[
https://issues.apache.org/jira/browse/MESOS-2256?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14290722#comment-14290722
]
Bill Farner commented on MESOS-2256:
------------------------------------
I agree with Jay on the ideal - having best-effort communication of progress
would be _amazing_. At the very least, the scheduler needs a reliable
indication that the task has been received by the slave. This is the only
reason we have a state timeout in Aurora, to give up when we think a task has
not successfully made the scheduler -> master -> slave relay, and we believe
we're unlikely to hear about that task again.
> mesos should send scheduler status updates for docker pull progress
> -------------------------------------------------------------------
>
> Key: MESOS-2256
> URL: https://issues.apache.org/jira/browse/MESOS-2256
> Project: Mesos
> Issue Type: Bug
> Components: docker
> Reporter: Jay Buffington
>
> The scheduler doesn't get any status updates during the "docker pull". It's
> not unreasonable for a 1.5GB docker image to take 15+ min to pull over a WAN
> link. The scheduler does not have enough information to be able to set a
> reasonable timeout, since it doesn't know how big the image is.
> It also isn't able to display download progress to the end user (in the case
> of aurora/marathon) so the user has no idea how fast the download is going or
> percentage complete, so they can't make a decision if they should cancel it
> or retry.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
