Github user steveloughran commented on the pull request:
https://github.com/apache/spark/pull/5423#issuecomment-139004547
I've not been pushing this up much. Current status
1. multiple attempts supported: every event -> a separate entity; these get
aggregated for the reports.
1. more efficient listing of apps/attempts means that background refreshes
put less load on ATS, so the scheduled background refresh code is
re-instated...though the on-demand option is still there and can trigger faster
updates.
1. The RM gets queried for running spark apps âthe records are used to
recognise attempts that finished without a record, e.g. AM crash. The policy is
* failed/finished/killed ==> completed
* Not in the list and the ApplicationAttemptInfo last updated over an
hour ago: assume failed. (The hour long delay is to handle transient loss of
apps from listing)
Hitting the RM for status on running apps makes a big difference; it keeps
killed apps out the incomplete list.
This is what the view looks like with some single attempt apps & double
attempts.
I plan to to go from listing attempts by yarn attempt ID to the simpler
"1", "2", "3" numbering; most of the code to do that is there, I've just not
switched over to it.
---
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]
