yuanbiao
/
emqx
1
0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

refactor(resman): work with state cache atomically 

Also ensure that cache entries are always consistent with `Data`,
so that most of the code could rely on reading the cached entry
most of the time.
This commit is contained in:
Andrew Mayorov 2023-03-13 18:28:28 +03:00
parent b3e7e51094
commit e411c5d5f8
No known key found for this signature in database
GPG Key ID: 2837C62ACFBFED5D
3 changed files with 423 additions and 279 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

158
apps/emqx_resource/src/emqx_resource_manager.erl
View File

@ -18,6 +18,7 @@
-include("emqx_resource.hrl"). -include("emqx_resource.hrl").
-include_lib("emqx/include/logger.hrl"). -include_lib("emqx/include/logger.hrl").
-include_lib("snabbkaffe/include/trace.hrl").
% API % API
-export([ -export([
@ -303,26 +304,30 @@ start_link(MgrId, ResId, Group, ResourceType, Config, Opts) ->
query_mode = maps:get(query_mode, Opts, sync), query_mode = maps:get(query_mode, Opts, sync),
config = Config, config = Config,
opts = Opts, opts = Opts,
status = connecting,
state = undefined, state = undefined,
error = undefined error = undefined
}, },
gen_statem:start_link(?MODULE, {Data, Opts}, []). gen_statem:start_link(?MODULE, {Data, Opts}, []).
init({Data, Opts}) -> init({DataIn, Opts}) ->
process_flag(trap_exit, true), process_flag(trap_exit, true),
%% init the cache so that lookup/1 will always return something Data = DataIn#data{pid = self()},
DataWithPid = Data#data{pid = self()},
insert_cache(DataWithPid#data.id, DataWithPid#data.group, DataWithPid),
case maps:get(start_after_created, Opts, ?START_AFTER_CREATED) of case maps:get(start_after_created, Opts, ?START_AFTER_CREATED) of
true -> {ok, connecting, DataWithPid, {next_event, internal, start_resource}}; true ->
false -> {ok, stopped, DataWithPid} %% init the cache so that lookup/1 will always return something
UpdatedData = update_state(Data#data{status = connecting}),
{ok, connecting, UpdatedData, {next_event, internal, start_resource}};
false ->
%% init the cache so that lookup/1 will always return something
UpdatedData = update_state(Data#data{status = stopped}),
{ok, stopped, UpdatedData}
end. end.
terminate({shutdown, removed}, _State, _Data) ->
ok;
terminate(_Reason, _State, Data) -> terminate(_Reason, _State, Data) ->
_ = stop_resource(Data), _ = maybe_stop_resource(Data),
_ = maybe_clear_alarm(Data#data.id), ok = delete_cache(Data#data.id, Data#data.manager_id),
delete_cache(Data#data.id, Data#data.manager_id),
ok. ok.
%% Behavior callback %% Behavior callback
@ -333,11 +338,12 @@ callback_mode() -> [handle_event_function, state_enter].
