Github user steveloughran commented on the pull request:
https://github.com/apache/spark/pull/6935#issuecomment-158971646
As discussed on #6545, I think we could actually switch to live UI
playback, provided the number of cached UIs was kept down.
What must be done here is keeping a limit on the number of threads doing
playback; a pool will be needed equal to the size of the loaded UIs, or, if the
provider starts them, the number of cached UIs must be limited *and* when
removed from the cache, the thread has to be returned to the pool.
This could not only handle incomplete applications, it should be able to
offer a more responsive load of completed ones -the UI could be registered
while event playback was in progress.
Proposal
1. Supplement `getAppUI(appId: String, attemptId: Option[String]):
Option[SparkUI]` with one to start Spark UI playback; some case class returned
with the UI, playback info (in progress), and a closure to call when releasing
the cached entry
1. when loading a complete attempt, the async playback would be executed
until the playback finished, then the close() operation called.
1. when loading an in progress app attempt, the async playback would
continue until one of: app completes, app release from cache, history server
halted (for in-VM tests)
---
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]
