spark-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anirudh Ramanathan (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (SPARK-24105) Spark 2.3.0 on kubernetes
Date Thu, 26 Apr 2018 20:47:00 GMT

    [ https://issues.apache.org/jira/browse/SPARK-24105?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16454847#comment-16454847
] 

Anirudh Ramanathan commented on SPARK-24105:
--------------------------------------------

> To avoid this deadlock, its required to support node selector (in future affinity/anti-affinity)
configruation by driver & executor.

Would inter-pod anti-affinity be a better bet here for this use-case?
In the extreme case, this is a gang scheduling issue IMO, where we don't want to schedule
drivers if there are no executors that can be scheduled.
There's some work on gang scheduling ongoing in https://github.com/kubernetes/kubernetes/issues/61012
under sig-scheduling.

> Spark 2.3.0 on kubernetes
> -------------------------
>
>                 Key: SPARK-24105
>                 URL: https://issues.apache.org/jira/browse/SPARK-24105
>             Project: Spark
>          Issue Type: Improvement
>          Components: Kubernetes
>    Affects Versions: 2.3.0
>            Reporter: Lenin
>            Priority: Major
>
> Right now its only possible to define node selector configurations thruspark.kubernetes.node.selector.[labelKey].
This gets used for both driver & executor pods. Without the capability to isolate driver
& executor pods, the cluster can run into a livelock scenario, where if there are a
lot of spark submits, can cause the driver pods to fill up the cluster capacity, with no
room for executor pods to do any work.
>  
> To avoid this deadlock, its required to support node selector (in future affinity/anti-affinity)
configruation by driver & executor.
>  



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org


Mime
View raw message