Merge pull request #12119 from ieQu1/EMQX-10342

Make stream renew timer configurable

apps/emqx/src/emqx_persistent_session_ds.erl

@@ -429,7 +429,8 @@ handle_timeout(
     {ok, Publishes, Session};
 handle_timeout(_ClientInfo, ?TIMER_GET_STREAMS, Session) ->
     renew_streams(Session),
-    {ok, [], emqx_session:ensure_timer(?TIMER_GET_STREAMS, 100, Session)};
+    Interval = emqx_config:get([session_persistence, renew_streams_interval]),
+    {ok, [], emqx_session:ensure_timer(?TIMER_GET_STREAMS, Interval, Session)};
 handle_timeout(_ClientInfo, ?TIMER_BUMP_LAST_ALIVE_AT, Session0) ->
     %% Note: we take a pessimistic approach here and assume that the client will be alive
     %% until the next bump timeout.  With this, we avoid garbage collecting this session

apps/emqx/src/emqx_persistent_session_ds_gc_worker.erl

@@ -82,7 +82,7 @@ try_gc() ->
     CoreNodes = mria_membership:running_core_nodelist(),
     Res = global:trans(
         {?MODULE, self()},
-        fun() -> ?tp_span(ds_session_gc, #{}, start_gc()) end,
+        fun() -> ?tp_span(debug, ds_session_gc, #{}, start_gc()) end,
         CoreNodes,
         %% Note: we set retries to 1 here because, in rare occasions, GC might start at the
         %% same time in more than one node, and each one will abort the other.  By allowing

apps/emqx/src/emqx_schema.erl

@@ -1805,6 +1805,14 @@ fields("session_persistence") ->
                     desc => ?DESC(session_ds_last_alive_update_interval)
                 }
             )},
+        {"renew_streams_interval",
+            sc(
+                timeout_duration(),
+                #{
+                    default => <<"5000ms">>,
+                    importance => ?IMPORTANCE_HIDDEN
+                }
+            )},
         {"session_gc_interval",
             sc(
                 timeout_duration(),

apps/emqx_durable_storage/src/emqx_ds.erl

@@ -195,7 +195,7 @@ store_batch(DB, Msgs) ->
 %% replay. This function returns stream together with its
 %% "coordinate": `stream_rank()'.
 %%
-%% Stream rank is a tuple of two integers, let's call them X and Y. If
+%% Stream rank is a tuple of two terms, let's call them X and Y. If
 %% X coordinate of two streams is different, they are independent and
 %% can be replayed in parallel. If it's the same, then the stream with
 %% smaller Y coordinate should be replayed first. If Y coordinates are