[
https://issues.apache.org/jira/browse/MESOS-786?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13809756#comment-13809756
]
Vinod Kone commented on MESOS-786:
----------------------------------
The *updated* semantics are supposed to guarantee that both 2) and 3) will
result in reregistered() call back. And I think that follows your state
diagram? Did I understand your question correctly?
> Update semantics of when framework registered()/reregistered() get called
> -------------------------------------------------------------------------
>
> Key: MESOS-786
> URL: https://issues.apache.org/jira/browse/MESOS-786
> Project: Mesos
> Issue Type: Bug
> Reporter: Vinod Kone
> Fix For: 0.16.0
>
>
> Current semantics:
> 1) Framework connects w/ master very first time --> registered()
> 2) Framework reconnects w/ same master after a zk blip --> reregistered()
> 3) Framework reconnects w/ failed over master --> registered()
> 4) Failed over framework connects w/ same master --> registered()
> 5) Failed over framework connects w/ failed over master --> registered()
> Updated semantics:
> Everything same except
> 3) Framework reconnects w/ failed over master --> reregistered()
--
This message was sent by Atlassian JIRA
(v6.1#6144)
