fix:check wal not exist again before register to push & move push logic from...

fix:check wal not exist again before register to push & move push logic from write thread to query thread

include/common/tmsgdef.h

@@ -310,6 +310,7 @@ enum {
   TD_DEF_MSG_TYPE(TDMT_VND_TMQ_ADD_CHECKINFO, "vnode-tmq-add-checkinfo", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_TMQ_DEL_CHECKINFO, "vnode-del-checkinfo", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_TMQ_CONSUME, "vnode-tmq-consume", SMqPollReq, SMqDataBlkRsp)
+  TD_DEF_MSG_TYPE(TDMT_VND_TMQ_CONSUME_PUSH, "vnode-tmq-consume-push", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_TMQ_VG_WALINFO, "vnode-tmq-vg-walinfo", SMqPollReq, SMqDataBlkRsp)
   TD_DEF_MSG_TYPE(TDMT_VND_TMQ_MAX_MSG, "vnd-tmq-max", NULL, NULL)
 

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

@@ -712,6 +712,7 @@ SArray *vmGetMsgHandles() {
   if (dmSetMgmtHandle(pArray, TDMT_VND_TMQ_ADD_CHECKINFO, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_TMQ_DEL_CHECKINFO, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_TMQ_CONSUME, vmPutMsgToQueryQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_VND_TMQ_CONSUME_PUSH, vmPutMsgToQueryQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_TMQ_VG_WALINFO, vmPutMsgToFetchQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_DELETE, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_BATCH_DEL, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;

source/dnode/vnode/src/inc/vnodeInt.h

@@ -226,6 +226,7 @@ int32_t tqProcessDeleteSubReq(STQ* pTq, int64_t version, char* msg, int32_t msgL
 int32_t tqProcessOffsetCommitReq(STQ* pTq, int64_t version, char* msg, int32_t msgLen);
 int32_t tqProcessSeekReq(STQ* pTq, int64_t sversion, char* msg, int32_t msgLen);
 int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg);
+int32_t tqProcessPollPush(STQ* pTq, SRpcMsg* pMsg);
 int32_t tqProcessVgWalInfoReq(STQ* pTq, SRpcMsg* pMsg);
 
 // tq-stream

source/dnode/vnode/src/tq/tq.c

@@ -264,7 +264,7 @@ int32_t tqSendDataRsp(STqHandle* pHandle, const SRpcMsg* pMsg, const SMqPollReq*
   tFormatOffset(buf1, 80, &pRsp->reqOffset);
   tFormatOffset(buf2, 80, &pRsp->rspOffset);
 
-  tqDebug("vgId:%d consumer:0x%" PRIx64 " (epoch %d) send rsp, block num:%d, req:%s, rsp:%s, reqId:0x%" PRIx64, vgId,
+  tqDebug("tmq poll vgId:%d consumer:0x%" PRIx64 " (epoch %d) send rsp, block num:%d, req:%s, rsp:%s, reqId:0x%" PRIx64, vgId,
           pReq->consumerId, pReq->epoch, pRsp->blockNum, buf1, buf2, pReq->reqId);
 
   return 0;
@@ -421,6 +421,35 @@ int32_t tqCheckColModifiable(STQ* pTq, int64_t tbUid, int32_t colId) {
   return 0;
 }
 
+int32_t tqProcessPollPush(STQ* pTq, SRpcMsg* pMsg) {
+  int32_t vgId = TD_VID(pTq->pVnode);
+  taosWLockLatch(&pTq->lock);
+  if (taosHashGetSize(pTq->pPushMgr) > 0) {
+    void* pIter = taosHashIterate(pTq->pPushMgr, NULL);
+
+    while (pIter) {
+      STqHandle* pHandle = *(STqHandle**)pIter;
+      tqDebug("vgId:%d start set submit for pHandle:%p, consumer:0x%" PRIx64, vgId, pHandle, pHandle->consumerId);
+
+      if (ASSERT(pHandle->msg != NULL)) {
+        tqError("pHandle->msg should not be null");
+        break;
+      }else{
+        SRpcMsg msg = {.msgType = TDMT_VND_TMQ_CONSUME, .pCont = pHandle->msg->pCont, .contLen = pHandle->msg->contLen, .info = pHandle->msg->info};
+        tmsgPutToQueue(&pTq->pVnode->msgCb, QUERY_QUEUE, &msg);
+        taosMemoryFree(pHandle->msg);
+        pHandle->msg = NULL;
+      }
+
+      pIter = taosHashIterate(pTq->pPushMgr, pIter);
+    }
+
+    taosHashClear(pTq->pPushMgr);
+  }
+  taosWLockLatch(&pTq->lock);
+  return 0;
+}
+
 int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
   SMqPollReq req = {0};
   int        code = 0;

source/dnode/vnode/src/tq/tqPush.c

@@ -20,30 +20,15 @@ int32_t tqProcessSubmitReqForSubscribe(STQ* pTq) {
   int32_t vgId = TD_VID(pTq->pVnode);
 
