Github user JoshRosen commented on the pull request:
https://github.com/apache/spark/pull/3825#issuecomment-68429462
I'd rather use fewer Akka features than more, since this will make it
easier to replace Akka with our own RPC layer in the future. Therefore, I'd
much prefer to not allow exceptions to trigger actor restarts / state clearing.
I think that adding an experimental Akka feature like persistence would be a
huge risk for little obvious gain.
I'm not sure if the "heartbeat from unknown worker" can ever occur if we
don't clear the master's state because I think that workers only begin sending
heartbeats once a master has ack'd their registration in which case the master
would know that it was a previously-registered worker and instruct it to
reconnect.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
