[ 
https://issues.apache.org/jira/browse/FLINK-7876?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16220470#comment-16220470
 ] 

ASF GitHub Bot commented on FLINK-7876:
---------------------------------------

Github user zentol commented on a diff in the pull request:

    https://github.com/apache/flink/pull/4872#discussion_r147149524
  
    --- Diff: 
flink-runtime/src/main/java/org/apache/flink/runtime/metrics/MetricRegistryImpl.java
 ---
    @@ -239,7 +239,15 @@ public void shutdown() {
     
                        if (queryService != null) {
                                stopTimeout = new FiniteDuration(1L, 
TimeUnit.SECONDS);
    -                           stopFuture = 
Patterns.gracefulStop(queryService, stopTimeout);
    +
    +                           try {
    +                                   stopFuture = 
Patterns.gracefulStop(queryService, stopTimeout);
    +                           } catch (IllegalStateException ignored) {
    +                                   // this can happen if the underlying 
actor system has been stopped before shutting
    +                                   // the metric registry down
    +                                   // TODO: Pull the MetricQueryService 
actor out of the MetricRegistry
    +                                   LOG.debug("Cannot gracefully stop the 
metric query service actor.");
    --- End diff --
    
    include exception


> Register TaskManagerMetricGroup under ResourceID instead of InstanceID
> ----------------------------------------------------------------------
>
>                 Key: FLINK-7876
>                 URL: https://issues.apache.org/jira/browse/FLINK-7876
>             Project: Flink
>          Issue Type: Improvement
>          Components: Metrics
>    Affects Versions: 1.4.0
>            Reporter: Till Rohrmann
>            Assignee: Till Rohrmann
>            Priority: Minor
>              Labels: flip-6
>
> Currently, the {{TaskManager}} registers the {{TaskManagerMetricGroup}} under 
> its {{InstanceID}} and thereby binding its metrics effectively to the 
> lifetime of its registration with the {{JobManager}}. This has also 
> implications how the REST handler retrieve the TaskManager metrics, namely by 
> its {{InstanceID}}.
> I would actually propose to register the {{TaskManagerMetricGroup}} under the 
> {{TaskManager}}/{{TaskExecutor}} {{ResourceID}} which is valid over the whole 
> lifetime of the {{TaskManager}}/{{TaskExecutor}}. That way we would also be 
> able to query metrics independent of the connection status to the 
> {{JobManager}}.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to