% Called during testing to force a specific state % Called during testing to force a specific state
handle_event({call, From}, set_resource_status_connecting, _State, Data) -> handle_event({call, From}, set_resource_status_connecting, _State, Data) ->
{next_state, connecting, Data#data{status = connecting}, [{reply, From, ok}]}; UpdatedData = update_state(Data#data{status = connecting}, Data),
{next_state, connecting, UpdatedData, [{reply, From, ok}]};
% Called when the resource is to be restarted % Called when the resource is to be restarted
handle_event({call, From}, restart, _State, Data) -> handle_event({call, From}, restart, _State, Data) ->
_ = stop_resource(Data), DataNext = stop_resource(Data),
start_resource(Data, From); start_resource(DataNext, From);
% Called when the resource is to be started (also used for manual reconnect) % Called when the resource is to be started (also used for manual reconnect)
handle_event({call, From}, start, State, Data) when handle_event({call, From}, start, State, Data) when
State =:= stopped orelse State =:= stopped orelse
@ -347,16 +353,14 @@ handle_event({call, From}, start, State, Data) when
handle_event({call, From}, start, _State, _Data) -> handle_event({call, From}, start, _State, _Data) ->
{keep_state_and_data, [{reply, From, ok}]}; {keep_state_and_data, [{reply, From, ok}]};
% Called when the resource received a `quit` message % Called when the resource received a `quit` message
handle_event(info, quit, stopped, _Data) ->
{stop, {shutdown, quit}};
handle_event(info, quit, _State, _Data) -> handle_event(info, quit, _State, _Data) ->
{stop, {shutdown, quit}}; {stop, {shutdown, quit}};
% Called when the resource is to be stopped % Called when the resource is to be stopped
handle_event({call, From}, stop, stopped, _Data) -> handle_event({call, From}, stop, stopped, _Data) ->
{keep_state_and_data, [{reply, From, ok}]}; {keep_state_and_data, [{reply, From, ok}]};
handle_event({call, From}, stop, _State, Data) -> handle_event({call, From}, stop, _State, Data) ->
Result = stop_resource(Data), UpdatedData = stop_resource(Data),
{next_state, stopped, Data, [{reply, From, Result}]}; {next_state, stopped, update_state(UpdatedData, Data), [{reply, From, ok}]};
% Called when a resource is to be stopped and removed. % Called when a resource is to be stopped and removed.
handle_event({call, From}, {remove, ClearMetrics}, _State, Data) -> handle_event({call, From}, {remove, ClearMetrics}, _State, Data) ->
handle_remove_event(From, ClearMetrics, Data); handle_remove_event(From, ClearMetrics, Data);
@ -371,11 +375,9 @@ handle_event({call, From}, health_check, stopped, _Data) ->
handle_event({call, From}, health_check, _State, Data) -> handle_event({call, From}, health_check, _State, Data) ->
handle_manually_health_check(From, Data); handle_manually_health_check(From, Data);
% State: CONNECTING % State: CONNECTING
handle_event(enter, _OldState, connecting, Data) -> handle_event(enter, _OldState, connecting = State, Data) ->
UpdatedData = Data#data{status = connecting}, ok = log_state_consistency(State, Data),
insert_cache(Data#data.id, Data#data.group, Data), {keep_state_and_data, [{state_timeout, 0, health_check}]};
Actions = [{state_timeout, 0, health_check}],
{keep_state, UpdatedData, Actions};
handle_event(internal, start_resource, connecting, Data) -> handle_event(internal, start_resource, connecting, Data) ->
start_resource(Data, undefined); start_resource(Data, undefined);
handle_event(state_timeout, health_check, connecting, Data) -> handle_event(state_timeout, health_check, connecting, Data) ->
@ -383,27 +385,23 @@ handle_event(state_timeout, health_check, connecting, Data) ->
%% State: CONNECTED %% State: CONNECTED
%% The connected state is entered after a successful on_start/2 of the callback mod %% The connected state is entered after a successful on_start/2 of the callback mod
%% and successful health_checks %% and successful health_checks
handle_event(enter, _OldState, connected, Data) -> handle_event(enter, _OldState, connected = State, Data) ->
UpdatedData = Data#data{status = connected}, ok = log_state_consistency(State, Data),
insert_cache(Data#data.id, Data#data.group, UpdatedData),
_ = emqx_alarm:deactivate(Data#data.id), _ = emqx_alarm:deactivate(Data#data.id),
Actions = [{state_timeout, health_check_interval(Data#data.opts), health_check}], {keep_state_and_data, health_check_actions(Data)};
{next_state, connected, UpdatedData, Actions};
handle_event(state_timeout, health_check, connected, Data) -> handle_event(state_timeout, health_check, connected, Data) ->
handle_connected_health_check(Data); handle_connected_health_check(Data);
%% State: DISCONNECTED %% State: DISCONNECTED
handle_event(enter, _OldState, disconnected, Data) -> handle_event(enter, _OldState, disconnected = State, Data) ->
UpdatedData = Data#data{status = disconnected}, ok = log_state_consistency(State, Data),
insert_cache(Data#data.id, Data#data.group, UpdatedData), {keep_state_and_data, retry_actions(Data)};
handle_disconnected_state_enter(UpdatedData);
handle_event(state_timeout, auto_retry, disconnected, Data) -> handle_event(state_timeout, auto_retry, disconnected, Data) ->
start_resource(Data, undefined); start_resource(Data, undefined);
%% State: STOPPED %% State: STOPPED
%% The stopped state is entered after the resource has been explicitly stopped %% The stopped state is entered after the resource has been explicitly stopped
handle_event(enter, _OldState, stopped, Data) -> handle_event(enter, _OldState, stopped = State, Data) ->
UpdatedData = Data#data{status = stopped}, ok = log_state_consistency(State, Data),
insert_cache(Data#data.id, Data#data.group, UpdatedData), {keep_state_and_data, []};
{next_state, stopped, UpdatedData};
% Ignore all other events % Ignore all other events
handle_event(EventType, EventData, State, Data) -> handle_event(EventType, EventData, State, Data) ->
?SLOG( ?SLOG(
@ -418,6 +416,22 @@ handle_event(EventType, EventData, State, Data) ->
), ),
keep_state_and_data. keep_state_and_data.
