AWS Thinkbox Discussion Forums

Offload DR Corona + Server Mode

(Updated for clarity)

I’ve built a rather hacky maxscript that can run offload Corona DR jobs on our farm, but I’m running into an issue with the “server” version of Max.
The script submits a maxscript job to the farm which uses dialogMonitorOps and uiAccessor to simulate button presses to initiate a DR job. Running the script with workstation mode turned-on works perfectly well, multiple computers pick up, the file renders fine, we can use bloom/glare/etc, but the downside is doing it this way requires the use of a Max license by the initial computer on the farm. This isn’t a big problem for one of our studios since we have some free max licenses, but our other studio is max’d out.

Running without workstation mode still allows the script to reserve servers, update the server list, I can see the job starting on other nodes on the farm, but the second the render is started it completes instantly.

I know this a pretty specific question about an edge use-case, but with the “server” mode documentation being pretty scant I’m wondering if you guys have any insight, or any idea of when you will implement offload Corona DR yourselves?

I’ve updated my original post, I realize it may have been unclear before.

Hi Steve,

Unfortunately I can’t answer your question but I’ve been contemplating the idea to write such a script for a couple of months now.
Would you mind sharing it with the community as I’m sure that I am not the only one looking for something like that.

Team Thinkbox, I’m sure that lots of users are looking for Corona DR, I second Steve in his request!

Thanks!

Chris

I can’t say when it’ll happen but we’ve been tracking it ourselves here. I’ve asked if the dev team have any insight here and I’d like to gauge how many users are in need of this one so I can try bumping the priority. Can you guys find a few more users who can chime in on how they’d like to make use of the feature?

Make sure you checkout Deadline 10 SP14 when it drops…hint hint… :wink:

That sounds great! When will it be released?
Chris

Really soon? :smiley:

Assuming it passes testing it should hit this week, but no guarantees as usual.

deadline.thinkboxsoftware.com/d … il-11-2018

So far everything is working well with Offload DR Corona.

Only thing I have noticed is that there is no Dynamic Start.
If all the servers aren’t present that are set during render submission it just waits.

Is this something that can be implemented like it was for vray?
Can the off load server count be changed after submission?

Thx.

1 Like

We definitely try to include dynamic joining for off-load jobs, but in this case there’s a technical limitation between Corona’s DR and our current implementation of the 3dsmaxLightning executable that prevents us from supporting it.

Cheers

Is there any work around for this?
Can you change the server amount after submission?

Not without killing the already running nodes unfortunately. There is an API call to add extra tasks to the end of a job without re-creating all tasks, but that’s not exposed in the UI as far as I know.

Is there a way to render one job at a time. Or to force all the slaves to fill up one job before searching for the second.
The issue I’m having is that if I set off all the rendering jobs in the queue at once the render nodes distribute over all of the render jobs, with out filling up all the require nodes for one. And when coronaDBR offload is active is just waits.
I have tried Pooling and Grouping machines which out any luck. Also changed the Job Scheduling to various option without any luck.
I may be missing something but not sure what.

The only way I got around this was by creating dependencies on the each job.

Actually, the issue you were just describing with the nodes distributing themselves is not the default behavior. When you look in the scheduling order configuration, that mode is called ‘balanced’ and normally we run it as first-in first-out.

Here’s the default as it is on the farm:

[attachment=0]2018-05-01 10_13_42-Configure Repository Options.png[/attachment]

I think if you were able to do it for just one job by using dependencies, that can be rather nice. The reason we implemented ‘balanced’ queuing was so that folks could get their test renders back and decide if they wanted to keep it rendering.

Is there any development on the Dynamic joining?

1 Like
Privacy | Site terms | Cookie preferences