Merge pull request #13843 from taosdata/fix/mnode

feat: make create db retry

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

@@ -472,7 +472,7 @@ static int32_t mndCreateDb(SMnode *pMnode, SRpcMsg *pReq, SCreateDbReq *pCreate,
   }
 
   int32_t code = -1;
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB, pReq);
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_DB, pReq);
   if (pTrans == NULL) goto _OVER;
 
   mDebug("trans:%d, used to create db:%s", pTrans->id, pCreate->db);

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

@@ -1347,13 +1347,11 @@ int32_t mndKillTrans(SMnode *pMnode, STrans *pTrans) {
 
   for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
     STransAction *pAction = taosArrayGet(pArray, i);
-    if (pAction->errCode != 0) {
-      mInfo("trans:%d, %s:%d set processed for kill msg received, errCode from %s to success", pTrans->id,
-            mndTransStr(pAction->stage), i, tstrerror(pAction->errCode));
-      pAction->msgSent = 1;
-      pAction->msgReceived = 1;
-      pAction->errCode = 0;
-    }
+    mInfo("trans:%d, %s:%d set processed for kill msg received, errCode from %s to success", pTrans->id,
+          mndTransStr(pAction->stage), i, tstrerror(pAction->errCode));
+    pAction->msgSent = 1;
+    pAction->msgReceived = 1;
+    pAction->errCode = 0;
   }
 
   mndTransExecute(pMnode, pTrans);

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

@@ -940,7 +940,7 @@ static int32_t mndAddSetVnodeStandByAction(SMnode *pMnode, STrans *pTrans, SDbOb
 
   action.pCont = pReq;
   action.contLen = contLen;
-  action.msgType = TDMT_DND_DROP_VNODE;
+  action.msgType = TDMT_SYNC_SET_VNODE_STANDBY;
   action.acceptableCode = TSDB_CODE_NODE_NOT_DEPLOYED;
 
   if (isRedo) {

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

@@ -66,7 +66,13 @@ static int32_t vnodeProcessAlterReplicaReq(SVnode *pVnode, SRpcMsg *pMsg) {
     vInfo("vgId:%d, replica:%d %s:%u", TD_VID(pVnode), r, pNode->nodeFqdn, pNode->nodePort);
   }
 
-  return syncReconfig(pVnode->sync, &cfg);
+  SRpcMsg rpcMsg = {.info = pMsg->info};
+  if (syncReconfigBuild(pVnode->sync, &cfg, &rpcMsg) != 0) {
+    vError("vgId:%d, failed to build reconfig msg since %s", TD_VID(pVnode), terrstr());
+    return -1;
+  }
+
+  return syncPropose(pVnode->sync, &rpcMsg, false);
 }
 
 void vnodeProposeMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs) {
@@ -241,6 +247,30 @@ static void vnodeSyncRollBackMsg(SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta
   syncRpcMsgLog2(logBuf, (SRpcMsg *)pMsg);
 }
 
+int32_t vnodeSnapshotStartRead(struct SSyncFSM *pFsm, void **ppReader) {
+  return 0;
+}
+
+int32_t vnodeSnapshotStopRead(struct SSyncFSM *pFsm, void *pReader) {
+  return 0;
+}
+
+int32_t vnodeSnapshotDoRead(struct SSyncFSM *pFsm, void *pReader, void **ppBuf, int32_t *len) {
+  return 0;
+}
+
+int32_t vnodeSnapshotStartWrite(struct SSyncFSM *pFsm, void **ppWriter) {
+  return 0;
+}
+
+int32_t vnodeSnapshotStopWrite(struct SSyncFSM *pFsm, void *pWriter, bool isApply) {
+  return 0;
+}
+
+int32_t vnodeSnapshotDoWrite(struct SSyncFSM *pFsm, void *pWriter, void *pBuf, int32_t len) {
+  return 0;
+}
+
 static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
   SSyncFSM *pFsm = taosMemoryCalloc(1, sizeof(SSyncFSM));
   pFsm->data = pVnode;
@@ -250,6 +280,14 @@ static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
   pFsm->FpGetSnapshot = vnodeSyncGetSnapshot;
   pFsm->FpRestoreFinishCb = NULL;
   pFsm->FpReConfigCb = vnodeSyncReconfig;
+
+  pFsm->FpSnapshotStartRead = vnodeSnapshotStartRead;
+  pFsm->FpSnapshotStopRead = vnodeSnapshotStopRead;
+  pFsm->FpSnapshotDoRead = vnodeSnapshotDoRead;
+  pFsm->FpSnapshotStartWrite = vnodeSnapshotStartWrite;
+  pFsm->FpSnapshotStopWrite = vnodeSnapshotStopWrite;
+  pFsm->FpSnapshotDoWrite = vnodeSnapshotDoWrite;
+
   return pFsm;
 }
 

tests/script/jenkins/basic.txt

@@ -58,7 +58,7 @@
 # ---- mnode
 ./test.sh -f tsim/mnode/basic1.sim
 ./test.sh -f tsim/mnode/basic2.sim
-#./test.sh -f tsim/mnode/basic3.sim
+./test.sh -f tsim/mnode/basic3.sim
 ./test.sh -f tsim/mnode/basic4.sim
 ./test.sh -f tsim/mnode/basic5.sim
 

tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim

@@ -163,26 +163,22 @@ endi
 print =============== step32: move follower2
 print redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
 sql redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
 
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
 sql show d1.tables
 if $rows != 1 then
   return -1
 endi
 
