fix/block-check-to-async-check

include/libs/sync/sync.h

@@ -250,6 +250,7 @@ void    syncPreStop(int64_t rid);
 void    syncPostStop(int64_t rid);
 int32_t syncPropose(int64_t rid, SRpcMsg* pMsg, bool isWeak, int64_t* seq);
 int32_t syncIsCatchUp(int64_t rid);
+ESyncRole syncGetRole(int64_t rid);
 int32_t syncProcessMsg(int64_t rid, SRpcMsg* pMsg);
 int32_t syncReconfig(int64_t rid, SSyncCfg* pCfg);
 int32_t syncBeginSnapshot(int64_t rid, int64_t lastApplyIndex);

include/util/taoserror.h

@@ -437,6 +437,8 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_VND_STOPPED                   TAOS_DEF_ERROR_CODE(0, 0x0529)
 #define TSDB_CODE_VND_DUP_REQUEST               TAOS_DEF_ERROR_CODE(0, 0x0530)
 #define TSDB_CODE_VND_QUERY_BUSY                TAOS_DEF_ERROR_CODE(0, 0x0531)
+#define TSDB_CODE_VND_NOT_CATCH_UP              TAOS_DEF_ERROR_CODE(0, 0x0532) // internal
+#define TSDB_CODE_VND_ALREADY_IS_VOTER          TAOS_DEF_ERROR_CODE(0, 0x0533) // internal
 
 // tsdb
 #define TSDB_CODE_TDB_INVALID_TABLE_ID          TAOS_DEF_ERROR_CODE(0, 0x0600)

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

@@ -336,13 +336,21 @@ int32_t vmProcessAlterVnodeTypeReq(SVnodeMgmt *pMgmt, SRpcMsg *pMsg) {
 
   SVnodeObj *pVnode = vmAcquireVnode(pMgmt, req.vgId);
   if (pVnode == NULL) {
-    dError("vgId:%d, failed to alter hashrange since %s", req.vgId, terrstr());
+    dError("vgId:%d, failed to alter vnode type since %s", req.vgId, terrstr());
     terrno = TSDB_CODE_VND_NOT_EXIST;
     return -1;
   }
 
+  if(vnodeGetRole(pVnode->pImpl) == TAOS_SYNC_ROLE_VOTER){
+    terrno = TSDB_CODE_VND_ALREADY_IS_VOTER;
+    vmReleaseVnode(pMgmt, pVnode);
+    return -1;
+  }
+
   dInfo("vgId:%d, checking node catch up", req.vgId);
-  if(vnodeIsCatchUp(pVnode->pImpl) != 0){
+  if(vnodeIsCatchUp(pVnode->pImpl) != 1){
+    terrno = TSDB_CODE_VND_NOT_CATCH_UP;
+    vmReleaseVnode(pMgmt, pVnode);
     return -1;
   }
 
@@ -365,6 +373,7 @@ int32_t vmProcessAlterVnodeTypeReq(SVnodeMgmt *pMgmt, SRpcMsg *pMsg) {
       req.selfIndex >= req.replica || req.learnerSelfIndex >= req.learnerReplica) {
     terrno = TSDB_CODE_INVALID_MSG;
     dError("vgId:%d, failed to alter replica since invalid msg", vgId);
+    vmReleaseVnode(pMgmt, pVnode);
     return -1;
   }
 
@@ -381,6 +390,7 @@ int32_t vmProcessAlterVnodeTypeReq(SVnodeMgmt *pMgmt, SRpcMsg *pMsg) {
     terrno = TSDB_CODE_INVALID_MSG;
     dError("vgId:%d, dnodeId:%d ep:%s:%u not matched with local dnode", vgId, pReplica->id, pReplica->fqdn,
            pReplica->port);
+    vmReleaseVnode(pMgmt, pVnode);
     return -1;
   }
 

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

@@ -1263,6 +1263,8 @@ int32_t mndAddAlterVnodeTypeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb,
   action.pCont = pReq;
   action.contLen = contLen;
   action.msgType = TDMT_DND_ALTER_VNODE_TYPE;
+  action.acceptableCode = TSDB_CODE_VND_ALREADY_IS_VOTER;
+  action.retryCode = TSDB_CODE_VND_NOT_CATCH_UP;
 
   if (mndTransAppendRedoAction(pTrans, &action) != 0) {
     taosMemoryFree(pReq);

source/dnode/vnode/inc/vnode.h

@@ -69,6 +69,7 @@ void    vnodeGetInfo(SVnode *pVnode, const char **dbname, int32_t *vgId);
 int32_t vnodeProcessCreateTSma(SVnode *pVnode, void *pCont, uint32_t contLen);
 int32_t vnodeGetAllTableList(SVnode *pVnode, uint64_t uid, SArray *list);
 int32_t vnodeIsCatchUp(SVnode *pVnode);
+ESyncRole vnodeGetRole(SVnode *pVnode);
 
 int32_t vnodeGetCtbIdList(SVnode *pVnode, int64_t suid, SArray *list);
 int32_t vnodeGetCtbIdListByFilter(SVnode *pVnode, int64_t suid, SArray *list, bool (*filter)(void *arg), void *arg);

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

@@ -431,6 +431,10 @@ int32_t vnodeIsCatchUp(SVnode *pVnode){
   return syncIsCatchUp(pVnode->sync);
 }
 
+ESyncRole vnodeGetRole(SVnode *pVnode){
+  return syncGetRole(pVnode->sync);
+}
+
 void vnodeStop(SVnode *pVnode) {}
 
 int64_t vnodeGetSyncHandle(SVnode *pVnode) { return pVnode->sync; }

source/libs/sync/src/syncMain.c

@@ -580,25 +580,37 @@ int32_t syncIsCatchUp(int64_t rid) {
     return -1;
   }
 
-  while(1){
+  int32_t isCatchUp = 0;
   if(pSyncNode->pLogBuf->totalIndex < 0 || pSyncNode->pLogBuf->commitIndex < 0 ||
       pSyncNode->pLogBuf->totalIndex < pSyncNode->pLogBuf->commitIndex ||
       pSyncNode->pLogBuf->totalIndex - pSyncNode->pLogBuf->commitIndex > SYNC_LEARNER_CATCHUP){
     sInfo("vgId:%d, Not catch up, wait one second, totalIndex:%" PRId64 " commitIndex:%" PRId64 " matchIndex:%" PRId64, 
                                   pSyncNode->vgId, pSyncNode->pLogBuf->totalIndex, pSyncNode->pLogBuf->commitIndex, 
                                   pSyncNode->pLogBuf->matchIndex);
-      taosSsleep(1);
+    isCatchUp = 0;
   }
   else{
     sInfo("vgId:%d, Catch up, totalIndex:%" PRId64 " commitIndex:%" PRId64 " matchIndex:%" PRId64, 
                                   pSyncNode->vgId, pSyncNode->pLogBuf->totalIndex, pSyncNode->pLogBuf->commitIndex,
                                   pSyncNode->pLogBuf->matchIndex);
-      break;
-    }
+    isCatchUp = 1;
   }
   
   syncNodeRelease(pSyncNode);
-  return 0;
+  return isCatchUp;
+}
+
+ESyncRole syncGetRole(int64_t rid) {
+  SSyncNode* pSyncNode = syncNodeAcquire(rid);
+  if (pSyncNode == NULL) {
+    sError("sync Node Acquire error since %d", errno);
+    return -1;
+  }
+
+  ESyncRole role = pSyncNode->raftCfg.cfg.nodeInfo[pSyncNode->raftCfg.cfg.myIndex].nodeRole;
+  
+  syncNodeRelease(pSyncNode);
+  return role;
 }
 
 int32_t syncNodePropose(SSyncNode* pSyncNode, SRpcMsg* pMsg, bool isWeak, int64_t* seq) {