[
https://issues.apache.org/jira/browse/SPARK-10928?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14979061#comment-14979061
]
holdenk commented on SPARK-10928:
---------------------------------
So looking at the MesosSchedulerBackend it requires 1 for the mesos executor
and 1 for the task its self.
> Spark Mesos finegrain mode with single core CPUs, rendered useless
> ------------------------------------------------------------------
>
> Key: SPARK-10928
> URL: https://issues.apache.org/jira/browse/SPARK-10928
> Project: Spark
> Issue Type: Bug
> Components: Mesos
> Reporter: Ravi Sanwal
>
> If we run spark on Mesos cluster with only single core CPUs with coarse
> grained mode it runs fine. Although when we try to do the same using fine
> grained mode, it fails, because the required number of CPUs on any node is >1
> (1 for task and at least 0.1 for executor overhead).
> So there is inherently a mismatch in configuration where same set of
> environment renders one mode useless.
> Steps to reproduce.
> Create a mesos cluster with cpu resources on each node limited to 1.
> Submit a task in coarse grained mode -> runs fine.
> Submit a task in fine grained mode -> does not run.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
