Merge pull request #3482 from taosdata/feature/crash

Feature/crash

src/dnode/src/dnodeModule.c

@@ -146,8 +146,8 @@ void dnodeProcessModuleStatus(uint32_t moduleStatus) {
   }
 }
 
-void dnodeCheckModules() {
-  if (tsModuleStatus & TSDB_MOD_MNODE) return;
+bool dnodeCheckMnodeStarting() {
+  if (tsModuleStatus & TSDB_MOD_MNODE) return false;
 
   SDMMnodeInfos *mnodes = dnodeGetMnodeInfos();
   for (int32_t i = 0; i < mnodes->nodeNum; ++i) {
@@ -156,7 +156,9 @@ void dnodeCheckModules() {
       uint32_t moduleStatus = tsModuleStatus | (1 << TSDB_MOD_MNODE);;
       dInfo("start mnode module, module status:%d, new status:%d", tsModuleStatus, moduleStatus);
       dnodeProcessModuleStatus(moduleStatus);
-      break;
+      return true;
     }
   }
+
+  return false;
 }

src/inc/dnode.h

@@ -43,7 +43,7 @@ void    dnodeGetMnodeEpSetForPeer(void *epSet);
 void    dnodeGetMnodeEpSetForShell(void *epSet);
 void *  dnodeGetMnodeInfos();
 int32_t dnodeGetDnodeId();
-void    dnodeCheckModules();
+bool    dnodeCheckMnodeStarting();
 
 void  dnodeAddClientRspHandle(uint8_t msgType, void (*fp)(SRpcMsg *rpcMsg));
 void  dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg);

src/mnode/src/mnodeSdb.c

@@ -298,8 +298,9 @@ static void sdbUpdateSyncTmrFp(void *param, void *tmrId) { sdbUpdateSync(); }
 void sdbUpdateSync() {
   if (!mnodeIsRunning()) {
     mDebug("mnode not start yet, update sync info later");
-    dnodeCheckModules();
+    if (dnodeCheckMnodeStarting()) {
       taosTmrReset(sdbUpdateSyncTmrFp, 1000, NULL, tsMnodeTmr, &tsUpdateSyncTmr);
+    }
     return;
   }
   mDebug("update sync info in sdb");

tests/script/unique/arbitrator/dn3_mn1_r2_vnode_delDir.sim

@@ -53,7 +53,7 @@ system sh/exec.sh -n dnode2 -s start
 system sh/exec.sh -n dnode3 -s start
 sql create dnode $hostname2
 sql create dnode $hostname3
-sleep 3000
+sleep 5000
 
 $sleepTimer = 3000
 
@@ -225,6 +225,7 @@ if $data00 != $totalRows then
 endi
 
 print ============== step5: stop dnode2, and remove its vnode
+sleep 5000
 system sh/exec.sh -n dnode2 -s stop -x SIGINT
 sleep $sleepTimer
 

tests/script/unique/arbitrator/replica_changeWithArbitrator.sim

@@ -193,6 +193,7 @@ if $data00 != $totalRows then
 endi
 
 print ============== step5: stop dnode1
+sleep 5000
 system sh/exec.sh -n dnode1 -s stop
 sleep 3000
 

tests/script/unique/dnode/alternativeRole.sim

@@ -18,13 +18,14 @@ system sh/cfg.sh -n dnode3 -c numOfMnodes -v 3
 
 print ========== step1
 system sh/exec.sh -n dnode1 -s start
+sleep 3000
 sql connect
+
 sql create dnode $hostname2
 system sh/exec.sh -n dnode2 -s start
-sleep 3000
 sql create dnode $hostname3
 system sh/exec.sh -n dnode3 -s start
-sleep 3000
+sleep 5000
 
 sql show dnodes
 print dnode1 $data5_1