Nuke Submitter Crashing

Deadline
42

Haven’t seen this, but sounds like a bug.
The intermittent occurrence suggests it’s related to threading again, although I couldn’t imagine why this particular issue is happening.

In such situations it might be best to break in the debugger and step over the code. Make it so that it breaks only when the bug occurs, e.g.

if len(nodes) > 1:
   remote_pdb.set_trace()
43

The crashes appear fully random. Presumably they are based on concurrent access to some shared resource from multiple threads. By the way the nuke API is not thread safe, so any calls to it should be serialized or made from the main thread.

There is one small piece installed in the user’s $HOME/.nuke directory, but the main part of the submitter is reloaded from the farm each time.

44

Thanks Mois!
We did notice around line 2432 there’s a list declaration that looks like it has a casing that is different from what’s being used.
tempwriteNodes = [] # type: List[Any]
The var inside the subsequent code is using
tempWriteNodes
We’ve updated this and so far have not hit previously mentioned error. I can’t confirm that this is the cause of the bug(because I was never able to replicate it on my workstation ;)), but so far so good.
Thanks,
-Jake

45

I’ll see about getting that typo fixed, but I don’t think it’s the root of your issue.

That typo’s been there since Deadline 10.1, and call to sorted happens in every branch of the following if/else ladder before it gets evaluated. So I figure that’s what’s keeping this from breaking for everyone else in the past.

What version of Deadline are you running? Line 2432 in my 10.3.2.1. checkout isn’t the line with the typo.

We did fix an issue with “Selected Nodes Only” and “Submit Write Nodes As Separate Jobs” before, but in that case no job would be created.

46

Huh, a static analyzer should have caught that. Justin do you guys use lsp or something?

47

Hi Justin,
I was referring to the line numbers from this post:

We are using 10.3.1.4, but I’ve forked ours with changes from the above link and a bunch of internal customizations.

Thanks,
-Jake

48

Oh! I’m not convinced that’d fix it given that typo’s been around for so long, but results are results.

49

I’m not certain, but I’ll be finding out what’s going wrong or unnoticed in our linting.

1 Like