[
https://issues.apache.org/jira/browse/AURORA-1059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14290723#comment-14290723
]
Bill Farner commented on AURORA-1059:
-------------------------------------
You're right, MESOS-2256 will be the inhibitor here, especially since the slave
is 100% responsible for the fetch in this case.
> Instances stay in "ASSIGNED" while docker pull completes
> --------------------------------------------------------
>
> Key: AURORA-1059
> URL: https://issues.apache.org/jira/browse/AURORA-1059
> Project: Aurora
> Issue Type: Story
> Reporter: Jay Buffington
>
> We need a new state between ASSIGNED and RUNNING, such as DOWNLOADING. This
> would only be used for tasks that have ContainerInfo set and mesos slave
> downloads the container before launching the task
> Docker images can be large (~2GB is not unheard of) and can take 15-20
> minutes to download over the WAN. Aurora needs to take the download progress
> into account rather than timing out due to transient_task_state_timeout or
> UpdateConfig's restart_threshold.
> This likely depends on MESOS-2256
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
