Merge pull request #22440 from taosdata/fix/retry_resolve_fqdn

merge fix/retry_resolve_fqdn to main

include/common/tglobal.h

@@ -199,6 +199,7 @@ extern bool    tsFilterScalarMode;
 extern int32_t tsKeepTimeOffset;
 extern int32_t tsMaxStreamBackendCache;
 extern int32_t tsPQSortMemThreshold;
+extern int32_t tsResolveFQDNRetryTime;
 
 // #define NEEDTO_COMPRESSS_MSG(size) (tsCompressMsgSize != -1 && (size) > tsCompressMsgSize)
 

source/common/src/tglobal.c

@@ -236,6 +236,7 @@ int64_t tsStreamBufferSize = 128 * 1024 * 1024;
 int64_t tsCheckpointInterval = 3 * 60 * 60 * 1000;
 bool    tsFilterScalarMode = false;
 int32_t tsKeepTimeOffset = 0;  // latency of data migration
+int     tsResolveFQDNRetryTime = 100; //seconds
 
 #ifndef _STORAGE
 int32_t taosSetTfsCfg(SConfig *pCfg) {
@@ -580,6 +581,7 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddInt32(pCfg, "keepTimeOffset", tsKeepTimeOffset, 0, 23, CFG_SCOPE_SERVER) != 0) return -1;
   if (cfgAddInt32(pCfg, "maxStreamBackendCache", tsMaxStreamBackendCache, 16, 1024, CFG_SCOPE_SERVER) != 0) return -1;
   if (cfgAddInt32(pCfg, "pqSortMemThreshold", tsPQSortMemThreshold, 1, 10240, CFG_SCOPE_SERVER) != 0) return -1;
+  if (cfgAddInt32(pCfg, "resolveFQDNRetryTime", tsResolveFQDNRetryTime, 1, 10240, 0) != 0) return -1;
 
   GRANT_CFG_ADD;
   return 0;
@@ -979,6 +981,7 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsKeepTimeOffset = cfgGetItem(pCfg, "keepTimeOffset")->i32;
   tsMaxStreamBackendCache = cfgGetItem(pCfg, "maxStreamBackendCache")->i32;
   tsPQSortMemThreshold = cfgGetItem(pCfg, "pqSortMemThreshold")->i32;
+  tsResolveFQDNRetryTime = cfgGetItem(pCfg, "resolveFQDNRetryTime")->i32;
 
   GRANT_CFG_GET;
   return 0;

source/libs/sync/src/syncUtil.c

@@ -21,6 +21,7 @@
 #include "syncRaftCfg.h"
 #include "syncRaftStore.h"
 #include "syncSnapshot.h"
+#include "tglobal.h"
 
 void syncCfg2SimpleStr(const SSyncCfg* pCfg, char* buf, int32_t bufLen) {
   int32_t len = snprintf(buf, bufLen, "{num:%d, as:%d, [", pCfg->replicaNum, pCfg->myIndex);
@@ -41,7 +42,22 @@ void syncUtilNodeInfo2EpSet(const SNodeInfo* pInfo, SEpSet* pEpSet) {
 }
 
 bool syncUtilNodeInfo2RaftId(const SNodeInfo* pInfo, SyncGroupId vgId, SRaftId* raftId) {
-  uint32_t ipv4 = taosGetIpv4FromFqdn(pInfo->nodeFqdn);
+  uint32_t ipv4 = 0xFFFFFFFF;
+  sDebug("vgId:%d, start to resolve sync addr fqdn in %d seconds, "
+        "dnode:%d cluster:%" PRId64 " fqdn:%s port:%u ", 
+        vgId, tsResolveFQDNRetryTime,
+        pInfo->nodeId, pInfo->clusterId, pInfo->nodeFqdn, pInfo->nodePort);
+  for(int i = 0; i < tsResolveFQDNRetryTime; i++){
+    ipv4 = taosGetIpv4FromFqdn(pInfo->nodeFqdn);
+    if (ipv4 == 0xFFFFFFFF || ipv4 == 1) {
+      sError("failed to resolve ipv4 addr, fqdn:%s, wait one second", pInfo->nodeFqdn);
+      taosSsleep(1);
+    }
+    else{
+      break;
+    }
+  }
+  
   if (ipv4 == 0xFFFFFFFF || ipv4 == 1) {
     sError("failed to resolve ipv4 addr, fqdn:%s", pInfo->nodeFqdn);
     terrno = TSDB_CODE_TSC_INVALID_FQDN;