[
https://issues.apache.org/jira/browse/SPARK-16630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16432495#comment-16432495
]
Attila Zsolt Piros commented on SPARK-16630:
--------------------------------------------
Let me illustrate my problem with an example:
- the limit for blacklisted nodes is configured to 2
- we have one node blacklisted close to the yarn allocator ("host1" ->
expiryTime1), this is the new code I am working on
- scheduler requests a new executors along with blacklisted nodes (task-level):
"host2", "host3"
(org.apache.spark.deploy.yarn.YarnAllocator#requestTotalExecutorsWithPreferredLocalities)
So I have to choose 2 nodes to communicate towards YARN. My idea to pass
expiryTime2 and expiryTime3 to the YarnAllocator to choose the most relevant 2
nodes (the one which expires latter are the more relevant).
For this in the case class RequestExecutors the nodeBlacklist field type is
changed to Map[String, Long] from Set[String].
> Blacklist a node if executors won't launch on it.
> -------------------------------------------------
>
> Key: SPARK-16630
> URL: https://issues.apache.org/jira/browse/SPARK-16630
> Project: Spark
> Issue Type: Improvement
> Components: YARN
> Affects Versions: 1.6.2
> Reporter: Thomas Graves
> Priority: Major
>
> On YARN, its possible that a node is messed or misconfigured such that a
> container won't launch on it. For instance if the Spark external shuffle
> handler didn't get loaded on it , maybe its just some other hardware issue or
> hadoop configuration issue.
> It would be nice we could recognize this happening and stop trying to launch
> executors on it since that could end up causing us to hit our max number of
> executor failures and then kill the job.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
