I started testing the VRED submission script. It didn’t really work like I had hoped but I know this is down to the fact that the VRED python api is extremely limited. (I have told them this several times and told them again this evening)
We typically use the render queue and have a list of things in there. We kind of see/treat them like maya render layers so the hope is that deadline will be able to read each render job in the render queue list and submit each one to an individual render node. Like the deadline submission scripts can do with maya and ‘Submit Render Layers As Separate Jobs’.
Right now it will just run all jobs listed on the render queue which is fine. This is still ok as we can at least submit multiple scene files to the farm.
I have not really tested the other parts of the submitter as we tend to just use the render queue. But I will try and have a play and see how it goes.
It would also be nice to have a submission script within Vred but I know that is also a vred limitation.
I would love to be able to help you with the ability to pull jobs out of the render queue and run them separately… however we run into 2 main issues, from the monitor we have no idea what is in the render queue and their api only contains 1 function to do with their render queue… run all render jobs.
For integrated submitter, it would currently be physically possibly to create an integrated submitter, however since we can pull effectively nothing from the scene (not even the scene name) it makes no sense to create on. If they expose more stuff to their python api then I will make sure to add it.
Let us know how it turns out for you, also have you tried the cluster yet.
Yeah totally understand and I have endorsed those points to the Vred autodesk people. However the problem is we are not one of their big customers (ie an automotive company) so our demands are lower down the list. We are just the ones that make pretty images using it. However they are becoming a bit more receptive though so fingers crossed. It will help if a company like yours can also push them on this! It is really unacceptable for a piece of software that costs as much as it does to have this shit of an api (pardon my french)
I have not tried the clustering yet. I am not totally sure how that works? I know how to use clustering within vred but not sure how the Deadline Vred Cluster process works. Could you explain?
I have requested some stuff from them, and we are looking at pushing some more.
When using Deadline’s Vred Cluster process will effectively reserve machines and start up the vred cluster process on the slaves (only works if they do not have the cluster service running currently). Doing this means you do not have to remove the slaves from your farm, and it will pick up however many machines you need. You then use it as per normal on your local machine. If you have used the VrayDBR jobs it works similar to those.
Let me know if that made sense or if you need more information
Grant
Right but you still have to set the vred file to use cluster in the scene file before you submit a scene?
I am sort of familiar with the VrayDBR. Our guys in London use it so I remember them showing it to me but we have not used it here as we do not use much vray here. Mainly VRED.
If you could just step me through the workflow that would be helpful.
Sure I can step through the workflow.
When you want to render an image locally you send a VREDCluster Job with a Cluster Count of however many slaves you want to render the job. You then wait for however many slaves as you want to pick it up when those slaves are “rendering”. After this in Vred scene set the Hostnames to the machine names of the slaves that picked it up and begin your render locally.
The main advantage this gives you over cluster rendering normally is that you do not have to manually remove machines from your farm.
So the workflow is…have scene in vred, open deadline, submit a Vred Cluster job with the Cluster Count set to the number of machines you want to render with. Go back to vred, and then enable cluster, entering the names of the machines the Vred Cluster job started on.
Will try it out.
When a render finishes, will the vred cluster job automatically terminate/complete as well? Or do you have to manually suspend the cluster job in deadline to makes the nodes available for other deadline jobs?
I am testing submitting a vred scene as an Animation but it seems not to take into account the camera name I enter or the animation clip.
It always just renders the perspective camera.
Any ideas? I can send you the test scene I have to try.
I do not plan on rendering this way as we would rather use the render queue but with vred the render queue submission is very limited right now so just seeing what else is possible.
Here is an updated VRED Plugin that should fix at least the camera name issue (Vred’s python api ignores the selected view when rendering apparently). It looks like something similar is going wrong with the clip, however I am not sure what it is at the moment.
Hmm no that does not seem to have fixed the camera issue. Maybe because it is a separate camera I imported and not an actual viewpoint. I will do a test.
Ah yeah, it looks like I am covering only the case of using a viewport not a separate camera, so I will make it so it will check if a viewport is specified it will use that viewport and the perspective cam, and if it is not a viewport it will try using the specified camera. Which gives me no info on if it passed or failed so we will hope. I should have a new one shortly that you can test.
Clips to be honest I do not know. I never use Clips. I thought it was mandatory for the submission but obviously not which is good.
Would it be possible to add a -no gui option so when you render it does not load up the vred application gui everytime. I believe there is a no gui setting for vred.
No gui was in already and I had Commented it out for testing purposes. if you just uncomment line 93 in the latest version (arguments += "-nogui ") then that should be working again.