PipelineFX Forum

Qube! => General => Topic started by: freelanceit on October 29, 2015, 11:44:28 AM

Title: The job that wouldn't die!
Post by: freelanceit on October 29, 2015, 11:44:28 AM

Hi all,

We're running 6.4.0.b and I've got a job that's sitting rendering on one of our Linux farm nodes and I can't get rid of it.
Some of the instances failed, some have been killed but some remain running. In the job properties it shows as Status: Dying.
I've tried killing the job (it ignores this), I've stopped and restarted the worker, I've rebooted the box and strangely I can't see any process running on the box that might be the job?

Any ideas?

Adam.

Title: Re: The job that wouldn't die!
Post by: jburk on December 22, 2015, 07:59:15 PM

There were some bugs in the early days of the 'dying' state that have been fixed since.  The dying state is intended to be "I've asked for this job to be killed, but the workers haven't all reported that they've successfully shut down the instances".

I'd suggest that you "remove" the job from Qube, and restart the worker service on those hosts that still think they're running it.   If that doesn't solve it, are you able to open a support ticket (still on valid support & maintenance)?  Qube 6.6. added a "data integrity check" menu to WranglerView, but since you're running an older version you may have to resort to SQL to get rid of this entirely.