[
https://issues.apache.org/jira/browse/MESOS-4571?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15138136#comment-15138136
]
Vinod Kone commented on MESOS-4571:
-----------------------------------
Yea, seems weird. I don't remember why it was changed anymore.
Feel free to send a patch to drop the status update between slave -> master (or
master -> scheduler) instead of between executor -> slave. Note that
FUTURE_MESSAGE() takes "from" and "to" pids are arguments that you can use here.
> SlaveRecoveryTest.RecoverStatusUpdateManager is not consistent with its
> description
> -----------------------------------------------------------------------------------
>
> Key: MESOS-4571
> URL: https://issues.apache.org/jira/browse/MESOS-4571
> Project: Mesos
> Issue Type: Improvement
> Components: tests
> Reporter: Shuai Lin
> Labels: newbie++
>
> The description is:
> {code}
> The slave is killed before the update reaches the scheduler. When the slave
> comes back up it resends the unacknowledged update"
> {code}
> But in the test code, the {{StatusUpdateMessage}} is dropped, so there is no
> checkpointing for status update involved at all. It's the executor resends
> the unacknowledged status updates when reconnecting, rather than the slave
> "resends the unacknowledged update". This means this test is the same with
> {{SlaveRecoveryTest.ReconnectExecutor}}, which doesn't make sense.
> To fix it, we should drop the {{StatusUpdateAcknowledgementMessage}} instead
> of the {{StatusUpdateMessage}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
