[MESOS-1462] External Containerizer can leave a task indefinitely in STAGING if the `launch` fails

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Won't Fix
• Affects Version/s: 0.19.0
• Labels: mesosphere
• Assignee: Till Toenshoff
• Reporter: Tom Arnfeld
• Created: 06/Jun/14 23:33
• Updated: 16/Aug/16 19:20
• Resolved: 16/Aug/16 19:20

Description

Not sure where else to create issues regarding RC software, but I guess here is as good as anywhere to record it...

I mentioned it to tillt before so i'm not sure if he thought of a fix (or has one unpushed). Essentially when you launch a task through an external containerizer, if the launch command of the external process fails for whatever reason (e.g. the fetcher throws an error) the task will sit in the STAGING state and never be terminated.

At this point, I think it's acceptable to think executor hasn't registered yet, though that's not guaranteed. I'm seeing this behaviour on 0.19.0-rc3.

Ping bmahler - I think this is worth holding up the vote to investigate.

Issue Links

• relates to: MESOS-2583 Tasks getting stuck in staging