log_state_consistency(State, #data{status = State} = Data) ->
log_cache_consistency(read_cache(Data#data.id), Data);
log_state_consistency(State, Data) ->
?tp(warning, "inconsistent_state", #{
state => State,
data => Data
}).
log_cache_consistency({_, Data}, Data) ->
ok;
log_cache_consistency({_, DataCached}, Data) ->
?tp(warning, "inconsistent_cache", #{
cache => DataCached,
data => Data
}).
%%------------------------------------------------------------------------------ %%------------------------------------------------------------------------------
%% internal functions %% internal functions
%%------------------------------------------------------------------------------ %%------------------------------------------------------------------------------
@ -451,10 +465,12 @@ delete_cache(ResId, MgrId) ->
end. end.
do_delete_cache(<<?TEST_ID_PREFIX, _/binary>> = ResId) -> do_delete_cache(<<?TEST_ID_PREFIX, _/binary>> = ResId) ->
ets:delete(?ETS_TABLE, {owner, ResId}), true = ets:delete(?ETS_TABLE, {owner, ResId}),
ets:delete(?ETS_TABLE, ResId); true = ets:delete(?ETS_TABLE, ResId),
ok;
do_delete_cache(ResId) -> do_delete_cache(ResId) ->
ets:delete(?ETS_TABLE, ResId). true = ets:delete(?ETS_TABLE, ResId),
ok.
set_new_owner(ResId) -> set_new_owner(ResId) ->
MgrId = make_manager_id(ResId), MgrId = make_manager_id(ResId),
@ -471,9 +487,6 @@ get_owner(ResId) ->
[] -> not_found [] -> not_found
end. end.
handle_disconnected_state_enter(Data) ->
{next_state, disconnected, Data, retry_actions(Data)}.
retry_actions(Data) -> retry_actions(Data) ->
case maps:get(auto_restart_interval, Data#data.opts, ?AUTO_RESTART_INTERVAL) of case maps:get(auto_restart_interval, Data#data.opts, ?AUTO_RESTART_INTERVAL) of
undefined -> undefined ->
@ -482,24 +495,27 @@ retry_actions(Data) ->
[{state_timeout, RetryInterval, auto_retry}] [{state_timeout, RetryInterval, auto_retry}]
end. end.
health_check_actions(Data) ->
[{state_timeout, health_check_interval(Data#data.opts), health_check}].
handle_remove_event(From, ClearMetrics, Data) -> handle_remove_event(From, ClearMetrics, Data) ->
stop_resource(Data), _ = stop_resource(Data),
ok = delete_cache(Data#data.id, Data#data.manager_id),
ok = emqx_resource_buffer_worker_sup:stop_workers(Data#data.id, Data#data.opts), ok = emqx_resource_buffer_worker_sup:stop_workers(Data#data.id, Data#data.opts),
case ClearMetrics of case ClearMetrics of
true -> ok = emqx_metrics_worker:clear_metrics(?RES_METRICS, Data#data.id); true -> ok = emqx_metrics_worker:clear_metrics(?RES_METRICS, Data#data.id);
false -> ok false -> ok
end, end,
{stop_and_reply, normal, [{reply, From, ok}]}. {stop_and_reply, {shutdown, removed}, [{reply, From, ok}]}.
