Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Re-evaluate stream SAC group after connection down event #13657

Merged
merged 1 commit into from
Mar 31, 2025
Merged

Re-evaluate stream SAC group after connection down event #13657

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
96 changes: 45 additions & 51 deletions deps/rabbit/src/rabbit_stream_sac_coordinator.erl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -229,7 +229,7 @@ apply(#command_unregister_consumer{vhost = VirtualHost,
of
{value, Consumer} ->
G1 = remove_from_group(Consumer, Group0),
handle_consumer_removal(G1, Consumer, Stream, ConsumerName);
handle_consumer_removal(G1, Stream, ConsumerName, Consumer#consumer.active);
false ->
{Group0, []}
end,
Expand Down Expand Up @@ -414,50 +414,44 @@ handle_connection_down(Pid,
{State0, []};
{Groups, PidsGroups1} ->
State1 = State0#?MODULE{pids_groups = PidsGroups1},
%% iterate other the groups that this PID affects
maps:fold(fun({VirtualHost, Stream, ConsumerName}, _,
{#?MODULE{groups = ConsumerGroups} = S0, Eff0}) ->
case lookup_group(VirtualHost,
Stream,
ConsumerName,
ConsumerGroups)
of
undefined -> {S0, Eff0};
#group{consumers = Consumers} ->
%% iterate over the consumers of the group
%% and unregister the ones from this PID.
%% It may not be optimal, computing the new active consumer
%% from the purged group and notifying the remaining consumers
%% appropriately should avoid unwanted notifications and even rebalancing.
lists:foldl(fun (#consumer{pid = P,
subscription_id =
SubId},
{StateSub0, EffSub0})
when P == Pid ->
{StateSub1, ok, E} =
?MODULE:apply(#command_unregister_consumer{vhost
=
VirtualHost,
stream
=
Stream,
consumer_name
=
ConsumerName,
connection_pid
=
Pid,
subscription_id
=
SubId},
StateSub0),
{StateSub1, EffSub0 ++ E};
(_Consumer, Acc) -> Acc
end,
{S0, Eff0}, Consumers)
end
end,
{State1, []}, Groups)
maps:fold(fun(G, _, Acc) ->
handle_group_after_connection_down(Pid, Acc, G)
end, {State1, []}, Groups)
end.

handle_group_after_connection_down(Pid,
{#?MODULE{groups = Groups0} = S0, Eff0},
{VirtualHost, Stream, ConsumerName}) ->
case lookup_group(VirtualHost,
Stream,
ConsumerName,
Groups0) of
undefined ->
{S0, Eff0};
#group{consumers = Consumers0} = G0 ->
%% remove the connection consumers from the group state
%% keep flags to know what happened
{Consumers1, ActiveRemoved, AnyRemoved} =
lists:foldl(
fun(#consumer{pid = P, active = S}, {L, ActiveFlag, _}) when P == Pid ->
{L, S or ActiveFlag, true};
(C, {L, ActiveFlag, AnyFlag}) ->
{L ++ [C], ActiveFlag, AnyFlag}
end, {[], false, false}, Consumers0),

case AnyRemoved of
true ->
G1 = G0#group{consumers = Consumers1},
{G2, Effects} = handle_consumer_removal(G1, Stream, ConsumerName, ActiveRemoved),
Groups1 = update_groups(VirtualHost,
Stream,
ConsumerName,
G2,
Groups0),
{S0#?MODULE{groups = Groups1}, Effects ++ Eff0};
false ->
{S0, Eff0}
end
end.

