Deadline thinks a slave is still rendering a task even though it’s long done. I tried to set the Max Render Time at submission so it would error out, but it ignores it. I can’t pause the job or anything.
Anything you can do to help would be great.
Thanks
It looks like the slave has lost tract of the task. This could be due to the slave crashing, or to a network hiccup (generally, Deadline does a very good job of not losing tasks when the network goes down, but on rare occasions they can slip through). The slave itself keeps track of task timeouts, so if it’s lost the task, it doesn’t know that it’s passed the max time.
Deadline should periodically clean up orphaned tasks like this. If you’re running Pulse, there is a 1 in 10 chance that these files will be purged during each Repository Cleanup operation, which by default should occur every minute. So essentially, if you’re running Pulse,these files should be purged once every 10 minutes (on average). If you’re not running Pulse, then the slaves will do this randomly themselves between render tasks.
If you right-click on the tasks in the list, you can choose to requeue them or mark them as complete. This will fix this specific case. Does this problem happen a lot for you, or was this just a one off issue. If it happens a lot, there might be some underlying problem (ie: network, etc) that could be resulting in the behavior.
Cheers,
It’s happened enough for me to notice. We don’t run Pulse. I’ll let our IT guy know what’s going on and ask if anything weird is going on with the network.
You said, “then the slaves will do this randomly themselves between render tasks”
Is there a way to make the slaves do this on a regular basis?
Thanks
You can’t make the slaves do this on a regular basis, but you could
just run Pulse as an alternative. You could just run it on one of your
render nodes and not configure it in the repository options. This way,
it doesn’t act as a proxy and interfere with rendering, but will still
clean things up on a regular basis.
Cheers,
Ryan
