We’ve been pushing out a lot of fusion comps recently and I’ve noticed that occasionally a task will still be running a long time after it should have completed (based on the average render time for the other tasks)…
Say the average is a minute, these “stuck” tasks are running 10’s minutes to well over an hour (we have a high timeout)…
When I VNC into the machine, the log says that the render has completed, but deadline does not seem to have noticed…
This usually happens on 1 or 2 tasks per job and if we mark the task as failed and resubmit it, it normally goes through fine (and normally on the same machine)??
We’ve recently run into a similar issue with Deadline 3.0 here, and the problem was occurring when Deadline was trying to shutdown Fusion gracefully to finish the render. The current workaround we’re implementing is just to kill Fusion. When this problem occurs for you, is Fusion still running?
Another thing that we stumbled across with this problem was that there was always 2 eyeonscript processes running in the Task Manager when this happened. Killing one of them would allow the other to continue (depending on which one we killed, this may or may not result in a Deadline error).
I wonder if we should be killing any rogue eyeonscript processes before launching new ones (currently, we only do this when initially starting up the job, but not between tasks). Although this may be overkill, because since applying the workaround on our farm that I mentioned earlier, we haven’t seen this hanging issue.