fix: avoid error 500 when node is re-joining cluster

Fixes https://emqx.atlassian.net/browse/EMQX-9899

apps/emqx/src/emqx_rpc.erl

@@ -147,7 +147,7 @@ unwrap_erpc({throw, A}) ->
     {error, A};
 unwrap_erpc({error, {exception, Err, _Stack}}) ->
     {error, Err};
-unwrap_erpc({error, {exit, Err}}) ->
+unwrap_erpc({exit, Err}) ->
     {error, Err};
 unwrap_erpc({error, {erpc, Err}}) ->
     {error, Err}.

apps/emqx_authz/src/emqx_authz_api_mnesia.erl

@@ -423,8 +423,8 @@ users(get, #{query_string := QueryString}) ->
     of
         {error, page_limit_invalid} ->
             {400, #{code => <<"INVALID_PARAMETER">>, message => <<"page_limit_invalid">>}};
-        {error, Node, {badrpc, R}} ->
+        {error, Node, Error} ->
-            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, R])),
+            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, Error])),
             {500, #{code => <<"NODE_DOWN">>, message => Message}};
         Result ->
             {200, Result}
@@ -459,8 +459,8 @@ clients(get, #{query_string := QueryString}) ->
     of
         {error, page_limit_invalid} ->
             {400, #{code => <<"INVALID_PARAMETER">>, message => <<"page_limit_invalid">>}};
-        {error, Node, {badrpc, R}} ->
+        {error, Node, Error} ->
-            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, R])),
+            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, Error])),
             {500, #{code => <<"NODE_DOWN">>, message => Message}};
         Result ->
             {200, Result}

apps/emqx_bridge/src/emqx_bridge_api.erl

@@ -756,7 +756,14 @@ format_bridge_info([FirstBridge | _] = Bridges) ->
     }).
 
 format_bridge_metrics(Bridges) ->
-    NodeMetrics = collect_metrics(Bridges),
+    FilteredBridges = lists:filter(
+        fun
+            ({_Node, Metric}) when is_map(Metric) -> true;
+            (_) -> false
+        end,
+        Bridges
+    ),
+    NodeMetrics = collect_metrics(FilteredBridges),
     #{
         metrics => aggregate_metrics(NodeMetrics),
         node_metrics => NodeMetrics

apps/emqx_ctl/src/emqx_ctl.app.src

@@ -1,6 +1,6 @@
 {application, emqx_ctl, [
     {description, "Backend for emqx_ctl script"},
-    {vsn, "0.1.0"},
+    {vsn, "0.1.1"},
     {registered, []},
     {mod, {emqx_ctl_app, []}},
     {applications, [

apps/emqx_ctl/src/emqx_ctl.erl

@@ -228,7 +228,7 @@ handle_call({register_command, Cmd, MF, Opts}, _From, State = #state{seq = Seq})
             ets:insert(?CMD_TAB, {{Seq, Cmd}, MF, Opts}),
             {reply, ok, next_seq(State)};
         [[OriginSeq] | _] ->
-            ?LOG_WARNING(#{msg => "CMD_overidden", cmd => Cmd, mf => MF}),
+            ?LOG_WARNING(#{msg => "CMD_overridden", cmd => Cmd, mf => MF}),
             true = ets:insert(?CMD_TAB, {{OriginSeq, Cmd}, MF, Opts}),
             {reply, ok, State}
     end;

apps/emqx_gateway/src/emqx_gateway.app.src

@@ -1,7 +1,7 @@
 %% -*- mode: erlang -*-
 {application, emqx_gateway, [
     {description, "The Gateway management application"},
-    {vsn, "0.1.16"},
+    {vsn, "0.1.17"},
     {registered, []},
     {mod, {emqx_gateway_app, []}},
     {applications, [kernel, stdlib, emqx, emqx_authn, emqx_ctl]},

apps/emqx_gateway/src/emqx_gateway_api_clients.erl

@@ -133,8 +133,10 @@ clients(get, #{
         case Result of
             {error, page_limit_invalid} ->
                 {400, #{code => <<"INVALID_PARAMETER">>, message => <<"page_limit_invalid">>}};
-            {error, Node, {badrpc, R}} ->
+            {error, Node, Error} ->
-                Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, R])),
+                Message = list_to_binary(
+                    io_lib:format("bad rpc call ~p, Reason ~p", [Node, Error])
+                ),
                 {500, #{code => <<"NODE_DOWN">>, message => Message}};
             Response ->
                 {200, Response}

