[
https://issues.apache.org/jira/browse/FLINK-11603?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16774180#comment-16774180
]
Till Rohrmann commented on FLINK-11603:
---------------------------------------
Yes [~Tison] we do need a separate {{RpcService}} if we want to keep the
metric's system independent of the rest.
> Port the MetricQueryService to the new RpcEndpoint
> --------------------------------------------------
>
> Key: FLINK-11603
> URL: https://issues.apache.org/jira/browse/FLINK-11603
> Project: Flink
> Issue Type: Improvement
> Components: Metrics
> Reporter: TisonKun
> Assignee: TisonKun
> Priority: Major
>
> Given that a series TODO mention {{This is a temporary hack until we have
> ported the MetricQueryService to the new RpcEndpoint}}, I'd like to give it a
> try to implement the RpcEndpoint version of MetricQueryService.
> Basically, port {{onRecieve}} to
> 1. {{addMetric(metricName, metric, group)}}
> 2. {{removeMetric(metric)}}
> 3. {{createDump()}}
> And then adjust tests and replace {{metricServiceQueryPath}} with a
> corresponding {{RpcGateway}}.
> I'd like to learn that if the statement if true --- when we call a
> Runnable/Callable with runAsync/callAsync, then the Runnable/Callable is
> running in the main thread of the underlying RPC service, specifically, in
> the actor thread?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