   taosWLockLatch(&pTq->lock);
-
   if (taosHashGetSize(pTq->pPushMgr) > 0) {
-    void* pIter = taosHashIterate(pTq->pPushMgr, NULL);
-
-    while (pIter) {
-      STqHandle* pHandle = *(STqHandle**)pIter;
-      tqDebug("vgId:%d start set submit for pHandle:%p, consumer:0x%" PRIx64, vgId, pHandle, pHandle->consumerId);
-
-      if (ASSERT(pHandle->msg != NULL)) {
-        tqError("pHandle->msg should not be null");
-        break;
-      }else{
-        SRpcMsg msg = {.msgType = TDMT_VND_TMQ_CONSUME, .pCont = pHandle->msg->pCont, .contLen = pHandle->msg->contLen, .info = pHandle->msg->info};
-        tmsgPutToQueue(&pTq->pVnode->msgCb, QUERY_QUEUE, &msg);
-        taosMemoryFree(pHandle->msg);
-        pHandle->msg = NULL;
-      }
-
-      pIter = taosHashIterate(pTq->pPushMgr, pIter);
-    }
-
-    taosHashClear(pTq->pPushMgr);
+    SRpcMsg msg = {.msgType = TDMT_VND_TMQ_CONSUME_PUSH};
+    msg.pCont = rpcMallocCont(sizeof(SMsgHead));
+    msg.contLen = sizeof(SMsgHead);
+    SMsgHead *pHead = msg.pCont;
+    pHead->vgId = vgId;
+    pHead->contLen = msg.contLen;
+    tmsgPutToQueue(&pTq->pVnode->msgCb, QUERY_QUEUE, &msg);
   }
-
   // unlock
   taosWUnLockLatch(&pTq->lock);
   return 0;

source/dnode/vnode/src/tq/tqUtil.c

@@ -176,24 +176,20 @@ static int32_t extractDataAndRspForNormalSubscribe(STQ* pTq, STqHandle* pHandle,
   if (terrno == TSDB_CODE_WAL_LOG_NOT_EXIST && dataRsp.blockNum == 0) {
     // lock
     taosWLockLatch(&pTq->lock);
-    code = tqRegisterPushHandle(pTq, pHandle, pMsg);
-    taosWUnLockLatch(&pTq->lock);
-    tDeleteMqDataRsp(&dataRsp);
-    return code;
+    int64_t ver = walGetCommittedVer(pHandle->pWalReader->pWal);
+    if (pOffset->version >= ver || dataRsp.rspOffset.version >= ver){   //check if there are data again to avoid lost data
+      code = tqRegisterPushHandle(pTq, pHandle, pMsg);
+      taosWUnLockLatch(&pTq->lock);
+      goto end;
+    }else{
+      taosWUnLockLatch(&pTq->lock);
+    }
   }
 
-  // NOTE: this pHandle->consumerId may have been changed already.
   code = tqSendDataRsp(pHandle, pMsg, pRequest, (SMqDataRsp*)&dataRsp, TMQ_MSG_TYPE__POLL_RSP, vgId);
 
-end : {
-  char buf[80] = {0};
-  tFormatOffset(buf, 80, &dataRsp.rspOffset);
-  tqDebug("tmq poll: consumer:0x%" PRIx64 ", subkey %s, vgId:%d, rsp block:%d, rsp offset type:%s, reqId:0x%" PRIx64
-          " code:%d",
-          consumerId, pHandle->subKey, vgId, dataRsp.blockNum, buf, pRequest->reqId, code);
+end :
   tDeleteMqDataRsp(&dataRsp);
-}
-
   return code;
 }
 

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

@@ -506,7 +506,7 @@ int32_t vnodePreprocessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg) {
 
 int32_t vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg) {
   vTrace("message in vnode query queue is processing");
-  if ((pMsg->msgType == TDMT_SCH_QUERY || pMsg->msgType == TDMT_VND_TMQ_CONSUME) && !syncIsReadyForRead(pVnode->sync)) {
+  if ((pMsg->msgType == TDMT_SCH_QUERY || pMsg->msgType == TDMT_VND_TMQ_CONSUME || pMsg->msgType == TDMT_VND_TMQ_CONSUME_PUSH) && !syncIsReadyForRead(pVnode->sync)) {
     vnodeRedirectRpcMsg(pVnode, pMsg, terrno);
     return 0;
   }
@@ -527,6 +527,8 @@ int32_t vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg) {
       return qWorkerProcessCQueryMsg(&handle, pVnode->pQuery, pMsg, 0);
     case TDMT_VND_TMQ_CONSUME:
       return tqProcessPollReq(pVnode->pTq, pMsg);
+    case TDMT_VND_TMQ_CONSUME_PUSH:
+      return tqProcessPollPush(pVnode->pTq, pMsg);
     default:
       vError("unknown msg type:%d in query queue", pMsg->msgType);
       return TSDB_CODE_APP_ERROR;
@@ -560,8 +562,8 @@ int32_t vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo) {
       return vnodeGetTableCfg(pVnode, pMsg, true);
     case TDMT_VND_BATCH_META:
       return vnodeGetBatchMeta(pVnode, pMsg);
-    case TDMT_VND_TMQ_CONSUME:
-      return tqProcessPollReq(pVnode->pTq, pMsg);
+//    case TDMT_VND_TMQ_CONSUME:
+//      return tqProcessPollReq(pVnode->pTq, pMsg);
     case TDMT_VND_TMQ_VG_WALINFO:
       return tqProcessVgWalInfoReq(pVnode->pTq, pMsg);
     case TDMT_STREAM_TASK_RUN: