EDIT
Two machines will render, its just that the other machines ignore the limit and will try to render as well.
This generates an error and eventually the job fails because of too many errors.
Normally (although now I’m not sure the limit system actually works correctly) other machines will not pick up the job if the limit has been reached.
Hey Ricardo. We did find a bug recently where an existing limit could get a negative count which I’m sure is what was causing this offset for you. The problem with the fix is that it didn’t correct the limit, so re-creating it was required. Did you use a version of 8.0 prior to service pack 10?
Yes we did, I have recreated the limits though. It still wont work.
All other limits (apart from the negative limits problem, which is present after recreating them too) seem to work fine, and just Nuke seems to ignore the limit…
Could you get me a dump of your database? It should show off what’s going on with the configuration of your farm. Please send that direct to support@thinkboxsoftware.com. You can make a dump by running this in a command prompt:
c:\DeadlineDatabase8\mongo\application\bin\mongodump --port=27080
explorer .
If you want to strip it down, I’m most interested in your limits and plugin configuration for Nuke so you can delete the “jobs” json file. I’m curious if there are problems relating to node locked licenses.
It turns out you’re hitting a known issue with a negative “InOverage” number. We’re currently working on it, so I’ll see if I can get some details. It’s great to have more data on the situation.
Limits config:
Nuke Limit (machine or slave does the same)
limit: 2
No blacklist/whitelist
Machines excluded from limit: 3 machines (these are the machines the nuke artists use, to have more renders going using the interactive licenses)
Nuke Plugin Config:
machine that can use interactive license: (the 3 machines that are used by Nuke Artists)
Well, after more investigating it looks like there is some logic we need to play with, but it would be good to get more data.
Would you be willing to re-create a machine-limit version so we can get a dump of that? Also, any other tips we could use to try a re-create here? I guessed concurrent tasks might help, but that doesn’t seem like it’s the problem here… Maybe it’s because the limit is so small?
Hey, for the moment my solution is to make my nuke jobs dependent of each other and to manually set the limit of machines but I can’t keep doing this and I really can’t see why only Deadline Nuke jobs seem to have this problem.
Altough my other limits also still frequently get negative numbers.
Update here: We have some limit checking fixed in SP12 that’s due out this week… I’m hoping it’ll help out with these Nuke limit issues, but on the surface they seem unrelated to what was done. Would you be willing to give the update a shot and report back?
I’ll also see if maybe we can put some extra debug code so we can try catching things, but if you have more than 10 machines it’s going to be a challenge to dig up all the Slave logs.
Well, you could technically do that now I suppose… The issue you’re hitting is in the book keeping for who has a limit and who doesn’t so if you wanted to just isolate the limit to a specific set of machines that should work just fine.
I want to discuss some testing ideas with the dev guys for limits… Right now it’s a very hard system to troubleshoot.