clients not reporting chunk finished

[madCGi]

though we currently use max 2016 and vray 3.3 we have seen this issue using 2015 and older versions of vray. but we have been having more issues recently.

[ch3m1kal]

No domain right now, that's actually planned shortly and I'm curious if it will fix the issue in any way.
RenderPal runs as an app, since it's too complicated to handle permissions for services right now.

[madCGi]

No domain right now, that's actually planned shortly and I'm curious if it will fix the issue in any way.
RenderPal runs as an app, since it's too complicated to handle permissions for services right now.

well that helps me rule out whether our domain permissions are to blame and that running as an application or service makes no difference.

seems to me like its has to be one of the following:
1. problem with 3ds max > personally i think this is unlikely.
2. problem with RenderPal > though the problem is with RenderPal not registering completed jobs, i dont believe there is anything that should cause this issue.
3. temporary loss of connectivity due to server traffic/load > i feel like this is probally the cause of the problem, but it is very hard to diagnose and prevent.

whatever is causing this issue it could be solved if RenderPal checks periodically if there is a difference / discrepency between the clients event log and the job the server believes it is running.

[madCGi]

dont know if this has anything to do with it, but what anti-virus/security and firewalls are you running?

fyi: we use symantec.

[ch3m1kal]

Already ruled security software out.

Dedicated render slaves are separate from artist machines and they have the same issue even though they run no additional security and even the windows firewalls are disabled.
I think the server failing to acknowledge a finished chunk event is most likely, since the render slaves just sit there, with no current job showing up in the client and remotely disconnecting the stuck client causes the last rendered chunk to restart.
That might also explain why sometimes, though very rarely, I see chunk restart events "while machine was offline" even though the machine was operating properly the whole time.

I'm thinking of maybe trying to migrate the RenderPal server to a Linux environment and see if somehow that fixes anything, since currently that's running on Windows too.

[madCGi]

That is exactly the same problem we have been having with exactly the same symptoms. If RenderPal could somehow check for this that would fix it. Until that happens to properly diagnose it I guess I have to........ Purposely overload the server!?? :p

[Shoran Support]

I also believe that the only fix is a period client check; added to our to do list.

[madCGi]

I also believe that the only fix is a period client check; added to our to do list.

good to hear. i will post more if i find another workaround or solution, but hopefully this can be implemented and rolled out quickly.

[ch3m1kal]

I also believe that the only fix is a period client check; added to our to do list.

Maybe add a server acknowledge event after a client reports chunk finished, and if the client does not receive that it retries every x seconds?
Assuming it doesn't already work that way, of course, I haven't actually had time to analyze RenderPal's network traffic.

[ch3m1kal]

Hey, I've actually managed to catch one of the stuck clients with verbose message log enabled.

See if it helps in any way.