It is possible to do a 'Kill & Go on' as in Muster with Deadline?

Deadline
1

Hi there,

We are migrating out render farm from Muster to Deadline and one feature that we miss is the ‘Kill & Go on’ one. We used it constantly in Muster and haven’t found yet the way to replicate it in Deadline.

The way it works is like this:

We share all the renderfarm nodes between 2D and 3D, but we have some pools assigned to Nuke. If these pools are free (no Nuke render) the 3D guys can use them but within a low priority range (i.e 10-50). When a Nuke comper needs to render, he submit the job with a higher priority range (i.e 80-90) and if there is any 3D render in the same pools that he need uses the ‘Kill & Go on’ function, so the low priority 3D renders stops, the Nuke renders starts and, this is the important part, when it finishes the 3D render is resumed automatically.

I’ve been browsing through the docs but I didn’t found a way to do this. Any ideas?

Thanks!

./charlie

2

Hey!

Check out our “Job Is Interruptible” setting, should pretty much do the same thing here for you:
docs.thinkboxsoftware.com/produc … le#general

You can also control it via the option called: “Interruptible %” once enabled. Deadline 8.0 (currently in beta - contact beta@thinkboxsoftware.com if you are happy to sign NDA and test/provide feedback) will bring an additional setting here, based on “Time Threshold”, ie: if task has more than this amount of time remaining to render: X hrs, mins, secs.

Out of interest, don’t your “3D” guys get annoyed when say a, 1+ hour render gets killed, for a 20 sec Nuke render, for that slave to just go pick back up that same 3D render frame again, after it finishes with Nuke?

3

Nice. I think that ‘Job Interruptible’ will do the work. Is there anyway to make it a default?

Since the 3D guys are ‘borrowing’ the nodes it is OK to kill the job if you need it, that’s the theory.The reality: before do the ‘Kill & Go on’ you check the render times of the job and if a task takes a long time to render and is close to finish wait until is done to start your render (unless you are in a hurry).

Deadline seems to have a lot more of options to control priorities than Muster, I’m pretty sure we are gonna change a lot of the old workflows when we start using it in production.

Thanks for the help. I’ll test it and let you know how it goes.

4

ok, cool. The extra setting we have coming in Deadline 8 will help you protect against accidentally killing long running render jobs when this interruptible setting is used.

There are quite a few different ways to make this setting, like any other setting in Deadline, the default. At submission time, in the submitter code for a particular in-app submitter or monitor submitter (per application/host), as an event plugin, general manual script, etc.

I’ve made you an event plugin which I feel is the cleaner way to address these type of workflows. Advantage to using an event plugin, is you have global config/control over what jobs being submitted to which pools, will be subject to having the “JobIsInterruptible” setting applied to them. You can extend/customise the attached example as you see fit, copying it to your “/custom/events/” directory. So, to install, download, un-zip the attached zip file to /custom/events/SetJobInterruptible directory, containing the individual files as stored in the zip file. Edit the py file as you see fit.

SetJobInterruptible.zip (2.93 KB)

BTW, I will get this sample uploaded to where we have other examples which may be of interest:
github.com/ThinkboxSoftware/Dea … tom/events

Have a root around this GitHub site, lots of little gems in there :wink:

5

Awesome.

I’ll check the Github repo too.

Thanks again!