start_resource(Data, From) -> start_resource(Data, From) ->
%% in case the emqx_resource:call_start/2 hangs, the lookup/1 can read status from the cache %% in case the emqx_resource:call_start/2 hangs, the lookup/1 can read status from the cache
insert_cache(Data#data.id, Data#data.group, Data),
case emqx_resource:call_start(Data#data.manager_id, Data#data.mod, Data#data.config) of case emqx_resource:call_start(Data#data.manager_id, Data#data.mod, Data#data.config) of
{ok, ResourceState} -> {ok, ResourceState} ->
UpdatedData = Data#data{state = ResourceState, status = connecting}, UpdatedData = Data#data{status = connecting, state = ResourceState},
%% Perform an initial health_check immediately before transitioning into a connected state %% Perform an initial health_check immediately before transitioning into a connected state
Actions = maybe_reply([{state_timeout, 0, health_check}], From, ok), Actions = maybe_reply([{state_timeout, 0, health_check}], From, ok),
{next_state, connecting, UpdatedData, Actions}; {next_state, connecting, update_state(UpdatedData, Data), Actions};
{error, Reason} = Err -> {error, Reason} = Err ->
?SLOG(warning, #{ ?SLOG(warning, #{
msg => start_resource_failed, msg => start_resource_failed,
@ -509,34 +525,42 @@ start_resource(Data, From) ->
_ = maybe_alarm(disconnected, Data#data.id), _ = maybe_alarm(disconnected, Data#data.id),
%% Keep track of the error reason why the connection did not work %% Keep track of the error reason why the connection did not work
%% so that the Reason can be returned when the verification call is made. %% so that the Reason can be returned when the verification call is made.
UpdatedData = Data#data{error = Reason}, UpdatedData = Data#data{status = disconnected, error = Reason},
Actions = maybe_reply(retry_actions(UpdatedData), From, Err), Actions = maybe_reply(retry_actions(UpdatedData), From, Err),
{next_state, disconnected, UpdatedData, Actions} {next_state, disconnected, update_state(UpdatedData, Data), Actions}
end. end.
stop_resource(#data{state = undefined, id = ResId} = _Data) -> maybe_stop_resource(#data{status = Status} = Data) when Status /= stopped ->
_ = maybe_clear_alarm(ResId), stop_resource(Data);
ok = emqx_metrics_worker:reset_metrics(?RES_METRICS, ResId), maybe_stop_resource(#data{status = stopped} = Data) ->
ok; Data.
stop_resource(Data) ->
stop_resource(#data{state = ResState, id = ResId} = Data) ->
%% We don't care the return value of the Mod:on_stop/2. %% We don't care the return value of the Mod:on_stop/2.
%% The callback mod should make sure the resource is stopped after on_stop/2 %% The callback mod should make sure the resource is stopped after on_stop/2
%% is returned. %% is returned.
ResId = Data#data.id, case ResState /= undefined of
_ = emqx_resource:call_stop(Data#data.manager_id, Data#data.mod, Data#data.state), true ->
emqx_resource:call_stop(Data#data.manager_id, Data#data.mod, ResState);
false ->
ok
end,
_ = maybe_clear_alarm(ResId), _ = maybe_clear_alarm(ResId),
ok = emqx_metrics_worker:reset_metrics(?RES_METRICS, ResId), ok = emqx_metrics_worker:reset_metrics(?RES_METRICS, ResId),
ok. Data#data{status = stopped}.
make_test_id() -> make_test_id() ->
RandId = iolist_to_binary(emqx_misc:gen_id(16)), RandId = iolist_to_binary(emqx_misc:gen_id(16)),
<<?TEST_ID_PREFIX, RandId/binary>>. <<?TEST_ID_PREFIX, RandId/binary>>.
handle_manually_health_check(From, Data) -> handle_manually_health_check(From, Data) ->
with_health_check(Data, fun(Status, UpdatedData) -> with_health_check(
Data,
fun(Status, UpdatedData) ->
Actions = [{reply, From, {ok, Status}}], Actions = [{reply, From, {ok, Status}}],
{next_state, Status, UpdatedData, Actions} {next_state, Status, UpdatedData, Actions}
end). end
).
