not retry if init_sqc message may be sent out successfully

src/share/ob_errno.cpp

@@ -6357,6 +6357,18 @@ static const _error _error_OB_DELETE_SERVER_NOT_ALLOWED = {
       .oracle_str_error      = "ORA-00600: internal error code, arguments: -4745, delete server not allowed",
       .oracle_str_user_error = "ORA-00600: internal error code, arguments: -4745, %s"
 };
+static const _error _error_OB_PACKET_STATUS_UNKNOWN = {
+      .error_name            = "OB_PACKET_STATUS_UNKNOWN",
+      .error_cause           = "Internal Error",
+      .error_solution        = "Contact OceanBase Support",
+      .mysql_errno           = -1,
+      .sqlstate              = "HY000",
+      .str_error             = "Network error and packet status unknown. Abort auto retry.",
+      .str_user_error        = "Network error and packet status unknown. Abort auto retry.",
+      .oracle_errno          = 600,
+      .oracle_str_error      = "ORA-00600: internal error code, arguments: -4746, Network error and packet status unknown. Abort auto retry.",
+      .oracle_str_user_error = "ORA-00600: internal error code, arguments: -4746, Network error and packet status unknown. Abort auto retry."
+};
 static const _error _error_OB_ERR_PARSER_INIT = {
       .error_name            = "OB_ERR_PARSER_INIT",
       .error_cause           = "Internal Error",
@@ -22480,6 +22492,7 @@ struct ObStrErrorInit
     _errors[-OB_FREEZE_SERVICE_EPOCH_MISMATCH] = &_error_OB_FREEZE_SERVICE_EPOCH_MISMATCH;
     _errors[-OB_FROZEN_INFO_ALREADY_EXIST] = &_error_OB_FROZEN_INFO_ALREADY_EXIST;
     _errors[-OB_DELETE_SERVER_NOT_ALLOWED] = &_error_OB_DELETE_SERVER_NOT_ALLOWED;
+    _errors[-OB_PACKET_STATUS_UNKNOWN] = &_error_OB_PACKET_STATUS_UNKNOWN;
     _errors[-OB_ERR_PARSER_INIT] = &_error_OB_ERR_PARSER_INIT;
     _errors[-OB_ERR_PARSE_SQL] = &_error_OB_ERR_PARSE_SQL;
     _errors[-OB_ERR_RESOLVE_SQL] = &_error_OB_ERR_RESOLVE_SQL;

src/share/ob_errno.def

@@ -623,6 +623,7 @@ DEFINE_ERROR(OB_WAIT_TABLET_READY_TIMEOUT, -4742, -1, "HY000", "log stream wait
 DEFINE_ERROR(OB_FREEZE_SERVICE_EPOCH_MISMATCH, -4743, -1, "HY000", "freeze service epoch is not expected");
 DEFINE_ERROR_EXT_DEP(OB_FROZEN_INFO_ALREADY_EXIST, -4744, -1, "HY000", "already exist larger frozen_scn in __all_freeze_info", "%s");
 DEFINE_ERROR_EXT(OB_DELETE_SERVER_NOT_ALLOWED, -4745, -1, "HY000", "delete server not allowed", "%s");
+DEFINE_ERROR(OB_PACKET_STATUS_UNKNOWN, -4746, -1, "HY000", "Network error and packet status unknown. Abort auto retry.");
 
 ////////////////////////////////////////////////////////////////
 // SQL & Schema specific error code, -5000 ~ -6000

src/share/ob_errno.h

@@ -433,6 +433,7 @@ constexpr int OB_WAIT_REPLAY_TIMEOUT = -4741;
 constexpr int OB_WAIT_TABLET_READY_TIMEOUT = -4742;
 constexpr int OB_FREEZE_SERVICE_EPOCH_MISMATCH = -4743;
 constexpr int OB_DELETE_SERVER_NOT_ALLOWED = -4745;
+constexpr int OB_PACKET_STATUS_UNKNOWN = -4746;
 constexpr int OB_ERR_PARSER_INIT = -5000;
 constexpr int OB_ERR_PARSE_SQL = -5001;
 constexpr int OB_ERR_RESOLVE_SQL = -5002;
@@ -2126,6 +2127,7 @@ constexpr int OB_ERR_INVALID_DATE_MSG_FMT_V2 = -4219;
 #define OB_FREEZE_SERVICE_EPOCH_MISMATCH__USER_ERROR_MSG "freeze service epoch is not expected"
 #define OB_FROZEN_INFO_ALREADY_EXIST__USER_ERROR_MSG "%s"
 #define OB_DELETE_SERVER_NOT_ALLOWED__USER_ERROR_MSG "%s"
+#define OB_PACKET_STATUS_UNKNOWN__USER_ERROR_MSG "Network error and packet status unknown. Abort auto retry."
 #define OB_ERR_PARSER_INIT__USER_ERROR_MSG "Failed to init SQL parser"
 #define OB_ERR_PARSE_SQL__USER_ERROR_MSG "%s near \'%.*s\' at line %d"
 #define OB_ERR_RESOLVE_SQL__USER_ERROR_MSG "Resolve error"
@@ -3956,6 +3958,7 @@ constexpr int OB_ERR_INVALID_DATE_MSG_FMT_V2 = -4219;
 #define OB_FREEZE_SERVICE_EPOCH_MISMATCH__ORA_USER_ERROR_MSG "ORA-00600: internal error code, arguments: -4743, freeze service epoch is not expected"
 #define OB_FROZEN_INFO_ALREADY_EXIST__ORA_USER_ERROR_MSG "ORA-00600: internal error code, arguments: -4744, %s"
 #define OB_DELETE_SERVER_NOT_ALLOWED__ORA_USER_ERROR_MSG "ORA-00600: internal error code, arguments: -4745, %s"
+#define OB_PACKET_STATUS_UNKNOWN__ORA_USER_ERROR_MSG "ORA-00600: internal error code, arguments: -4746, Network error and packet status unknown. Abort auto retry."
 #define OB_ERR_PARSER_INIT__ORA_USER_ERROR_MSG "ORA-00600: internal error code, arguments: -5000, Failed to init SQL parser"
 #define OB_ERR_PARSE_SQL__ORA_USER_ERROR_MSG "ORA-00900: %s near \'%.*s\' at line %d"
 #define OB_ERR_RESOLVE_SQL__ORA_USER_ERROR_MSG "ORA-00600: internal error code, arguments: -5002, Resolve error"

