Chris Burroughs created MESOS-2270:
--------------------------------------
Summary: allow cgroups_enable_cfs to be configured per task
Key: MESOS-2270
URL: https://issues.apache.org/jira/browse/MESOS-2270
Project: Mesos
Issue Type: Improvement
Reporter: Chris Burroughs
Currently cgroups_enable_cfs is a per slave setting to enforce "at most"
constraints. Without "at most" latency sensitive tasks are at risk of being
configured in such a way that they only happen to meet SLAs when there is no
contention. However, it would be desirable for batch tasks to still be able to
utilize slack cpu resources. Giving batch tasks a ceiling prevents them from
doing so, limiting utilization. It would be more flexible if this could be
controlled per task.
NOTE: This would still force cpu "at least" == "at most". It would be nice if
these were separate settings to allow a range of task "batchy-ness". This
would also be closer to the lmctfy model [1].
Some previous discussion at:
http://mail-archives.apache.org/mod_mbox/mesos-user/201406.mbox/%[email protected]%3E
[1] https://github.com/google/lmctfy/blob/master/include/lmctfy.proto
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