handle_connecting_health_check(Data) -> handle_connecting_health_check(Data) ->
with_health_check( with_health_check(
@ -545,8 +569,7 @@ handle_connecting_health_check(Data) ->
(connected, UpdatedData) -> (connected, UpdatedData) ->
{next_state, connected, UpdatedData}; {next_state, connected, UpdatedData};
(connecting, UpdatedData) -> (connecting, UpdatedData) ->
Actions = [{state_timeout, health_check_interval(Data#data.opts), health_check}], {keep_state, UpdatedData, health_check_actions(UpdatedData)};
{keep_state, UpdatedData, Actions};
(disconnected, UpdatedData) -> (disconnected, UpdatedData) ->
{next_state, disconnected, UpdatedData} {next_state, disconnected, UpdatedData}
end end
@ -557,8 +580,7 @@ handle_connected_health_check(Data) ->
Data, Data,
fun fun
(connected, UpdatedData) -> (connected, UpdatedData) ->
Actions = [{state_timeout, health_check_interval(Data#data.opts), health_check}], {keep_state, UpdatedData, health_check_actions(UpdatedData)};
{keep_state, UpdatedData, Actions};
(Status, UpdatedData) -> (Status, UpdatedData) ->
?SLOG(warning, #{ ?SLOG(warning, #{
msg => health_check_failed, msg => health_check_failed,
@ -580,8 +602,16 @@ with_health_check(Data, Func) ->
UpdatedData = Data#data{ UpdatedData = Data#data{
state = NewState, status = Status, error = Err state = NewState, status = Status, error = Err
}, },
insert_cache(ResId, UpdatedData#data.group, UpdatedData), Func(Status, update_state(UpdatedData, Data)).
Func(Status, UpdatedData).
update_state(Data) ->
update_state(Data, undefined).
update_state(DataWas, DataWas) ->
DataWas;
update_state(Data, _DataWas) ->
_ = insert_cache(Data#data.id, Data#data.group, Data),
Data.
health_check_interval(Opts) -> health_check_interval(Opts) ->
maps:get(health_check_interval, Opts, ?HEALTHCHECK_INTERVAL). maps:get(health_check_interval, Opts, ?HEALTHCHECK_INTERVAL).

12
apps/emqx_resource/test/emqx_connector_demo.erl
View File

@ -75,8 +75,7 @@ on_start(InstId, #{name := Name} = Opts) ->
on_stop(_InstId, #{stop_error := true}) -> on_stop(_InstId, #{stop_error := true}) ->
{error, stop_error}; {error, stop_error};
on_stop(_InstId, #{pid := Pid}) -> on_stop(_InstId, #{pid := Pid}) ->
erlang:exit(Pid, shutdown), stop_counter_process(Pid).
ok.
on_query(_InstId, get_state, State) -> on_query(_InstId, get_state, State) ->
{ok, State}; {ok, State};
@ -247,6 +246,15 @@ spawn_counter_process(Name, Register) ->
true = maybe_register(Name, Pid, Register), true = maybe_register(Name, Pid, Register),
Pid. Pid.
stop_counter_process(Pid) ->
true = erlang:is_process_alive(Pid),
true = erlang:exit(Pid, shutdown),
receive
{'EXIT', Pid, shutdown} -> ok
after 5000 ->
{error, timeout}
end.
counter_loop() -> counter_loop() ->
counter_loop(#{ counter_loop(#{
counter => 0, counter => 0,

184
apps/emqx_resource/test/emqx_resource_SUITE.erl
View File

@ -72,48 +72,74 @@ t_check_config(_) ->
{error, _} = emqx_resource:check_config(?TEST_RESOURCE, #{invalid => config}). {error, _} = emqx_resource:check_config(?TEST_RESOURCE, #{invalid => config}).
