fix(sync) sync/mnode integration dead lock

source/libs/sync/inc/syncInt.h

@@ -149,7 +149,7 @@ typedef struct SSyncNode {
 
   // restore state
   bool       restoreFinish;
-  sem_t      restoreSem;
+  //sem_t      restoreSem;
   SSnapshot* pSnapshot;
 
 } SSyncNode;

source/libs/sync/src/syncAppendEntries.c

@@ -366,9 +366,12 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
                     ths->pFsm->FpRestoreFinish(ths->pFsm);
                   }
                   ths->restoreFinish = true;
+                  sInfo("==syncNodeOnAppendEntriesCb== restoreFinish set true %p vgId:%d", ths, ths->vgId);
 
+                  /*
                   tsem_post(&ths->restoreSem);
-                  sInfo("==syncNodeOnAppendEntriesCb== RestoreFinish tsem_post");
+                  sInfo("==syncNodeOnAppendEntriesCb== RestoreFinish tsem_post %p", ths);
+                  */
                 }
               }
 

source/libs/sync/src/syncCommit.c

@@ -143,9 +143,12 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
                 pSyncNode->pFsm->FpRestoreFinish(pSyncNode->pFsm);
               }
               pSyncNode->restoreFinish = true;
+              sInfo("==syncMaybeAdvanceCommitIndex== restoreFinish set true %p vgId:%d", pSyncNode, pSyncNode->vgId);
 
+              /*
               tsem_post(&pSyncNode->restoreSem);
-              sInfo("==syncMaybeAdvanceCommitIndex== RestoreFinish tsem_post");
+              sInfo("==syncMaybeAdvanceCommitIndex== RestoreFinish tsem_post %p", pSyncNode);
+              */
             }
           }
 

source/libs/sync/src/syncMain.c

@@ -499,7 +499,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
     pSyncNode->pSnapshot = taosMemoryMalloc(sizeof(SSnapshot));
     pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, pSyncNode->pSnapshot);
   }
-  tsem_init(&(pSyncNode->restoreSem), 0, 0);
+  //tsem_init(&(pSyncNode->restoreSem), 0, 0);
 
   // start in syncNodeStart
   // start raft
@@ -521,7 +521,19 @@ void syncNodeStart(SSyncNode* pSyncNode) {
     syncNodeAppendNoop(pSyncNode);
     syncMaybeAdvanceCommitIndex(pSyncNode);  // maybe only one replica
 
+    /*
+    sInfo("==syncNodeStart== RestoreFinish begin 1 replica tsem_wait %p", pSyncNode);
     tsem_wait(&pSyncNode->restoreSem);
+    sInfo("==syncNodeStart== RestoreFinish end 1 replica tsem_wait %p", pSyncNode);
+    */
+
+    /*
+    while (pSyncNode->restoreFinish != true) {
+      taosMsleep(10);
+    }
+    */
+
+    sInfo("==syncNodeStart== restoreFinish ok 1 replica %p vgId:%d", pSyncNode, pSyncNode->vgId);
     return;
   }
 
@@ -532,7 +544,18 @@ void syncNodeStart(SSyncNode* pSyncNode) {
   // ret = syncNodeStartPingTimer(pSyncNode);
   assert(ret == 0);
 
+  /*
+  sInfo("==syncNodeStart== RestoreFinish begin multi replica tsem_wait %p", pSyncNode);
   tsem_wait(&pSyncNode->restoreSem);
+  sInfo("==syncNodeStart== RestoreFinish end multi replica tsem_wait %p", pSyncNode);
+  */
+
+  /*
+  while (pSyncNode->restoreFinish != true) {
+    taosMsleep(10);
+  }
+  */
+  sInfo("==syncNodeStart== restoreFinish ok multi replica %p vgId:%d", pSyncNode, pSyncNode->vgId);
 }
 
 void syncNodeStartStandBy(SSyncNode* pSyncNode) {
@@ -573,7 +596,7 @@ void syncNodeClose(SSyncNode* pSyncNode) {
     taosMemoryFree(pSyncNode->pSnapshot);
   }
 
-  tsem_destroy(&pSyncNode->restoreSem);
+  //tsem_destroy(&pSyncNode->restoreSem);
 
   // free memory in syncFreeNode
   // taosMemoryFree(pSyncNode);

source/libs/transport/inc/transComm.h

@@ -94,7 +94,7 @@ typedef void* queue[2];
 /* Return the structure holding the given element. */
 #define QUEUE_DATA(e, type, field) ((type*)((void*)((char*)(e)-offsetof(type, field))))
 
-#define TRANS_RETRY_COUNT_LIMIT 20  // retry count limit
+#define TRANS_RETRY_COUNT_LIMIT 100  // retry count limit
 #define TRANS_RETRY_INTERVAL    15  // ms retry interval
 #define TRANS_CONN_TIMEOUT      3   // connect timeout