[TD-114] let create table wait until master selected

src/inc/mnode.h

@@ -149,7 +149,8 @@ typedef struct _vg_obj {
   int32_t         lbDnodeId;
   int32_t         lbTime;
   int8_t          status;
-  int8_t          reserved[14];
+  int8_t          inUse;
+  int8_t          reserved[13];
   int8_t          updateEnd[1];
   int32_t         refCount;
   struct _vg_obj *prev, *next;
@@ -243,6 +244,8 @@ typedef struct {
   int8_t   received;
   int8_t   successed;
   int8_t   expected;
+  int8_t   retry;
+  int8_t   maxRetry;
   int32_t  contLen;
   int32_t  code;
   void     *ahandle;

src/inc/taosmsg.h

@@ -510,7 +510,6 @@ typedef struct SRetrieveTableRsp {
 
 typedef struct {
   int32_t vgId;
-  int32_t vnode;
   int64_t totalStorage;
   int64_t compStorage;
   int64_t pointsWritten;

src/mnode/inc/mgmtShell.h

@@ -31,6 +31,7 @@ void mgmtAddShellShowMetaHandle(uint8_t showType, SShowMetaFp fp);
 void mgmtAddShellShowRetrieveHandle(uint8_t showType, SShowRetrieveFp fp);
 
 void mgmtAddToShellQueue(SQueuedMsg *queuedMsg);
+void mgmtDealyedAddToShellQueue(SQueuedMsg *queuedMsg);
 void mgmtSendSimpleResp(void *thandle, int32_t code);
 
 #ifdef __cplusplus

src/mnode/inc/mgmtVgroup.h

@@ -37,6 +37,7 @@ void    mgmtDropAllVgroups(SDbObj *pDropDb);
 
 void *  mgmtGetNextVgroup(void *pNode, SVgObj **pVgroup);
 void    mgmtUpdateVgroup(SVgObj *pVgroup);
+void    mgmtUpdateVgroupStatus(SVgObj *pVgroup, int32_t dnodeId, SVnodeLoad *pVload);
 
 void    mgmtCreateVgroup(SQueuedMsg *pMsg, SDbObj *pDb);
 void    mgmtDropVgroup(SVgObj *pVgroup, void *ahandle);

src/mnode/src/mgmtDnode.c

@@ -189,18 +189,21 @@ void clusterProcessDnodeStatusMsg(SRpcMsg *rpcMsg) {
  
   int32_t openVnodes = htons(pStatus->openVnodes);
   for (int32_t j = 0; j < openVnodes; ++j) {
-    pDnode->vload[j].vgId          = htonl(pStatus->load[j].vgId);
-    pDnode->vload[j].totalStorage  = htobe64(pStatus->load[j].totalStorage);
-    pDnode->vload[j].compStorage   = htobe64(pStatus->load[j].compStorage);
-    pDnode->vload[j].pointsWritten = htobe64(pStatus->load[j].pointsWritten);
+    SVnodeLoad *pVload = &pStatus->load[j];
+    pDnode->vload[j].vgId          = htonl(pVload->vgId);
+    pDnode->vload[j].totalStorage  = htobe64(pVload->totalStorage);
+    pDnode->vload[j].compStorage   = htobe64(pVload->compStorage);
+    pDnode->vload[j].pointsWritten = htobe64(pVload->pointsWritten);
     
     SVgObj *pVgroup = mgmtGetVgroup(pDnode->vload[j].vgId);
     if (pVgroup == NULL) {
       SRpcIpSet ipSet = mgmtGetIpSetFromIp(pDnode->privateIp);
       mPrint("dnode:%d, vgroup:%d not exist in mnode, drop it", pDnode->dnodeId, pDnode->vload[j].vgId);
       mgmtSendDropVnodeMsg(pDnode->vload[j].vgId, &ipSet, NULL);
+    } else {
+      mgmtUpdateVgroupStatus(pVgroup, pDnode->dnodeId, pVload);
+      mgmtReleaseVgroup(pVgroup);
     }
-    mgmtReleaseVgroup(pVgroup);
   }
 
   if (pDnode->status == TAOS_DN_STATUS_OFFLINE) {

src/mnode/src/mgmtProfile.c

@@ -806,6 +806,8 @@ void* mgmtCloneQueuedMsg(SQueuedMsg *pSrcMsg) {
   pDestMsg->msgType = pSrcMsg->msgType;
   pDestMsg->pCont   = pSrcMsg->pCont;
   pDestMsg->contLen = pSrcMsg->contLen;
+  pDestMsg->retry   = pSrcMsg->retry;
+  pDestMsg->maxRetry= pSrcMsg->maxRetry;
   pDestMsg->pUser   = pSrcMsg->pUser;
   pDestMsg->usePublicIp = pSrcMsg->usePublicIp;
 

src/mnode/src/mgmtShell.c

@@ -128,6 +128,15 @@ void mgmtAddToShellQueue(SQueuedMsg *queuedMsg) {
   taosScheduleTask(tsMgmtTranQhandle, &schedMsg);
 }
 
+static void mgmtDoDealyedAddToShellQueue(void *param, void *tmrId) {
+  mgmtAddToShellQueue(param);
+}
+
+void mgmtDealyedAddToShellQueue(SQueuedMsg *queuedMsg) {
+  void *unUsed = NULL;
+  taosTmrReset(mgmtDoDealyedAddToShellQueue, 1000, queuedMsg, tsMgmtTmr, &unUsed);
+}
+
 static void mgmtProcessMsgFromShell(SRpcMsg *rpcMsg) {
   if (rpcMsg == NULL || rpcMsg->pCont == NULL) {
     return;

src/mnode/src/mgmtTable.c

@@ -540,7 +540,7 @@ static void mgmtProcessCreateTableMsg(SQueuedMsg *pMsg) {
   SCMCreateTableMsg *pCreate = pMsg->pCont;
   
   pMsg->pTable = mgmtGetTable(pCreate->tableId);
-  if (pMsg->pTable != NULL) {
+  if (pMsg->pTable != NULL && pMsg->retry == 0) {
     if (pCreate->igExists) {
       mTrace("table:%s, is already exist", pCreate->tableId);
       mgmtSendSimpleResp(pMsg->thandle, TSDB_CODE_SUCCESS);
@@ -1300,7 +1300,11 @@ static void mgmtProcessCreateChildTableMsg(SQueuedMsg *pMsg) {
     return;
   }
 
-  pMsg->pTable = (STableObj *)mgmtDoCreateChildTable(pCreate, pVgroup, sid);
+  if (pMsg->retry == 0) {
+    pMsg->pTable = (STableObj *)mgmtDoCreateChildTable(pCreate, pVgroup, sid);
+  } else {
+    pMsg->pTable = mgmtGetTable(pCreate->tableId);
+  }
   if (pMsg->pTable == NULL) {
     mgmtSendSimpleResp(pMsg->thandle, terrno);
     return;
@@ -1315,6 +1319,7 @@ static void mgmtProcessCreateChildTableMsg(SQueuedMsg *pMsg) {
   SRpcIpSet ipSet = mgmtGetIpSetFromVgroup(pVgroup);
   SQueuedMsg *newMsg = mgmtCloneQueuedMsg(pMsg);
   newMsg->ahandle = pMsg->pTable;
+  newMsg->maxRetry = 5;
   mgmtIncTableRef(pMsg->pTable);
   SRpcMsg rpcMsg = {
       .handle  = newMsg,
@@ -1737,30 +1742,40 @@ static void mgmtProcessCreateChildTableRsp(SRpcMsg *rpcMsg) {
   queueMsg->received++;
 
   SChildTableObj *pTable = queueMsg->ahandle;
-  mTrace("table:%s, create table rsp received, thandle:%p ahandle:%p result:%s", pTable->info.tableId, queueMsg->thandle,
-         rpcMsg->handle, tstrerror(rpcMsg->code));
+  mTrace("table:%s, create table rsp received, thandle:%p result:%s", pTable->info.tableId, queueMsg->thandle,
+         tstrerror(rpcMsg->code));
 
   if (rpcMsg->code != TSDB_CODE_SUCCESS) {
-    SSdbOperDesc oper = {
-      .type = SDB_OPER_TYPE_GLOBAL,
-      .table = tsChildTableSdb,
-      .pObj = pTable
-    };
-    sdbDeleteRow(&oper);
+    if (queueMsg->retry++ < queueMsg->maxRetry) {
+      mTrace("table:%s, create table rsp received, retry:%d thandle:%p result:%s", pTable->info.tableId,
+             queueMsg->retry, queueMsg->thandle, tstrerror(rpcMsg->code));
+      mgmtDealyedAddToShellQueue(queueMsg);
+    } else {
+      mError("table:%s, failed to create in dnode, thandle:%p result:%s", pTable->info.tableId,
+             queueMsg->thandle, tstrerror(rpcMsg->code));
+      
+      SSdbOperDesc oper = {
+        .type = SDB_OPER_TYPE_GLOBAL,
+        .table = tsChildTableSdb,
+        .pObj = pTable
+      };
+      sdbDeleteRow(&oper);
     
