You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Benjamin Mahler (JIRA)" <ji...@apache.org> on 2014/11/03 20:29:33 UTC

[jira] [Created] (MESOS-2032) Update Maintenance design to account for persistent resources.

Benjamin Mahler created MESOS-2032:
--------------------------------------

             Summary: Update Maintenance design to account for persistent resources.
                 Key: MESOS-2032
                 URL: https://issues.apache.org/jira/browse/MESOS-2032
             Project: Mesos
          Issue Type: Task
          Components: framework, master, slave
            Reporter: Benjamin Mahler


With persistent resources and dynamic reservations, frameworks need to know how long the resources will be unavailable for maintenance operations.

This is because for persistent resources, the framework needs to understand how long the persistent resource will be unavailable. For example, if there will be a 10 minute reboot for a kernel upgrade, the framework will not want to re-replicate all of it's persistent data on the machine. Rather, tolerating one unavailable replica for the maintenance window would be preferred.

I'd like to do a revisit of the design to ensure it works well for persistent resources as well.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)