Github user squito commented on the issue:
https://github.com/apache/spark/pull/20640
Hi Everyone, just wanted to let you know that SPARK-16630 is progressing
now https://github.com/apache/spark/pull/21068 and after some discussion, the
implementation will actually need to live inside the cluster manager code for
yarn. But Mesos should be able to do something very similar, in fact I suspect
you could refactor a lot of that code so that its used by mesos as well -- in
principle the allocation blacklist just needs to know when container allocation
fails, and optionally the node count of the cluster. I don't totally
understand where things live in mesos and if the ApplicationMaster / Driver
distinction is present there as well.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
