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

feat(emqx_resource): include error with alarm for resource_down

This commit is contained in:
Stefan Strigler 2023-03-15 14:52:03 +01:00
parent 53825b9aba
commit c1384b6e6e
2 changed files with 14 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
apps/emqx_resource/src/emqx_resource_manager.erl
View File

@ -522,7 +522,7 @@ start_resource(Data, From) ->
id => Data#data.id, id => Data#data.id,
reason => Reason reason => Reason
}), }),
_ = maybe_alarm(disconnected, Data#data.id), _ = maybe_alarm(disconnected, Data#data.id, Data#data.error),
%% 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{status = disconnected, error = Reason}, UpdatedData = Data#data{status = disconnected, error = Reason},
@ -597,7 +597,7 @@ with_health_check(Data, Func) ->
ResId = Data#data.id, ResId = Data#data.id,
HCRes = emqx_resource:call_health_check(Data#data.manager_id, Data#data.mod, Data#data.state), HCRes = emqx_resource:call_health_check(Data#data.manager_id, Data#data.mod, Data#data.state),
{Status, NewState, Err} = parse_health_check_result(HCRes, Data), {Status, NewState, Err} = parse_health_check_result(HCRes, Data),
_ = maybe_alarm(Status, ResId), _ = maybe_alarm(Status, ResId, Err),
ok = maybe_resume_resource_workers(ResId, Status), ok = maybe_resume_resource_workers(ResId, Status),
UpdatedData = Data#data{ UpdatedData = Data#data{
state = NewState, status = Status, error = Err state = NewState, status = Status, error = Err
@ -616,15 +616,20 @@ update_state(Data, _DataWas) ->
health_check_interval(Opts) -> health_check_interval(Opts) ->
maps:get(health_check_interval, Opts, ?HEALTHCHECK_INTERVAL). maps:get(health_check_interval, Opts, ?HEALTHCHECK_INTERVAL).
maybe_alarm(connected, _ResId) -> maybe_alarm(connected, _ResId, _Error) ->
ok; ok;
maybe_alarm(_Status, <<?TEST_ID_PREFIX, _/binary>>) -> maybe_alarm(_Status, <<?TEST_ID_PREFIX, _/binary>>, _Error) ->
ok; ok;
maybe_alarm(_Status, ResId) -> maybe_alarm(_Status, ResId, Error) ->
HrError =
case Error of
undefined -> <<"Unknown reason">>;
_Else -> emqx_misc:readable_error_msg(Error)
end,
emqx_alarm:activate( emqx_alarm:activate(
ResId, ResId,
#{resource_id => ResId, reason => resource_down}, #{resource_id => ResId, reason => resource_down},
<<"resource down: ", ResId/binary>> <<"resource down: ", HrError/binary>>
). ).
maybe_resume_resource_workers(ResId, connected) -> maybe_resume_resource_workers(ResId, connected) ->

4
changes/ce/fix-10145.en.md
View File

@ -1 +1,3 @@
Fix `bridges` API to report error conditions for a failing bridge as `status_reason`. Fix `bridges` API to report error conditions for a failing bridge as
`status_reason`. Also when creating an alarm for a failing resource we include
this error condition with the alarm's message.