add log for user to check; show "System not ready" for not connection

issue;

src/common/inc/tglobal.h

@@ -53,6 +53,7 @@ extern char     tsCharset[];            // default encode string
 extern int32_t  tsEnableCoreFile;
 extern int32_t  tsCompressMsgSize;
 extern char     tsTempDir[];
+extern int32_t  tsFirstBoot;
 
 //query buffer management
 extern int32_t  tsQueryBufferSize;      // maximum allowed usage buffer for each data node during query processing

src/common/src/tglobal.c

@@ -60,6 +60,7 @@ char    tsCharset[TSDB_LOCALE_LEN] = {0};  // default encode string
 int32_t tsEnableCoreFile = 0;
 int32_t tsMaxBinaryDisplayWidth = 30;
 char    tsTempDir[TSDB_FILENAME_LEN] = "/tmp/";
+int32_t tsFirstBoot = 0;
 
 /*
  * denote if the server needs to compress response message at the application layer to client, including query rsp,
@@ -1368,11 +1369,16 @@ int32_t taosCheckGlobalCfg() {
 
   if (tsFirst[0] == 0) {
     strcpy(tsFirst, tsLocalEp);
+    taosGetFqdnPortFromEp(tsFirst, fqdn, &port);
   } else {
     taosGetFqdnPortFromEp(tsFirst, fqdn, &port);
     snprintf(tsFirst, sizeof(tsFirst), "%s:%u", fqdn, port);
   }
 
+  if(tscEmbedded && tsFirstBoot && strcmp(fqdn, tsLocalFqdn)){
+    userLog("Dnode %s is goning to join the cluster %s belongs, make sure this is what you want", tsLocalFqdn, fqdn);
+  }
+
   if (tsSecond[0] == 0) {
     strcpy(tsSecond, tsLocalEp);
   } else {

src/dnode/src/dnodeCfg.c

@@ -91,6 +91,7 @@ static int32_t dnodeReadCfg() {
   fp = fopen(file, "r");
   if (!fp) {
     dDebug("failed to read %s, file not exist", file);
+    tsFirstBoot = 1;
     goto PARSE_CFG_OVER;
   }
 

src/dnode/src/dnodeMgmt.c

@@ -58,6 +58,7 @@ static uint32_t   tsRebootTime;
 static taos_qset  tsMgmtQset = NULL;
 static taos_queue tsMgmtQueue = NULL;
 static pthread_t  tsQthread;
+static uint32_t   tsNumOfLostConn = 0;
 
 static void   dnodeProcessStatusRsp(SRpcMsg *pMsg);
 static void   dnodeSendStatusMsg(void *handle, void *tmrId);
@@ -467,10 +468,31 @@ static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) {
 static void dnodeProcessStatusRsp(SRpcMsg *pMsg) {
   if (pMsg->code != TSDB_CODE_SUCCESS) {
     dError("status rsp is received, error:%s", tstrerror(pMsg->code));
+    if(pMsg->code == TSDB_CODE_RPC_NETWORK_UNAVAIL) {
+      tsNumOfLostConn++;
+      if(tsNumOfLostConn && (tsNumOfLostConn % 10 == 0)) {
+        if(tsFirstBoot) {
+          userLog("Dnode has not joined the cluster yet, check the network/configuration and the other dnode status in cluster");
+        } else {
+          userLog("Dnode lost connection to the mnode, check the network and the other dnode status in cluster");
+        }
+      }
+    }
     taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer);
     return;
   }
 
+  if(tsNumOfLostConn) {
+    if(tsNumOfLostConn >= 10) {
+      if(tsFirstBoot) {
+        userLog("Done has joined the cluster now");
+      } else {
+        userLog("Dnode's connection to the mnode is ok now");
+      }
+    }
+    tsNumOfLostConn = 0;
+  }
+
   SStatusRsp *pStatusRsp = pMsg->pCont;
   SMnodeInfos *minfos = &pStatusRsp->mnodes;
   dnodeUpdateMInfos(minfos);

src/dnode/src/dnodeShell.c

@@ -160,6 +160,10 @@ static int dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char
 
   if (rpcRsp.code != 0) {
     dError("user:%s, auth msg received from mnodes, error:%s", user, tstrerror(rpcRsp.code));
+    if(rpcRsp.code == TSDB_CODE_RPC_NETWORK_UNAVAIL) {
+      dError("all mnodes unreachable, dnode not ready");
+      rpcRsp.code = TSDB_CODE_RPC_NOT_READY;
+    }
   } else {
     SAuthRsp *pRsp = rpcRsp.pCont;
     dDebug("user:%s, auth msg received from mnodes", user);

src/rpc/src/rpcMain.c

@@ -1372,7 +1372,9 @@ static void rpcProcessRetryTimer(void *param, void *tmrId) {
       // close the connection
       tDebug("%s, failed to send msg:%s to %s:%hu", pConn->info, taosMsg[pConn->outType], pConn->peerFqdn, pConn->peerPort);
       if (pConn->pContext) {
-        pConn->pContext->code = TSDB_CODE_RPC_NETWORK_UNAVAIL;
+        if(pConn->pContext->code == 0) {
+          pConn->pContext->code = TSDB_CODE_RPC_NETWORK_UNAVAIL;
+        }	      
         pConn->pContext->pConn = NULL;
         pConn->pReqMsg = NULL;
         taosTmrStart(rpcProcessConnError, 0, pConn->pContext, pRpc->tmrCtrl);

src/util/inc/tlog.h

@@ -49,6 +49,8 @@ void    taosPrintLongString(const char * flags, int32_t dflag, const char *forma
 
 void    taosDumpData(unsigned char *msg, int32_t len);
 
+#define userLog(...) { taosPrintLog("USR ", 255, __VA_ARGS__); }
+
 #ifdef __cplusplus
 }
 #endif