t_create_remove(_) -> t_create_remove(_) ->
{error, _} = emqx_resource:check_and_create_local( ?check_trace(
begin
?assertMatch(
{error, _},
emqx_resource:check_and_create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{unknown => test_resource} #{unknown => test_resource}
)
), ),
{ok, _} = emqx_resource:create( ?assertMatch(
{ok, _},
emqx_resource:create(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{name => test_resource} #{name => test_resource}
)
), ),
{ok, _} = emqx_resource:recreate( ?assertMatch(
{ok, _},
emqx_resource:recreate(
?ID, ?ID,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{name => test_resource}, #{name => test_resource},
#{} #{}
)
), ),
{ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state),
?assert(is_process_alive(Pid)), ?assert(is_process_alive(Pid)),
ok = emqx_resource:remove(?ID), ?assertEqual(ok, emqx_resource:remove(?ID)),
{error, _} = emqx_resource:remove(?ID), ?assertMatch({error, _}, emqx_resource:remove(?ID)),
?assertNot(is_process_alive(Pid)). ?assertNot(is_process_alive(Pid))
end,
fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
).
t_create_remove_local(_) -> t_create_remove_local(_) ->
{error, _} = emqx_resource:check_and_create_local( ?check_trace(
begin
?assertMatch(
{error, _},
emqx_resource:check_and_create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{unknown => test_resource} #{unknown => test_resource}
)
), ),
{ok, _} = emqx_resource:create_local( ?assertMatch(
{ok, _},
emqx_resource:create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{name => test_resource} #{name => test_resource}
)
), ),
emqx_resource:recreate_local( emqx_resource:recreate_local(
@ -122,6 +148,7 @@ t_create_remove_local(_) ->
#{name => test_resource}, #{name => test_resource},
#{} #{}
), ),
{ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state),
?assert(is_process_alive(Pid)), ?assert(is_process_alive(Pid)),
@ -135,23 +162,34 @@ t_create_remove_local(_) ->
#{} #{}
), ),
ok = emqx_resource:remove_local(?ID), ?assertEqual(ok, emqx_resource:remove_local(?ID)),
{error, _} = emqx_resource:remove_local(?ID), ?assertMatch({error, _}, emqx_resource:remove_local(?ID)),
?assertMatch( ?assertMatch(
?RESOURCE_ERROR(not_found), ?RESOURCE_ERROR(not_found),
emqx_resource:query(?ID, get_state) emqx_resource:query(?ID, get_state)
), ),
?assertNot(is_process_alive(Pid)).
?assertNot(is_process_alive(Pid))
end,
fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
).
t_do_not_start_after_created(_) -> t_do_not_start_after_created(_) ->
ct:pal("creating resource"), ?check_trace(
{ok, _} = emqx_resource:create_local( begin
?assertMatch(
{ok, _},
emqx_resource:create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{name => test_resource}, #{name => test_resource},
#{start_after_created => false} #{start_after_created => false}
)
), ),
%% the resource should remain `disconnected` after created %% the resource should remain `disconnected` after created
timer:sleep(200), timer:sleep(200),
@ -165,22 +203,25 @@ t_do_not_start_after_created(_) ->
), ),
%% start the resource manually.. %% start the resource manually..
ct:pal("starting resource manually"), ?assertEqual(ok, emqx_resource:start(?ID)),
ok = emqx_resource:start(?ID),
{ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state),
?assert(is_process_alive(Pid)), ?assert(is_process_alive(Pid)),
%% restart the resource %% restart the resource
ct:pal("restarting resource"), ?assertEqual(ok, emqx_resource:restart(?ID)),
ok = emqx_resource:restart(?ID),
?assertNot(is_process_alive(Pid)), ?assertNot(is_process_alive(Pid)),
{ok, #{pid := Pid2}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid2}} = emqx_resource:query(?ID, get_state),
?assert(is_process_alive(Pid2)), ?assert(is_process_alive(Pid2)),
ct:pal("removing resource"), ?assertEqual(ok, emqx_resource:remove_local(?ID)),
ok = emqx_resource:remove_local(?ID),
?assertNot(is_process_alive(Pid2)). ?assertNot(is_process_alive(Pid2))
end,
fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
).
