rid, SNodeInfo newLeader) {
   SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
   if (pSyncNode == NULL) {
-    return false;
+    return TAOS_SYNC_OTHER_ERROR;
   }
-  assert(rid == pSyncNode->rid);
+  ASSERT(rid == pSyncNode->rid);
   int32_t ret = 0;
 
   if (pSyncNode->replicaNum == 1) {
-    taosReleaseRef(tsNodeRefId, pSyncNode->rid);
     sError("only one replica, cannot drop leader");
+    taosReleaseRef(tsNodeRefId, pSyncNode->rid);
     return TAOS_SYNC_ONLY_ONE_REPLICA;
   }
 
@@ -220,26 +289,11 @@ int32_t syncLeaderTransferTo(int64_t rid, SNodeInfo newLeader) {
   syncLeaderTransfer2RpcMsg(pMsg, &rpcMsg);
   syncLeaderTransferDestroy(pMsg);
 
-  ret = syncPropose(rid, &rpcMsg, false);
-
+  ret = syncNodePropose(pSyncNode, &rpcMsg, false);
   taosReleaseRef(tsNodeRefId, pSyncNode->rid);
   return ret;
 }
 
-int32_t syncReconfigRaw(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg) {
-  int32_t ret = 0;
-  char*   newconfig = syncCfg2Str((SSyncCfg*)pNewCfg);
-
-  pRpcMsg->msgType = TDMT_SYNC_CONFIG_CHANGE;
-  pRpcMsg->info.noResp = 1;
-  pRpcMsg->contLen = strlen(newconfig) + 1;
-  pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
-  snprintf(pRpcMsg->pCont, pRpcMsg->contLen, "%s", newconfig);
-  taosMemoryFree(newconfig);
-
-  return ret;
-}
-
 bool syncCanLeaderTransfer(int64_t rid) {
   SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
   if (pSyncNode == NULL) {
@@ -272,8 +326,6 @@ bool syncCanLeaderTransfer(int64_t rid) {
   return matchOK;
 }
 
-int32_t syncGiveUpLeader(int64_t rid) { return 0; }
-
 int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
   int32_t ret = syncPropose(rid, pMsg, isWeak);
   return ret;
@@ -467,10 +519,19 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
 
   SSyncNode* pSyncNode = taosAcquireRef(tsNodeRefId, rid);
   if (pSyncNode == NULL) {
-    return TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    return TAOS_SYNC_OTHER_ERROR;
   }
   assert(rid == pSyncNode->rid);
   sDebug("vgId:%d sync event propose msgType:%s", pSyncNode->vgId, TMSG_INFO(pMsg->msgType));
+  ret = syncNodePropose(pSyncNode, pMsg, isWeak);
+
+  taosReleaseRef(tsNodeRefId, pSyncNode->rid);
+  return ret;
+}
+
+int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak) {
+  int32_t ret = TAOS_SYNC_PROPOSE_SUCCESS;
+  sDebug("vgId:%d sync event propose msgType:%s", pSyncNode->vgId, TMSG_INFO(pMsg->msgType));
 
   if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
     SRespStub stub;
@@ -485,15 +546,14 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
     if (pSyncNode->FpEqMsg != NULL && (*pSyncNode->FpEqMsg)(pSyncNode->msgcb, &rpcMsg) == 0) {
       ret = TAOS_SYNC_PROPOSE_SUCCESS;
     } else {
-      sTrace("syncPropose pSyncNode->FpEqMsg is NULL");
+      sError("syncPropose pSyncNode->FpEqMsg is NULL");
     }
     syncClientRequestDestroy(pSyncMsg);
   } else {
-    sTrace("syncPropose not leader, %s", syncUtilState2String(pSyncNode->state));
+    sError("syncPropose not leader, %s", syncUtilState2String(pSyncNode->state));
     ret = TAOS_SYNC_PROPOSE_NOT_LEADER;
   }
 
-  taosReleaseRef(tsNodeRefId, pSyncNode->rid);
   return ret;
 }