AWS Thinkbox Discussion Forums

Deadline not respecting priorities

Hello,
First off, I’ll say that it’s likely this is user error somehow. Since we upgraded to Deadline 10.1, we’ve had issues with Deadline not respecting render priorities. For example, let’s say I’ve got a 3DS Max job rendering at priority 60. I’ll submit an After Effects job at priority 90, and the machines will keep on chewing on the priority 60 job, even after completing the frames they were working on before I submitted the priority 90 job. I end up having to suspend the priority 60 job to get the machines to pick up the priority 90 job. We aren’t using pools, so that shouldn’t be the issue. It’s strange, because we’ve been using Deadline for a good while now and have only been noticing this as a persistent issue for the past month or so.

The fix is probably a simple one, so please let me know! Thanks!
Andy

Is the priority “60” job using pool that is higher in the hierarchy than the job that uses priority “90”? Cause situation like that would behave like you describe.

Check what your Job Scheduling settings are.

1 Like

To add to @panze’s note, it’s worth checking if the 60 job has ‘Enforce Sequential Rendering’ enabled (details in the docs), as that will keep that job from getting interrupted.

Thinking of that, is the 60 job Interruptible? If it’s not, it won’t get bumped out of the way by higher priority jobs. (Details on that here).

That last thing to check would be if the machine you’d expect to pick up the 90 job is even allowed to grab it, but since suspending the 60 seems to work, we don’t need to worry about that. (If we were worried, a great first step would be using the Find Render Candidates tool to see what’s what.)

Strangely, I’ve checked all of those possibilities and none of them seem to be applicable.

As mentioned above, we don’t use pools, and everyone’s pool is set to ‘none’.

Our job scheduling order is ‘Priority, First-in-first-out’.

None of the jobs are set to interruptible. My understanding is I’d only want to set the job to interruptible if I wanted the renders to release and move to the higher priority job mid-render. Correct me if I’m wrong on that.

The machines I’m trying to move to the higher priority job are render candidates for said job.

One thing we are doing that might be weird is that we aren’t using a group for 3DS max jobs, since all machines on our farm have Max. So we have an AE group, a Maya group, and various other groups that contain better or lesser machines, but when submitting a Max job, we’ve always just had the group set to ‘none’.

And just to reiterate, this was never a problem a month or so ago. Very weird.

One other pattern we’ve noticed is that this isn’t a problem with jobs that are submitted by the same person. Is it possible for some users to be weighted more than others?

I figured I’d mention that this seemed to work itself out after a few weeks of posting this thread. However, I just got running with a new computer and am experiencing this behavior again.

That’s really really odd. There’s nothing user-specific that could be contributing to priority. Given what we’ve got in here I still can’t figure what’s causing the inconsistency.

Is there anything else that’s changed? There must be something we’re not thinking of.

I’ve finally isolated it to my older jobs. Newer projects I’ve been working on don’t show this behavior. Projects I’m working on that have returned after being dormant for a while do show this behavior.

Privacy | Site terms | Cookie preferences