t_query(_) -> t_query(_) ->
{ok, _} = emqx_resource:create_local( {ok, _} = emqx_resource:create_local(
@ -771,33 +812,51 @@ t_query_counter_async_inflight_batch(_) ->
ok = emqx_resource:remove_local(?ID). ok = emqx_resource:remove_local(?ID).
t_healthy_timeout(_) -> t_healthy_timeout(_) ->
{ok, _} = emqx_resource:create_local( ?check_trace(
begin
?assertMatch(
{ok, _},
emqx_resource:create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{name => <<"bad_not_atom_name">>, register => true}, #{name => <<"bad_not_atom_name">>, register => true},
%% the ?TEST_RESOURCE always returns the `Mod:on_get_status/2` 300ms later. %% the ?TEST_RESOURCE always returns the `Mod:on_get_status/2` 300ms later.
#{health_check_interval => 200} #{health_check_interval => 200}
)
), ),
?assertMatch( ?assertMatch(
{error, {resource_error, #{reason := timeout}}}, {error, {resource_error, #{reason := timeout}}},
emqx_resource:query(?ID, get_state, #{timeout => 1_000}) emqx_resource:query(?ID, get_state, #{timeout => 1_000})
), ),
?assertMatch({ok, _Group, #{status := disconnected}}, emqx_resource_manager:ets_lookup(?ID)), ?assertMatch(
ok = emqx_resource:remove_local(?ID). {ok, _Group, #{status := disconnected}}, emqx_resource_manager:lookup(?ID)
),
?assertEqual(ok, emqx_resource:remove_local(?ID))
end,
fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
).
t_healthy(_) -> t_healthy(_) ->
{ok, _} = emqx_resource:create_local( ?check_trace(
begin
?assertMatch(
{ok, _},
emqx_resource:create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{name => test_resource} #{name => test_resource}
)
), ),
{ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid}} = emqx_resource:query(?ID, get_state),
timer:sleep(300), timer:sleep(300),
emqx_resource:set_resource_status_connecting(?ID), emqx_resource:set_resource_status_connecting(?ID),
{ok, connected} = emqx_resource:health_check(?ID), ?assertEqual({ok, connected}, emqx_resource:health_check(?ID)),
?assertMatch( ?assertMatch(
[#{status := connected}], [#{status := connected}],
emqx_resource:list_instances_verbose() emqx_resource:list_instances_verbose()
@ -812,21 +871,35 @@ t_healthy(_) ->
emqx_resource:list_instances_verbose() emqx_resource:list_instances_verbose()
), ),
ok = emqx_resource:remove_local(?ID). ?assertEqual(ok, emqx_resource:remove_local(?ID))
end,
fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
).
t_stop_start(_) -> t_stop_start(_) ->
{error, _} = emqx_resource:check_and_create( ?check_trace(
begin
?assertMatch(
{error, _},
emqx_resource:check_and_create(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{unknown => test_resource} #{unknown => test_resource}
)
), ),
{ok, _} = emqx_resource:check_and_create( ?assertMatch(
{ok, _},
emqx_resource:check_and_create(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{<<"name">> => <<"test_resource">>} #{<<"name">> => <<"test_resource">>}
)
), ),
%% add some metrics to test their persistence %% add some metrics to test their persistence
@ -836,11 +909,14 @@ t_stop_start(_) ->
emqx_resource_metrics:inflight_set(?ID, WorkerID1, 3), emqx_resource_metrics:inflight_set(?ID, WorkerID1, 3),
?assertEqual(5, emqx_resource_metrics:inflight_get(?ID)), ?assertEqual(5, emqx_resource_metrics:inflight_get(?ID)),
{ok, _} = emqx_resource:check_and_recreate( ?assertMatch(
{ok, _},
emqx_resource:check_and_recreate(
?ID, ?ID,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{<<"name">> => <<"test_resource">>}, #{<<"name">> => <<"test_resource">>},
#{} #{}
)
), ),
{ok, #{pid := Pid0}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid0}} = emqx_resource:query(?ID, get_state),
@ -861,7 +937,7 @@ t_stop_start(_) ->
emqx_resource:query(?ID, get_state) emqx_resource:query(?ID, get_state)
), ),
ok = emqx_resource:restart(?ID), ?assertEqual(ok, emqx_resource:restart(?ID)),
timer:sleep(300), timer:sleep(300),
{ok, #{pid := Pid1}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid1}} = emqx_resource:query(?ID, get_state),
@ -873,38 +949,54 @@ t_stop_start(_) ->
emqx_resource_metrics:inflight_set(?ID, WorkerID0, 1), emqx_resource_metrics:inflight_set(?ID, WorkerID0, 1),
emqx_resource_metrics:inflight_set(?ID, WorkerID1, 4), emqx_resource_metrics:inflight_set(?ID, WorkerID1, 4),
?assertEqual(5, emqx_resource_metrics:inflight_get(?ID)), ?assertEqual(5, emqx_resource_metrics:inflight_get(?ID)),
ok = emqx_resource:stop(?ID), ?assertEqual(ok, emqx_resource:stop(?ID)),
?assertEqual(0, emqx_resource_metrics:inflight_get(?ID)), ?assertEqual(0, emqx_resource_metrics:inflight_get(?ID))
end,
ok. fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
).