src/sql/engine/px/ob_px_rpc_processor.cpp

@@ -476,7 +476,10 @@ int ObFastInitSqcCB::deal_with_rpc_timeout_err_safely()
 
 {
   int ret = OB_SUCCESS;
-  ObDealWithRpcTimeoutCall call(addr_, retry_info_, timeout_ts_, trace_id_);
+  // only if it's sure init_sqc msg is not sent to sqc successfully, we can retry the query.
+  bool init_sqc_not_send_out = (get_error() == EASY_TIMEOUT_NOT_SENT_OUT
+      || get_error() == EASY_DISCONNECT_NOT_SENT_OUT);
+  ObDealWithRpcTimeoutCall call(addr_, retry_info_, timeout_ts_, trace_id_, init_sqc_not_send_out);
   call.ret_ = OB_TIMEOUT;
   ObGlobalInterruptManager *manager = ObGlobalInterruptManager::getInstance();
   if (OB_NOT_NULL(manager)) {
@@ -529,7 +532,12 @@ void ObDealWithRpcTimeoutCall::deal_with_rpc_timeout_err()
           LOG_WARN("fail to add invalid server distinctly", K_(trace_id), K(a_ret), K_(addr));
         }
       }
-      ret_ = OB_RPC_CONNECT_ERROR;
+      if (can_retry_) {
+        // return OB_RPC_CONNECT_ERROR to retry.
+        ret_ = OB_RPC_CONNECT_ERROR;
+      } else {
+        ret_ = OB_PACKET_STATUS_UNKNOWN;
+      }
     } else {
       LOG_DEBUG("rpc return OB_TIMEOUT, and it is actually timeout, "
                 "do not change error code", K(ret_),

src/sql/engine/px/ob_px_rpc_processor.h

@@ -120,8 +120,9 @@ public:
   ObDealWithRpcTimeoutCall(common::ObAddr addr,
       ObQueryRetryInfo *retry_info,
       int64_t timeout_ts,
-      common::ObCurTraceId::TraceId &trace_id) : addr_(addr), retry_info_(retry_info),
+      common::ObCurTraceId::TraceId &trace_id,
-      timeout_ts_(timeout_ts), trace_id_(trace_id), ret_(common::OB_TIMEOUT) {}
+      bool retry) : addr_(addr), retry_info_(retry_info),
+      timeout_ts_(timeout_ts), trace_id_(trace_id), ret_(common::OB_TIMEOUT), can_retry_(retry) {}
   ~ObDealWithRpcTimeoutCall() = default;
   void operator() (hash::HashMapPair<ObInterruptibleTaskID,
       ObInterruptCheckerNode *> &entry);
@@ -132,6 +133,7 @@ public:
   int64_t timeout_ts_;
   common::ObCurTraceId::TraceId trace_id_;
   int ret_;
+  bool can_retry_;
 };
 
 class ObFastInitSqcCB

src/sql/engine/px/ob_px_sqc_async_proxy.cpp

@@ -185,7 +185,14 @@ int ObPxSqcAsyncProxy::wait_all() {
         if (phy_plan_ctx_->get_timeout_timestamp() -
           ObTimeUtility::current_time() > 0) {
           error_index_ = idx;
-          ret = OB_RPC_CONNECT_ERROR;
+          bool init_sqc_not_send_out = (callback.get_error() == EASY_TIMEOUT_NOT_SENT_OUT
+              || callback.get_error() == EASY_DISCONNECT_NOT_SENT_OUT);
+          if (init_sqc_not_send_out) {
+            // only if it's sure init_sqc msg is not sent to sqc successfully, return OB_RPC_CONNECT_ERROR to retry.
+            ret = OB_RPC_CONNECT_ERROR;
+          } else {
+            ret = OB_PACKET_STATUS_UNKNOWN;
+          }
         } else {
           ret = OB_TIMEOUT;
         }