+return
+
 print =============== step33: move follower1
 print redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
 sql redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
 
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
 sql show d1.tables
 if $rows != 1 then
   return -1
@@ -191,12 +187,9 @@ endi
 print =============== step34: move follower2
 print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
 sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
 
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
 sql show d1.tables
 if $rows != 1 then
   return -1
@@ -205,15 +198,8 @@ endi
 print =============== step35: move follower1
 print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
 sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-sql show d1.tables
-if $rows != 1 then
-  return -1
-endi
-=======
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
 
 sql show d1.tables
 if $rows != 1 then
@@ -242,8 +228,6 @@ if $rows != 1 then
   return -1
 endi
 
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
 print =============== step38: move follower2
 print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
 sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
@@ -254,7 +238,6 @@ sql show d1.tables
 if $rows != 1 then
   return -1
 endi
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
 
 print =============== step39: move follower1
 print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1

tests/script/tsim/trans/create_db.sim

@@ -7,44 +7,28 @@ system sh/exec.sh -n dnode1 -s start
 system sh/exec.sh -n dnode2 -s start
 sql connect
 
-print =============== show dnodes
-sql show dnodes;
-if $rows != 1 then 
-  return -1
-endi
-
-if $data00 != 1 then 
-  return -1
-endi
-
-sql show mnodes;
-if $rows != 1 then 
-  return -1
-endi
-
-if $data00 != 1 then 
-  return -1
-endi
-
-if $data02 != leader then 
-  return -1
-endi
-
 print =============== create dnodes
 sql create dnode $hostname port 7200
-sleep 2000
 
-sql show dnodes;
-if $rows != 2 then 
-   return -1
-endi
-
-if $data00 != 1 then 
+$x = 0
+step1:
+	$x = $x + 1
+	sleep 1000
+	if $x == 10 then
+	  print ====> dnode not ready!
+		return -1
+	endi
+sql show dnodes
+print ===> $data00 $data01 $data02 $data03 $data04 $data05
+print ===> $data10 $data11 $data12 $data13 $data14 $data15
+if $rows != 2 then
   return -1
 endi
-
-if $data10 != 2 then 
-  return -1
+if $data(1)[4] != ready then
+  goto step1
+endi
+if $data(2)[4] != ready then
+  goto step1
 endi
 
 print =============== kill dnode2
@@ -68,7 +52,7 @@ if $data[0][0] != 7 then
   return -1
 endi
 
-if $data[0][2] != undoAction then
+if $data[0][2] != redoAction then
   return -1
 endi
 
@@ -80,14 +64,34 @@ sql_error create database d1 vgroups 2;
 
 print =============== start dnode2
 system sh/exec.sh -n dnode2 -s start
-sleep 3000
+
+$x = 0
+step2:
+	$x = $x + 1
+	sleep 1000
+	if $x == 10 then
+	  print ====> dnode not ready!
+		return -1
+	endi
+sql show dnodes
+print ===> $data00 $data01 $data02 $data03 $data04 $data05
+print ===> $data10 $data11 $data12 $data13 $data14 $data15
+if $rows != 2 then
+  return -1
+endi
+if $data(1)[4] != ready then
+  goto step2
+endi
+if $data(2)[4] != ready then
+  goto step2
+endi
 
 sql show transactions
 if $rows != 0 then
   return -1
 endi
 
-sql create database d1 vgroups 2;
+sql_error create database d1 vgroups 2;
 
 print =============== kill dnode2
 system sh/exec.sh -n dnode2 -s stop -x SIGINT
@@ -106,22 +110,31 @@ if $rows != 1 then
   return -1
 endi
 
-if $data[0][0] != 9 then
+if $data[0][0] != 8 then
   return -1
 endi
 
-if $data[0][2] != undoAction then
+if $data[0][2] != redoAction then
   return -1
 endi
 
 if $data[0][3] != d2 then
   return -1
 endi
-return
+
+sql show databases ;
+if $rows != 4 then
+  return -1
+endi
+print d2 ==> $data(d2)[19]
+if $data(d2)[19] != creating then
+  return -1
+endi
+
 sql_error create database d2 vgroups 2;
 
 print =============== kill transaction
-sql kill transaction 9;
+sql kill transaction 8;
 sleep 2000
 
 sql show transactions
@@ -131,7 +144,34 @@ endi
 
 print =============== start dnode2
 system sh/exec.sh -n dnode2 -s start
-sleep 3000
+
+$x = 0
+step3:
+	$x = $x + 1
+	sleep 1000
+	if $x == 10 then
+	  print ====> dnode not ready!
+		return -1
+	endi
+sql show dnodes
+print ===> $data00 $data01 $data02 $data03 $data04 $data05
+print ===> $data10 $data11 $data12 $data13 $data14 $data15
+if $rows != 2 then
+  return -1
+endi
+if $data(1)[4] != ready then
+  goto step3
+endi
+if $data(2)[4] != ready then
+  goto step3
+endi
+
+sql show transactions
+if $rows != 0 then
+  return -1
+endi
+
+sql drop database d2;
 
 sql show transactions
 if $rows != 0 then
@@ -145,6 +185,5 @@ sql_error kill transaction 3;
 sql_error kill transaction 4;
 sql_error kill transaction 5;
 
-return
 system sh/exec.sh -n dnode1 -s stop -x SIGINT
 system sh/exec.sh -n dnode2 -s stop -x SIGINT
\ No newline at end of file