apps/emqx_management/src/emqx_mgmt_api.erl

@@ -134,8 +134,8 @@ do_node_query(
     ResultAcc
 ) ->
     case do_query(Node, QueryState) of
-        {error, {badrpc, R}} ->
+        {error, Error} ->
-            {error, Node, {badrpc, R}};
+            {error, Node, Error};
         {Rows, NQueryState = #{complete := Complete}} ->
             case accumulate_query_rows(Node, Rows, NQueryState, ResultAcc) of
                 {enough, NResultAcc} ->
@@ -179,8 +179,8 @@ do_cluster_query(
     ResultAcc
 ) ->
     case do_query(Node, QueryState) of
-        {error, {badrpc, R}} ->
+        {error, Error} ->
-            {error, Node, {badrpc, R}};
+            {error, Node, Error};
         {Rows, NQueryState = #{complete := Complete}} ->
             case accumulate_query_rows(Node, Rows, NQueryState, ResultAcc) of
                 {enough, NResultAcc} ->
@@ -275,7 +275,7 @@ do_query(Node, QueryState) when Node =:= node() ->
     do_select(Node, QueryState);
 do_query(Node, QueryState) ->
     case
-        rpc:call(
+        catch rpc:call(
             Node,
             ?MODULE,
             do_query,
@@ -284,6 +284,7 @@ do_query(Node, QueryState) ->
         )
     of
         {badrpc, _} = R -> {error, R};
+        {'EXIT', _} = R -> {error, R};
         Ret -> Ret
     end.
 
@@ -298,6 +299,7 @@ do_select(
 ) ->
     QueryState = maybe_apply_total_query(Node, QueryState0),
     Result =
+        try
             case maps:get(continuation, QueryState, undefined) of
                 undefined ->
                     ets:select(Tab, Ms, Limit);
@@ -305,8 +307,16 @@ do_select(
                     %% XXX: Repair is necessary because we pass Continuation back
                     %% and forth through the nodes in the `do_cluster_query`
                     ets:select(ets:repair_continuation(Continuation, Ms))
+            end
+        catch
+            exit:_ = Exit ->
+                {error, Exit};
+            Type:Reason:Stack ->
+                {error, #{exception => Type, reason => Reason, stacktrace => Stack}}
         end,
     case Result of
+        {error, _} ->
+            {[], mark_complete(QueryState)};
         {Rows, '$end_of_table'} ->
             NRows = maybe_apply_fuzzy_filter(Rows, QueryState),
             {NRows, mark_complete(QueryState)};
@@ -354,7 +364,11 @@ counting_total_fun(_QueryState = #{match_spec := Ms, fuzzy_fun := undefined}) ->
     [{MatchHead, Conditions, _Return}] = Ms,
     CountingMs = [{MatchHead, Conditions, [true]}],
     fun(Tab) ->
+        try
             ets:select_count(Tab, CountingMs)
+        catch
+            _Type:_Reason -> 0
+        end
     end;
 counting_total_fun(_QueryState = #{fuzzy_fun := FuzzyFun}) when FuzzyFun =/= undefined ->
     %% XXX: Calculating the total number for a fuzzy searching is very very expensive

apps/emqx_management/src/emqx_mgmt_api_alarms.erl

@@ -123,8 +123,8 @@ alarms(get, #{query_string := QString}) ->
     of
         {error, page_limit_invalid} ->
             {400, #{code => <<"INVALID_PARAMETER">>, message => <<"page_limit_invalid">>}};
-        {error, Node, {badrpc, R}} ->
+        {error, Node, Error} ->
-            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, R])),
+            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, Error])),
             {500, #{code => <<"NODE_DOWN">>, message => Message}};
         Response ->
             {200, Response}

apps/emqx_management/src/emqx_mgmt_api_topics.erl

@@ -120,8 +120,8 @@ do_list(Params) ->
     of
         {error, page_limit_invalid} ->
             {400, #{code => <<"INVALID_PARAMETER">>, message => <<"page_limit_invalid">>}};
-        {error, Node, {badrpc, R}} ->
+        {error, Node, Error} ->
-            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, R])),
+            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, Error])),
             {500, #{code => <<"NODE_DOWN">>, message => Message}};
         Response ->
             {200, Response}

apps/emqx_rule_engine/src/emqx_rule_engine_api.erl

@@ -339,6 +339,9 @@ param_path_id() ->
     of
         {error, page_limit_invalid} ->
             {400, #{code => 'BAD_REQUEST', message => <<"page_limit_invalid">>}};
+        {error, Node, Error} ->
+            Message = list_to_binary(io_lib:format("bad rpc call ~p, Reason ~p", [Node, Error])),
+            {500, #{code => <<"NODE_DOWN">>, message => Message}};
         Result ->
             {200, Result}
     end;

changes/ce/fix-10760.en.md

@@ -0,0 +1 @@
+Fix Internal Error 500 that occurred sometimes when bridge statistics page was updated while a node was (re)joining the cluster.