Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/6082#issuecomment-103165158
@ehnalis Yes, that's the question. If the RM fulfills the request
instantly, this is theoretically exactly the improvement you say. If it
doesn't, it's just wasted heartbeat messages, and your change makes things
worse. Which one is realistic to expect though?
It depends on what the cluster is doing, and Thomas's test is pretty
realistic and suggests the RM doesn't react that fast. For example, it's not
clear that dumber changes like simply decreasing the overall heartbeat default
don't get the same result. The extra complexity here needs to be justified.
That said it's not that complex a change. But can you show a workload that
shows this improvement? It'd be useful to view the ball as being in your court
on that one.
---
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]
