Github user steveloughran commented on the pull request:
https://github.com/apache/spark/pull/8744#issuecomment-147749231
I've pushed out a rebase of this against master.
1. The asynchronous publishing of events with retries and, eventually
dropped events, appears to be stable, at least during unit and integration
tests.
1. scalastyle is set up also scan the yarn/history source trees (the code
is still there), and is happy.
1. I've renamed Service, SchedulerExtensionService, as I proposed earlier.
That helps distinguish it from 'YARN services', which are entirely independent.
1. I've also moved the test of extension service instantiation into the
`yarn/src/test` source tree as
`org.apache.spark.scheduler.cluster.ExtensionServiceIntegrationSuite`. With
that isolated test, I can submit the extension as a self-contained patch for
independent review. Note how the extension service takes a case class
{{SchedulerExtensionServiceBinding}} as its binding, rathe than a series of
parameters. That's so that in future, if new binding attributes were added
(currently: context, yarn app ID and yarn attempt ID), it would remain
backwards compatible with existing scheduler extensions.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
