Github user steveloughran commented on the pull request:
https://github.com/apache/spark/pull/4688#issuecomment-85992286
On AM failover, YARN refreshes the tokens for the AM so that it the NM
localizer can restart the app. when the AM collects these tokens (by the normal
mechanism), they should have the normal expiry time. Provided the AM gets these
to the containers when launched, the executors get current tokens too.
Life is more complex on work-preserving-AM-restart âwhen the containers
stay up during AM failover. They don't get the new tokens automatically, so a
mechanism like this patch is needed. the Spark AM doesn't attempt this (yet)
âit requires all state to be persisted somewhere or rebuildable, the code to
actually do the rebuilding, and a large set of tests to verify it works.
---
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]
