Swarm with VRAY Next issue

[jmueller]

Since upgrading to VRAY Next I've not been able to use SWARM for rendering. All of the hosts show with status "ENABLED" and never change to "Assigned" or other status. Rendering works fine with SWARM disabled, but when enabled the render stops after building the light cache and never does the actual rendering. I've tried using a tag to send only to machines that also have VRAY Next installed.

[jskt]

I've had that issue as well. I wish I could say what made it stop, but it just sort of decided to.

Our setup is a bit odd (2 different subnets, as well as starting/stopping the Swarm service with Deadline), and I suspect that had something to do with it.

I also very definitely needed to use assigned Master nodes rather than allow the nodes to decide Ad-hoc, which has apretty counterintuitive setup procedure (Master nodes themselves need to auto-detect). Whenever a machine talks to the swarm with incorrect master settings it tries to take over, the swarm gets fragmented, and things don't settle back to normal for a while.

Another thing that very definitely needed to be done by hand was upload the VRay package (C:\Program Files\Chaos Group\V-Ray\V-Ray for Revit\Libraries\vrayappsdk\appsdk-win[restOfLongFileName].7z) to all the swarm nodes through the web interface. For some reason the Revit version doesn't seem to handle this task at all.

[Ana Lyubenova]

Uploading the V-Ray package to the Swarm nodes is automatic.
jmueller , can you please reproduce the problem and send me the latest V-Ray for Revit .log file from C:\Users\your username\AppData\Roaming\Chaos Group\V-Ray for Revit 20XX\Logs

[jskt]

Ana,

Uploading the V-Ray package to the Swarm nodes is not automatic here, at least with V-Ray Next for Revit (Rhino works fine, and VRfR 3.x worked fine). How would we go about troubleshooting this? Is it possible that this is related to our configuration that uses permanent Master nodes that never participate in renders?

[ivaylocg]

Hello,

From what I understand you have to manually upload the V-Ray build to the Swarm node, otherwise the rendering does not start? Is there any specific error message that appears in the render log in that case? Also, do you see anything being logged in the Swarm node log? It would be best to reproduce the problem over a remote session, so someone from our Support team can take a better look.

[jskt]

Hello,

From what I understand you have to manually upload the V-Ray build to the Swarm node, otherwise the rendering does not start? Is there any specific error message that appears in the render log in that case? Also, do you see anything being logged in the Swarm node log? It would be best to reproduce the problem over a remote session, so someone from our Support team can take a better look.

We've done a remote session, which sort of half-fixed the problem. I have everything fairly stable now, and don't want to break anything. From what I recall it looked like licensing errors (the vray standalone would start up but never download assets or begin rendering), but we had plenty in the pool and really extensive firewall exceptions. Logs did not provide deep insight into what was going on, since the errors related to absence of action rather than wrong action.

I think what really fixed it was hunting down all the nodes that were set to auto-detect master machines (due to having installed from the .exe rather than with our silent install batch file) and stopping them from confusing everybody. On the architectural level, it seems like this (auto-detecting nodes promoting themselves to Master and fragmenting the swarm when the other nodes use a specific master) is a major point of weakness for however Swarm holds elections under the hood.