You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by "Adam B (JIRA)" <ji...@apache.org> on 2014/04/07 22:59:16 UTC
[jira] [Commented] (MESOS-982) Relax slave (re-)registration
retries and add a backoff mechanism.
[ https://issues.apache.org/jira/browse/MESOS-982?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13962246#comment-13962246 ]
Adam B commented on MESOS-982:
------------------------------
Shouldn't we still be concerned about the network load on the master, especially for master failover in a large-scale cluster with thousands of slaves? Or in high-latency networks?
Admittedly, it's not a blocker for the registrar, but we might still want to add some retry/backoff logic.
> Relax slave (re-)registration retries and add a backoff mechanism.
> ------------------------------------------------------------------
>
> Key: MESOS-982
> URL: https://issues.apache.org/jira/browse/MESOS-982
> Project: Mesos
> Issue Type: Sub-task
> Components: slave
> Reporter: Benjamin Mahler
> Assignee: Vinod Kone
> Fix For: 0.19.0
>
>
> With the Registrar in place, the master must persist the registration attempt of slaves.
> Slaves will currently retry registration every 1 second, until registration succeeds. With the addition of the persistence in the master, we should relax this retry time and add a back-off mechanism to avoid placing excessive load on the master.
--
This message was sent by Atlassian JIRA
(v6.2#6252)