do_register_consumer(VirtualHost,
Expand Down Expand Up @@ -576,9 +570,9 @@ do_register_consumer(VirtualHost,
handle_consumer_removal(#group{consumers = []} = G, _, _, _) ->
{G, []};
handle_consumer_removal(#group{partition_index = -1} = Group0,
Consumer, Stream, ConsumerName) ->
case Consumer of
#consumer{active = true} ->
Stream, ConsumerName, ActiveRemoved) ->
case ActiveRemoved of
true ->
%% this is the active consumer we remove, computing the new one
Group1 = compute_active_consumer(Group0),
case lookup_active_consumer(Group1) of
Expand All @@ -589,11 +583,11 @@ handle_consumer_removal(#group{partition_index = -1} = Group0,
%% no active consumer found in the group, nothing to do
{Group1, []}
end;
#consumer{active = false} ->
false ->
%% not the active consumer, nothing to do.
{Group0, []}
end;
handle_consumer_removal(Group0, Consumer, Stream, ConsumerName) ->
handle_consumer_removal(Group0, Stream, ConsumerName, ActiveRemoved) ->
case lookup_active_consumer(Group0) of
{value,
#consumer{pid = ActPid, subscription_id = ActSubId} =
Expand All @@ -612,7 +606,7 @@ handle_consumer_removal(Group0, Consumer, Stream, ConsumerName) ->
Stream, ConsumerName, false, true)]}
end;
false ->
case Consumer#consumer.active of
case ActiveRemoved of
true ->
%% the active one is going away, picking a new one
#consumer{pid = P, subscription_id = SID} =
Expand Down
187 changes: 175 additions & 12 deletions deps/rabbit/test/rabbit_stream_sac_coordinator_SUITE.erl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -312,29 +312,27 @@ ensure_monitors_test(_) ->

ok.

handle_connection_down_test(_) ->
handle_connection_down_sac_should_get_activated_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Pid1 = spawn(fun() -> ok end),
Group =
cgroup([consumer(Pid0, 0, true), consumer(Pid1, 1, false),
consumer(Pid0, 2, false)]),
State0 =
state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]),
Pid1 => maps:from_list([{GroupId, true}])}),
Group = cgroup([consumer(Pid0, 0, true),
consumer(Pid1, 1, false),
consumer(Pid0, 2, false)]),
State0 = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]),
Pid1 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups1, groups = Groups1} = State1,
Effects1} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State0),
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State0),
assertSize(1, PidsGroups1),
assertSize(1, maps:get(Pid1, PidsGroups1)),
assertSendMessageEffect(Pid1, 1, Stream, ConsumerName, true, Effects1),
?assertEqual(#{GroupId => cgroup([consumer(Pid1, 1, true)])},
Groups1),
{#?STATE{pids_groups = PidsGroups2, groups = Groups2} = _State2,
assertHasGroup(GroupId, cgroup([consumer(Pid1, 1, true)]), Groups1),
{#?STATE{pids_groups = PidsGroups2, groups = Groups2},
Effects2} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid1, State1),
assertEmpty(PidsGroups2),
Expand All @@ -343,6 +341,168 @@ handle_connection_down_test(_) ->

ok.

handle_connection_down_sac_active_does_not_change_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Pid1 = spawn(fun() -> ok end),
Group = cgroup([consumer(Pid1, 0, true),
consumer(Pid0, 1, false),
consumer(Pid0, 2, false)]),
State = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]),
Pid1 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups, groups = Groups},
Effects} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State),
assertSize(1, PidsGroups),
assertSize(1, maps:get(Pid1, PidsGroups)),
assertEmpty(Effects),
assertHasGroup(GroupId, cgroup([consumer(Pid1, 0, true)]), Groups),
ok.

handle_connection_down_sac_no_more_consumers_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Group = cgroup([consumer(Pid0, 0, true),
consumer(Pid0, 1, false)]),
State = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups, groups = Groups},
Effects} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State),
assertEmpty(PidsGroups),
assertEmpty(Groups),
assertEmpty(Effects),
ok.

handle_connection_down_sac_no_consumers_in_down_connection_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Pid1 = spawn(fun() -> ok end),
Group = cgroup([consumer(Pid1, 0, true),
consumer(Pid1, 1, false)]),
State = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]), %% should not be there
Pid1 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups, groups = Groups},
Effects} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State),

assertSize(1, PidsGroups),
assertSize(1, maps:get(Pid1, PidsGroups)),
assertEmpty(Effects),
assertHasGroup(GroupId, cgroup([consumer(Pid1, 0, true), consumer(Pid1, 1, false)]),
Groups),
ok.

handle_connection_down_super_stream_active_stays_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Pid1 = spawn(fun() -> ok end),
Group = cgroup(1, [consumer(Pid0, 0, false),
consumer(Pid0, 1, true),
consumer(Pid1, 2, false),
consumer(Pid1, 3, false)]),
State = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]),
Pid1 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups, groups = Groups},
Effects} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid1, State),
assertSize(1, PidsGroups),
assertSize(1, maps:get(Pid0, PidsGroups)),
assertEmpty(Effects),
assertHasGroup(GroupId, cgroup(1, [consumer(Pid0, 0, false), consumer(Pid0, 1, true)]),
Groups),
ok.

