You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Andrei Sekretenko (Jira)" <ji...@apache.org> on 2020/11/03 19:14:00 UTC
[jira] [Comment Edited] (MESOS-10161) Constraints-based offer
filtering
[ https://issues.apache.org/jira/browse/MESOS-10161?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17225617#comment-17225617 ]
Andrei Sekretenko edited comment on MESOS-10161 at 11/3/20, 7:13 PM:
---------------------------------------------------------------------
Marathon now is capable of using offer constraints as an experimental feature: https://jira.d2iq.com/browse/MARATHON-8764
I've moved related tasks that are not going to land into 1.11.0 but might be needed for the offer constraints to graduate from experimental status into stable into https://issues.apache.org/jira/browse/MESOS-10195 .
was (Author: asekretenko):
Marathon now is capable of using offer constraints as an experimental feature: https://jira.d2iq.com/browse/MARATHON-8764
I've moved related tasks that are not going to land into 1.11.0 but might be needed for the offer constraints to graduate from experimental status into stable into https://issues.apache.org/jira/browse/MESOS-10195,
> Constraints-based offer filtering
> ---------------------------------
>
> Key: MESOS-10161
> URL: https://issues.apache.org/jira/browse/MESOS-10161
> Project: Mesos
> Issue Type: Epic
> Components: allocation
> Reporter: Andrei Sekretenko
> Assignee: Andrei Sekretenko
> Priority: Major
>
> This epic tracks work for implementing offer constraints for the scheduler API, i.e. an interface for frameworks to specify constraints on agent attributes in offers.
> By specifying offer constraints, a "picky" framework can greatly reduce a time needed to get a suitable offer, especially in the case when the corresponding role is close to be out of quota.
> One example which receives a huge benefit from offer constraints is running a "daemon set": when a framework needs to launch a task on every node of a large cluster, but the quota is just enough for all these tasks.
> Design doc: https://docs.google.com/document/d/1MV048BwjLSoa8sn_5hs4kIH4YJMf6-Gsqbij3YuT1No/edit#heading=h.wq9atl6k4yq0
--
This message was sent by Atlassian Jira
(v8.3.4#803005)