Merge pull request #10353 from rabbitmq/list-running-opt

Do not contact disconnected nodes in rabbit_nodes:list_running/0

Author: kjnilsson

deps/rabbit/src/rabbit_mnesia_rename.erl

@@ -144,6 +144,9 @@ maybe_finish(AllNodes) ->
 finish(FromNode, ToNode, AllNodes) ->
     case node() of
         ToNode ->
+            %% filter_running/1 does not perform reconnections so we need
+            %% to do that manually
+            _ = [net_kernel:connect_node(N) || N <- AllNodes],
             case rabbit_nodes:filter_running(AllNodes) of
                 [] -> finish_primary(FromNode, ToNode);
                 _  -> finish_secondary(FromNode, ToNode, AllNodes)

deps/rabbit/src/rabbit_node_monitor.erl

@@ -168,7 +168,7 @@ notify_node_up() ->
 
 notify_joined_cluster() ->
     NewMember = node(),
-    Nodes = rabbit_nodes:list_running() -- [NewMember],
+    Nodes = alive_rabbit_nodes() -- [NewMember],
     gen_server:abcast(Nodes, ?SERVER,
                       {joined_cluster, node(), rabbit_db_cluster:node_type()}),
 
@@ -177,7 +177,7 @@ notify_joined_cluster() ->
 -spec notify_left_cluster(node()) -> 'ok'.
 
 notify_left_cluster(Node) ->
-    Nodes = rabbit_nodes:list_running(),
+    Nodes = alive_rabbit_nodes(),
     gen_server:abcast(Nodes, ?SERVER, {left_cluster, Node}),
     ok.
 
@@ -1020,19 +1020,24 @@ alive_nodes() -> rabbit_nodes:list_reachable().
 
 alive_nodes(Nodes) -> rabbit_nodes:filter_reachable(Nodes).
 
-alive_rabbit_nodes() -> rabbit_nodes:list_running().
+alive_rabbit_nodes() ->
+    alive_rabbit_nodes(rabbit_nodes:list_members()).
 
 -spec alive_rabbit_nodes([node()]) -> [node()].
 
 alive_rabbit_nodes(Nodes) ->
+    ok = ping(Nodes),
     rabbit_nodes:filter_running(Nodes).
 
 %% This one is allowed to connect!
 
 -spec ping_all() -> 'ok'.
 
 ping_all() ->
-    [net_adm:ping(N) || N <- rabbit_nodes:list_members()],
+    ping(rabbit_nodes:list_members()).
+
+ping(Nodes) ->
+    _ = [net_adm:ping(N) || N <- Nodes],
     ok.
 
 possibly_partitioned_nodes() ->

deps/rabbit/src/rabbit_nodes.erl

@@ -366,9 +366,10 @@ filter_not_running(Nodes) ->
 do_filter_running(Members) ->
     %% All clustered members where `rabbit' is running, regardless if they are
     %% under maintenance or not.
+    ReachableMembers = do_filter_reachable(Members),
     Rets = erpc:multicall(
-             Members, rabbit, is_running, [], ?FILTER_RPC_TIMEOUT),
-    RetPerMember = lists:zip(Members, Rets),
+             ReachableMembers, rabbit, is_running, [], ?FILTER_RPC_TIMEOUT),
+    RetPerMember = lists:zip(ReachableMembers, Rets),
     lists:filtermap(
       fun
           ({Member, {ok, true}}) ->

deps/rabbit/test/clustering_management_SUITE.erl

@@ -924,29 +924,8 @@ change_cluster_when_node_offline(Config) ->
     assert_clustered([Rabbit, Hare]),
     ok = start_app(Config, Bunny),
     assert_not_clustered(Bunny),
+    ok.
 
-    %% Now the same, but Rabbit is a RAM node, and we bring up Bunny
-    %% before
-    ok = stop_app(Config, Rabbit),
-    ok = change_cluster_node_type(Config, Rabbit, ram),
-    ok = start_app(Config, Rabbit),
-    stop_join_start(Config, Bunny, Hare),
-    assert_cluster_status(
-      {[Rabbit, Hare, Bunny], [Hare, Bunny], [Rabbit, Hare, Bunny]},
-      [Rabbit, Hare, Bunny]),
-    ok = stop_app(Config, Rabbit),
-    ok = stop_app(Config, Bunny),
-    ok = reset(Config, Bunny),
-    ok = start_app(Config, Bunny),
-    assert_not_clustered(Bunny),
-    assert_cluster_status({[Rabbit, Hare], [Hare], [Hare]}, [Hare]),
-    assert_cluster_status(
-      {[Rabbit, Hare, Bunny], [Hare, Bunny], [Hare, Bunny]},
-      [Rabbit]),
-    ok = start_app(Config, Rabbit),
-    assert_cluster_status({[Rabbit, Hare], [Hare], [Rabbit, Hare]},
-                          [Rabbit, Hare]),
-    assert_not_clustered(Bunny).
 
 update_cluster_nodes(Config) ->
     [Rabbit, Hare, Bunny] = cluster_members(Config),