handle_connection_down_super_stream_active_changes_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Pid1 = spawn(fun() -> ok end),
Group = cgroup(1, [consumer(Pid0, 0, false),
consumer(Pid1, 1, true),
consumer(Pid0, 2, false),
consumer(Pid1, 3, false)]),
State = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]),
Pid1 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups, groups = Groups},
Effects} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State),
assertSize(1, PidsGroups),
assertSize(1, maps:get(Pid1, PidsGroups)),
assertSendMessageSteppingDownEffect(Pid1, 1, Stream, ConsumerName, Effects),
assertHasGroup(GroupId, cgroup(1, [consumer(Pid1, 1, false), consumer(Pid1, 3, false)]),
Groups),
ok.

handle_connection_down_super_stream_activate_in_remaining_connection_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Pid1 = spawn(fun() -> ok end),
Group = cgroup(1, [consumer(Pid0, 0, false),
consumer(Pid0, 1, true),
consumer(Pid1, 2, false),
consumer(Pid1, 3, false)]),
State = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]),
Pid1 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups, groups = Groups},
Effects} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State),
assertSize(1, PidsGroups),
assertSize(1, maps:get(Pid1, PidsGroups)),
assertSendMessageEffect(Pid1, 3, Stream, ConsumerName, true, Effects),
assertHasGroup(GroupId, cgroup(1, [consumer(Pid1, 2, false), consumer(Pid1, 3, true)]),
Groups),
ok.

handle_connection_down_super_stream_no_active_removed_or_present_test(_) ->
Stream = <<"stream">>,
ConsumerName = <<"app">>,
GroupId = {<<"/">>, Stream, ConsumerName},
Pid0 = self(),
Pid1 = spawn(fun() -> ok end),
%% this is a weird case that should not happen in the wild,
%% we test the logic in the code nevertheless.
%% No active consumer in the group
Group = cgroup(1, [consumer(Pid0, 0, false),
consumer(Pid0, 1, false),
consumer(Pid1, 2, false),
consumer(Pid1, 3, false)]),
State = state(#{GroupId => Group},
#{Pid0 => maps:from_list([{GroupId, true}]),
Pid1 => maps:from_list([{GroupId, true}])}),

{#?STATE{pids_groups = PidsGroups, groups = Groups},
Effects} =
rabbit_stream_sac_coordinator:handle_connection_down(Pid0, State),
assertSize(1, PidsGroups),
assertSize(1, maps:get(Pid1, PidsGroups)),
assertEmpty(Effects),
assertHasGroup(GroupId, cgroup(1, [consumer(Pid1, 2, false), consumer(Pid1, 3, false)]),
Groups),
ok.

assertSize(Expected, []) ->
?assertEqual(Expected, 0);
assertSize(Expected, Map) when is_map(Map) ->
Expand All @@ -353,6 +513,9 @@ assertSize(Expected, List) when is_list(List) ->
assertEmpty(Data) ->
assertSize(0, Data).

assertHasGroup(GroupId, Group, Groups) ->
?assertEqual(#{GroupId => Group}, Groups).

consumer(Pid, SubId, Active) ->
#consumer{pid = Pid,
subscription_id = SubId,
Expand Down
14 changes: 2 additions & 12 deletions deps/rabbitmq_stream/src/rabbit_stream_reader.erl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -598,26 +598,16 @@ augment_infos_with_user_provided_connection_name(Infos,
end.

close(Transport,
#stream_connection{socket = S, virtual_host = VirtualHost,
outstanding_requests = Requests},
#stream_connection{socket = S},
#stream_connection_state{consumers = Consumers}) ->
[begin
%% we discard the result (updated requests) because they are no longer used
_ = maybe_unregister_consumer(VirtualHost, Consumer,
single_active_consumer(Properties),
Requests),
case Log of
undefined ->
ok; %% segment may not be defined on subscription (single active consumer)
L ->
osiris_log:close(L)
end
end
|| #consumer{log = Log,
configuration =
#consumer_configuration{properties = Properties}} =
Consumer
<- maps:values(Consumers)],
end || #consumer{log = Log} <- maps:values(Consumers)],
Transport:shutdown(S, write),
Transport:close(S).

Expand Down
Loading