fix: add pre-finished stage, trans won't be dropped in follower if it has undo actions while rollback

source/dnode/mnode/impl/inc/mndDef.h

@@ -108,7 +108,8 @@ typedef enum {
   TRN_STAGE_UNDO_ACTION = 3,
   TRN_STAGE_COMMIT = 4,
   TRN_STAGE_COMMIT_ACTION = 5,
-  TRN_STAGE_FINISHED = 6
+  TRN_STAGE_FINISHED = 6,
+  TRN_STAGE_PRE_FINISH = 7
 } ETrnStage;
 
 typedef enum {

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

@@ -460,6 +460,8 @@ static const char *mndTransStr(ETrnStage stage) {
       return "commitAction";
     case TRN_STAGE_FINISHED:
       return "finished";
+    case TRN_STAGE_PRE_FINISH:
+      return "pre-finish";
     default:
       return "invalid";
   }
@@ -602,6 +604,11 @@ static int32_t mndTransActionUpdate(SSdb *pSdb, STrans *pOld, STrans *pNew) {
     mTrace("trans:%d, stage from rollback to undoAction since perform update action", pNew->id);
   }
 
+  if (pOld->stage == TRN_STAGE_PRE_FINISH) {
+    pOld->stage = TRN_STAGE_FINISHED;
+    mTrace("trans:%d, stage from pre-finish to finished since perform update action", pNew->id);
+  }
+
   return 0;
 }
 
@@ -931,6 +938,16 @@ static int32_t mndTransRollback(SMnode *pMnode, STrans *pTrans) {
   return 0;
 }
 
+static int32_t mndTransPreFinish(SMnode *pMnode, STrans *pTrans) {
+  mInfo("trans:%d, pre-finish transaction", pTrans->id);
+  if (mndTransSync(pMnode, pTrans) != 0) {
+    mError("trans:%d, failed to pre-finish since %s", pTrans->id, terrstr());
+    return -1;
+  }
+  mInfo("trans:%d, pre-finish finished", pTrans->id);
+  return 0;
+}
+
 static void mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans) {
   bool    sendRsp = false;
   int32_t code = pTrans->code;
@@ -1437,7 +1454,7 @@ static bool mndTransPerformCommitActionStage(SMnode *pMnode, STrans *pTrans) {
 
   if (code == 0) {
     pTrans->code = 0;
-    pTrans->stage = TRN_STAGE_FINISHED;
+    pTrans->stage = TRN_STAGE_FINISHED; // TRN_STAGE_PRE_FINISH is not necessary
     mInfo("trans:%d, stage from commitAction to finished", pTrans->id);
     continueExec = true;
   } else {
@@ -1455,8 +1472,8 @@ static bool mndTransPerformUndoActionStage(SMnode *pMnode, STrans *pTrans) {
   int32_t code = mndTransExecuteUndoActions(pMnode, pTrans);
 
   if (code == 0) {
-    pTrans->stage = TRN_STAGE_FINISHED;
+    pTrans->stage = TRN_STAGE_PRE_FINISH;
-    mInfo("trans:%d, stage from undoAction to finished", pTrans->id);
+    mInfo("trans:%d, stage from undoAction to pre-finish", pTrans->id);
     continueExec = true;
   } else if (code == TSDB_CODE_ACTION_IN_PROGRESS) {
     mInfo("trans:%d, stage keep on undoAction since %s", pTrans->id, tstrerror(code));
@@ -1489,6 +1506,25 @@ static bool mndTransPerformRollbackStage(SMnode *pMnode, STrans *pTrans) {
   return continueExec;
 }
 
+static bool mndTransPerfromPreFinishedStage(SMnode *pMnode, STrans *pTrans) {
+  if (mndCannotExecuteTransAction(pMnode)) return false;
+
+  bool    continueExec = true;
+  int32_t code = mndTransPreFinish(pMnode, pTrans);
+
+  if (code == 0) {
+    pTrans->stage = TRN_STAGE_FINISHED;
+    mInfo("trans:%d, stage from pre-finish to finish", pTrans->id);
+    continueExec = true;
+  } else {
+    pTrans->failedTimes++;
+    mError("trans:%d, stage keep on pre-finish since %s, failedTimes:%d", pTrans->id, terrstr(), pTrans->failedTimes);
+    continueExec = false;
+  }
+
+  return continueExec;
+}
+
 static bool mndTransPerfromFinishedStage(SMnode *pMnode, STrans *pTrans) {
   bool continueExec = false;
 
@@ -1545,6 +1581,14 @@ void mndTransExecute(SMnode *pMnode, STrans *pTrans, bool isLeader) {
       case TRN_STAGE_UNDO_ACTION:
         continueExec = mndTransPerformUndoActionStage(pMnode, pTrans);
         break;
+      case TRN_STAGE_PRE_FINISH:
+        if (isLeader) {
+          continueExec = mndTransPerfromPreFinishedStage(pMnode, pTrans);
+        } else {
+          mInfo("trans:%d, can not pre-finish since not leader", pTrans->id);
+          continueExec = false;
+        }
+        break;
       case TRN_STAGE_FINISHED:
         continueExec = mndTransPerfromFinishedStage(pMnode, pTrans);
         break;

tests/script/tsim/trans/create_stb.sim

@@ -0,0 +1,72 @@
+system sh/stop_dnodes.sh
+system sh/deploy.sh -n dnode1 -i 1
+system sh/deploy.sh -n dnode2 -i 2
+system sh/deploy.sh -n dnode3 -i 3
+system sh/deploy.sh -n dnode4 -i 4
+system sh/cfg.sh -n dnode1 -c transPullupInterval -v 1
+system sh/cfg.sh -n dnode2 -c transPullupInterval -v 1
+system sh/cfg.sh -n dnode3 -c transPullupInterval -v 1
+system sh/cfg.sh -n dnode4 -c transPullupInterval -v 1
+system sh/exec.sh -n dnode1 -s start
+system sh/exec.sh -n dnode2 -s start
+system sh/exec.sh -n dnode3 -s start
+system sh/exec.sh -n dnode4 -s start
+sql connect
+
+print =============== step1: create dnodes
+sql create dnode $hostname port 7200
+sql create dnode $hostname port 7300
+sql create dnode $hostname port 7400
+
+$x = 0
+step1: 
+	$x = $x + 1
+	sleep 1000
+	if $x == 10 then
+		return -1
+	endi
+sql select * from information_schema.ins_dnodes -x step1
+if $data(1)[4] != ready then
+  goto step1
+endi
+if $data(2)[4] != ready then
+  goto step1
+endi
+if $data(3)[4] != ready then
+  goto step1
+endi
+
+print =============== step2: create mnode 2 and 3
+sql create mnode on dnode 2
+sql create mnode on dnode 3
+sql create database db vgroups 2
+
+print =============== step3: kill dnode4
+system sh/exec.sh -n dnode4 -s stop -x SIGKILL
+sql use db
+sql_error create table stb (ts timestamp, i int) tags (j int)
+
+print =============== step4: create database
+sql show transactions
+if $rows != 1 then
+  return -1
+endi
+
+sleep 3000
+system sh/exec.sh -n dnode4 -s start
+
+$wt = 0
+step4:
+$wt = $wt + 1
+sleep 1000
+if $wt == 200 then
+  print ====> transaction already running
+  return -1
+endi
+
+sql show transactions
+if $rows != 0 then
+  print wait 1 seconds to alter
+  goto step4
+endi
+