t_stop_start_local(_) -> t_stop_start_local(_) ->
{error, _} = emqx_resource:check_and_create_local( ?check_trace(
begin
?assertMatch(
{error, _},
emqx_resource:check_and_create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{unknown => test_resource} #{unknown => test_resource}
)
), ),
{ok, _} = emqx_resource:check_and_create_local( ?assertMatch(
{ok, _},
emqx_resource:check_and_create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{<<"name">> => <<"test_resource">>} #{<<"name">> => <<"test_resource">>}
)
), ),
{ok, _} = emqx_resource:check_and_recreate_local( ?assertMatch(
{ok, _},
emqx_resource:check_and_recreate_local(
?ID, ?ID,
?TEST_RESOURCE, ?TEST_RESOURCE,
#{<<"name">> => <<"test_resource">>}, #{<<"name">> => <<"test_resource">>},
#{} #{}
)
), ),
{ok, #{pid := Pid0}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid0}} = emqx_resource:query(?ID, get_state),
?assert(is_process_alive(Pid0)), ?assert(is_process_alive(Pid0)),
ok = emqx_resource:stop(?ID), ?assertEqual(ok, emqx_resource:stop(?ID)),
?assertNot(is_process_alive(Pid0)), ?assertNot(is_process_alive(Pid0)),
@ -913,11 +1005,17 @@ t_stop_start_local(_) ->
emqx_resource:query(?ID, get_state) emqx_resource:query(?ID, get_state)
), ),
ok = emqx_resource:restart(?ID), ?assertEqual(ok, emqx_resource:restart(?ID)),
{ok, #{pid := Pid1}} = emqx_resource:query(?ID, get_state), {ok, #{pid := Pid1}} = emqx_resource:query(?ID, get_state),
?assert(is_process_alive(Pid1)). ?assert(is_process_alive(Pid1))
end,
fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
).
t_list_filter(_) -> t_list_filter(_) ->
{ok, _} = emqx_resource:create_local( {ok, _} = emqx_resource:create_local(
@ -1031,6 +1129,8 @@ t_auto_retry(_) ->
?assertEqual(ok, Res). ?assertEqual(ok, Res).
t_health_check_disconnected(_) -> t_health_check_disconnected(_) ->
?check_trace(
begin
_ = emqx_resource:create_local( _ = emqx_resource:create_local(
?ID, ?ID,
?DEFAULT_RESOURCE_GROUP, ?DEFAULT_RESOURCE_GROUP,
@ -1041,6 +1141,12 @@ t_health_check_disconnected(_) ->
?assertEqual( ?assertEqual(
{ok, disconnected}, {ok, disconnected},
emqx_resource:health_check(?ID) emqx_resource:health_check(?ID)
)
end,
fun(Trace) ->
?assertEqual([], ?of_kind("inconsistent_state", Trace)),
?assertEqual([], ?of_kind("inconsistent_cache", Trace))
end
). ).
t_unblock_only_required_buffer_workers(_) -> t_unblock_only_required_buffer_workers(_) ->