You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Susan X. Huynh (JIRA)" <ji...@apache.org> on 2017/07/18 18:07:00 UTC
[jira] [Updated] (SPARK-21458) Tear down the framework when
failover_timeout > 0 (Mesos cluster mode)
[ https://issues.apache.org/jira/browse/SPARK-21458?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Susan X. Huynh updated SPARK-21458:
-----------------------------------
Description:
When the driver failover_timeout was always set to zero, we relied on the Mesos master to detect the disconnected driver and tear down the framework. When failover_timeout is nonzero, we have to make sure that the driver framework is torn down in all cases. Some cases require an explicit teardown are:
# When a driver job submission is killed by the user
# In --supervise mode, when a driver fails
Note: the driver and executors do stop running. The only issue is the the framework shows up as "Inactive" rather than "Completed" without the teardown.
was:
When the driver failover_timeout was always set to zero, we relied on the Mesos master to detect the disconnected driver and tear down the framework. When failover_timeout is nonzero, we have to make sure that the driver framework is torn down in all cases. Some cases require an explicit teardown are:
# When a driver job submission is killed by the user
# In --supervise mode, when a driver fails
> Tear down the framework when failover_timeout > 0 (Mesos cluster mode)
> ----------------------------------------------------------------------
>
> Key: SPARK-21458
> URL: https://issues.apache.org/jira/browse/SPARK-21458
> Project: Spark
> Issue Type: Sub-task
> Components: Mesos
> Affects Versions: 2.2.0
> Reporter: Susan X. Huynh
>
> When the driver failover_timeout was always set to zero, we relied on the Mesos master to detect the disconnected driver and tear down the framework. When failover_timeout is nonzero, we have to make sure that the driver framework is torn down in all cases. Some cases require an explicit teardown are:
> # When a driver job submission is killed by the user
> # In --supervise mode, when a driver fails
> Note: the driver and executors do stop running. The only issue is the the framework shows up as "Inactive" rather than "Completed" without the teardown.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org