Error hunting tools

Deadline
1

I propose deadline come with some error hunting tools.

Here is my off-the-top of my head list.

  • Search for errors by keyword.
  • Analysis based on similarities of errors. Grouping job, machine, or user errors together.
  • Marking a job or slave to capture screenshots, process list, stacks etc. I want to see a screenshot with my errors if they keep coming up.
  • Mark errors as read. Or mark in time when you last looked at errors. I don’t want to look at old errors, but i also don’t want to delete them.
2

Great ideas!

The first 2 should be possible in the next major Deadline release. Whether or not we have time to implement them will depend on a lot of things, but the system itself should make doing this pretty straightforward.

Additional information like process lists should be possible too. Screen grabs is definitely an interesting idea. Would result in more storage space taken up by the reports though…

Marking errors as read could be tricky. It’s not like an mail client where you’re the only one reading your email. Reports can be viewed by everyone, and you would probably want the “mark as read” to be on a per user basis.

Cheers,

  • Ryan
3

Marking errors read would definitely be a per user basis. And I would couple that with the feature that groups them together.

And yes, screengrabs would be more more disk space. I would want that to only happen when you are “hunting.” So you would have to say, "hmmm, I don’t know what the text errors mean, let me mark this as, ‘get more info.’ Then that job or slave will collect more info during the render, like stacks, screengrabs, performence graphs at a certain polling frequency. Once solved, it stops collecting the info.

B.