-    mError("table:%s, failed to create in dnode, reason:%s", pTable->info.tableId, tstrerror(rpcMsg->code));
-    mgmtSendSimpleResp(queueMsg->thandle, rpcMsg->code);
+      mgmtSendSimpleResp(queueMsg->thandle, rpcMsg->code);
+      mgmtFreeQueuedMsg(queueMsg);
+    }    
   } else {
-    mTrace("table:%s, created in dnode", pTable->info.tableId);
+    mTrace("table:%s, created in dnode, thandle:%p result:%s", pTable->info.tableId, queueMsg->thandle,
+           tstrerror(rpcMsg->code));
+
     if (queueMsg->msgType != TSDB_MSG_TYPE_CM_CREATE_TABLE) {
       mTrace("table:%s, start to get meta", pTable->info.tableId);
-      mgmtAddToShellQueue(mgmtCloneQueuedMsg(queueMsg));
+      mgmtAddToShellQueue(queueMsg);
     } else {
       mgmtSendSimpleResp(queueMsg->thandle, rpcMsg->code);
+      mgmtFreeQueuedMsg(queueMsg);
     }
   }
-
-  mgmtFreeQueuedMsg(queueMsg);
 }
 
 // not implemented yet

src/mnode/src/mgmtVgroup.c

@@ -18,6 +18,7 @@
 #include "taoserror.h"
 #include "tlog.h"
 #include "tbalance.h"
+#include "tsync.h"
 #include "tcluster.h"
 #include "mnode.h"
 #include "mgmtDb.h"
@@ -209,6 +210,18 @@ void mgmtUpdateVgroup(SVgObj *pVgroup) {
   mgmtSendCreateVgroupMsg(pVgroup, NULL);
 }
 
+void mgmtUpdateVgroupStatus(SVgObj *pVgroup, int32_t dnodeId, SVnodeLoad *pVload) {
+  if (pVload->role == TAOS_SYNC_ROLE_MASTER) {
+    for (int32_t i = 0; i < pVgroup->numOfVnodes; ++i) {
+      SVnodeGid *pVgid = &pVgroup->vnodeGid[i];
+      if (pVgid->dnodeId == dnodeId) {
+        pVgroup->inUse = i;
+        break;
+      }
+    }
+  }
+}
+
 SVgObj *mgmtGetAvailableVgroup(SDbObj *pDb) {
   return pDb->pHead;
 }

src/vnode/main/src/vnodeMain.c

@@ -285,7 +285,6 @@ static void vnodeBuildVloadMsg(char *pNode, void * param) {
 
   SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++];
   pLoad->vgId = htonl(pVnode->vgId);
-  pLoad->vnode = htonl(pVnode->vgId);
   pLoad->status = pVnode->status;
   pLoad->role = pVnode->role;
 }

src/vnode/main/src/vnodeWrite.c

@@ -57,7 +57,8 @@ int32_t vnodeProcessWrite(void *param1, int qtype, void *param2, void *item) {
     if (pVnode->status != TAOS_VN_STATUS_READY) 
       return TSDB_CODE_NOT_ACTIVE_VNODE;
 
-    // if (pVnode->replica > 1 && pVnode->role != TAOS_SYNC_ROLE_MASTER)
+    if (pVnode->syncCfg.replica > 1 && pVnode->role != TAOS_SYNC_ROLE_MASTER)
+      return TSDB_CODE_NO_MASTER;
 
     // assign version
     pVnode->version++;

tests/script/tmp/prepare.sim

 system sh/stop_dnodes.sh
 system sh/deploy.sh -n dnode1 -m 192.168.0.1 -i 192.168.0.1
-system sh/deploy.sh -n dnode2 -m 192.168.0.1 -i 192.168.0.2
\ No newline at end of file
+system sh/deploy.sh -n dnode2 -m 192.168.0.1 -i 192.168.0.2
+system sh/deploy.sh -n dnode3 -m 192.168.0.1 -i 192.168.0.3
\ No newline at end of file