[
https://issues.apache.org/jira/browse/MESOS-6295?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Joseph Wu updated MESOS-6295:
-----------------------------
Shepherd: Joseph Wu
Sounds good. But instead of adding logging around it, it would be better to
change the log level to {{VLOG(1)}}.
There is already logging for any QoS corrections the agent performs or fails to
perform. The count of QoS corrections might only be useful when
testing/debugging your module.
> Excessive logging on agent when oversubscription modules are attached
> ----------------------------------------------------------------------
>
> Key: MESOS-6295
> URL: https://issues.apache.org/jira/browse/MESOS-6295
> Project: Mesos
> Issue Type: Bug
> Components: oversubscription
> Reporter: Kunal Thakar
>
> The mesos agent is logging a INFO log in the agent that prints out the number
> of QoS corrections received from the module [1]. By default, this method gets
> called in a tight loop, which results in the logs getting flooded with noise.
> {code}
> mesos-slave[14591]: I0930 17:31:19.278774 14598 slave.cpp:5049] Received 0
> QoS corrections
> mesos-slave[14591]: I0930 17:31:19.280088 14602 slave.cpp:5049] Received 0
> QoS corrections
> mesos-slave[14591]: I0930 17:31:19.281497 14599 slave.cpp:5049] Received 0
> QoS corrections
> mesos-slave[14591]: I0930 17:31:19.282856 14614 slave.cpp:5049] Received 0
> QoS corrections
> {code}
> Even after reducing the QoS checking frequency, this line does not make for a
> good INFO log as it is periodic. Can we change it to only log when there are
> non zero number of corrections received from the module? I can send a patch
> if this is acceptable.
> [1] https://github.com/apache/mesos/blob/master/src/slave/slave.cpp#L5523
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
