All tasks of job marked failed if only one actually failed

Deadline
1

Hi all,

I just wrote a custom Deadline plugin, based off of the stock PythonSubmission plugin for Deadline (but modified to pass along task start and end frames for handling frame chunks within tasks of the same job, instead of using a separate job per frame chunk).

I’ve got things working pretty well … my submission will submit only one job, but with multiple tasks if frame chunks are specified … and I can resubmit and change the frame list range and chunk size, and the task renders pick up the new correct frames to render. So that’s all good.

But, I’ve noticed behaviour where all tasks for the job are marked as failed, even if they actually completed (and returned 0), if one of them ended up failing (i.e. returned non-zero).

The desired behaviour I need to get working is to have any actually failed tasks to be marked as failed, have any actually completed tasks to be marked as complete, and then to have the job marked as failed if any one of its tasks failed … otherwise marking the job as complete if all tasks completed.

Anyone know of a way for me to accomplish this? Some task settings or configuration that I can specify for my plugin? Or some job info setting?

Any help on this would be greatly appreciated!

Many thanks,
Mike

2

Hello Mike,

Could I have you send over your custom script so I can have some of our coders look it over? Thanks.

3

Hi Dwight,

Okay, I’ll go ahead and create a support ticket to send you code.

Will aim to get the ticket created and information added to it by end of day today.

Many thanks,
Mike

4

Sorry, I forgot to post the solution from the support ticket I had opened on this.

It turned out the issue was just with the Failure settings in the custom job info file for our custom submission. We previously were just working with jobs and not tasks, and only had these settings in our job info file:

OverrideJobFailureDetection=True
FailureDetectionJobErrors=3

… but with the desire to let tasks error individually, without necessarily failing the job, and all the other tasks, right away … we found out (through more trial and error) that we needed these settings in the job info file:

OverrideJobFailureDetection=True
FailureDetectionJobErrors=0
OverrideTaskFailureDetection=True
FailureDetectionTaskErrors=1

Hopefully this helps with anyone else out there working with task chunks.

Cheers!