[
https://issues.apache.org/jira/browse/BEAM-3310?focusedWorklogId=115423&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-115423
]
ASF GitHub Bot logged work on BEAM-3310:
----------------------------------------
Author: ASF GitHub Bot
Created on: 25/Jun/18 13:54
Start Date: 25/Jun/18 13:54
Worklog Time Spent: 10m
Work Description: echauchot commented on a change in pull request #4548:
[BEAM-3310] Metrics pusher
URL: https://github.com/apache/beam/pull/4548#discussion_r196330081
##########
File path:
runners/flink/src/main/java/org/apache/beam/runners/flink/FlinkRunner.java
##########
@@ -120,7 +121,9 @@ public PipelineResult run(Pipeline pipeline) {
if (result instanceof DetachedEnvironment.DetachedJobExecutionResult) {
LOG.info("Pipeline submitted in Detached mode");
- return new FlinkDetachedRunnerResult();
+ FlinkDetachedRunnerResult flinkDetachedRunnerResult = new
FlinkDetachedRunnerResult();
+ // no metricsPusher because metrics are not supported in detached mode
Review comment:
@zorro786 in Flink there is no driver like there is in Spark. So user main()
code (pipeline.run and thus MetricsPusher thread) will run inside the Flink
client. In default submission mode, the client lives until the end of pipeline
so no issue. In detached mode, the client lives only the time of submission of
the job so, as the Metrics thread is a daemon, it would be leaking after client
is stopped. If we set it not to be a daemon, then it will stop with the client
and no metrics will be pushed. So the current architecture of the MetricsPusher
will not support detached mode in Flink. One of the solutions is to rewrite it
all using the DAG modification but IMHO I find it overkill to change it all
just to support one sub-case on one only runner and more over only metrics
pushing and not pulling will be supported. Is there a strong need for
supporting metrics export in detached mode even if regular metrics pulling
(PipelineResult.metrics()) are not supported?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 115423)
Time Spent: 14h (was: 13h 50m)
> Push metrics to a backend in an runner agnostic way
> ---------------------------------------------------
>
> Key: BEAM-3310
> URL: https://issues.apache.org/jira/browse/BEAM-3310
> Project: Beam
> Issue Type: New Feature
> Components: runner-extensions-metrics, sdk-java-core
> Reporter: Etienne Chauchot
> Assignee: Etienne Chauchot
> Priority: Major
> Time Spent: 14h
> Remaining Estimate: 0h
>
> The idea is to avoid relying on the runners to provide access to the metrics
> (either at the end of the pipeline or while it runs) because they don't have
> all the same capabilities towards metrics (e.g. spark runner configures sinks
> like csv, graphite or in memory sinks using the spark engine conf). The
> target is to push the metrics in the common runner code so that no matter the
> chosen runner, a user can get his metrics out of beam.
> Here is the link to the discussion thread on the dev ML:
> https://lists.apache.org/thread.html/01a80d62f2df6b84bfa41f05e15fda900178f882877c294fed8be91e@%3Cdev.beam.apache.org%3E
> And the design doc:
> https://s.apache.org/runner_independent_metrics_extraction
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
