Merge pull request #9119 from emqx/1009-fix-sticky-dispatch-should-check-subscription-table

fix(shared): check sticky if sticky pid is still a member

CHANGES-4.3.md

@@ -58,6 +58,12 @@ File format:
   - For wildcard deliveries, the re-dispatch used the wrong topic (the publishing topic,
     but not the subscribing topic), caused messages to be lost when dispatching.
 
+- Fix shared subscription group member unsubscribe issue when 'sticky' strategy is used.
+  Prior to this fix, if a previously picked member unsubscribes from the group (without reconnect)
+  the message is still dispatched to it.
+  This issue only occurs when unsubscribe with the session kept.
+  Fixed in [#9119](https://github.com/emqx/emqx/pull/9119)
+
 ## v4.3.20
 
 ### Bug fixes

src/emqx_shared_sub.erl

@@ -337,7 +337,8 @@ fetch_sender_ref({Sender, Ref}) -> {Sender, Ref}.
 
 pick(sticky, ClientId, SourceTopic, Group, Topic, FailedSubs) ->
     Sub0 = erlang:get({shared_sub_sticky, Group, Topic}),
-    case is_active_sub(Sub0, FailedSubs) of
+    All = subscribers(Group, Topic),
+    case is_active_sub(Sub0, FailedSubs, All) of
         true ->
             %% the old subscriber is still alive
             %% keep using it for sticky strategy
@@ -507,8 +508,10 @@ update_stats(State) ->
     State.
 
 %% Return 'true' if the subscriber process is alive AND not in the failed list
-is_active_sub(Pid, FailedSubs) ->
-    not maps:is_key(Pid, FailedSubs) andalso is_alive_sub(Pid).
+is_active_sub(Pid, FailedSubs, All) ->
+    lists:member(Pid, All) andalso
+        (not maps:is_key(Pid, FailedSubs)) andalso
+        is_alive_sub(Pid).
 
 %% erlang:is_process_alive/1 does not work with remote pid.
 is_alive_sub(Pid) when ?IS_LOCAL_PID(Pid) ->