Hi
Met this quite strange problem on one of the nodes.
So we had DR process going on , and couple of buckets left - about 5. They were busy with one of the nodes (dual xeon , so it had more than 5 threads to use) , so we noticed that it takes quite a lot of time.
Checking out the task manager i found out that that node had processor busy on 0% , though scene was loaded and connection was initiated. There is absolutely nothing in VRay logs , like it is usual rendering process going on.
We are using Max 2014 and VRay 3.20.02 (still did not switch to 20.03).
So we decided to kill VRayspawner on that machine - buckets were picked up immediately by another spawners and rendered in no time.
Any guess why that can happen ?
How many other machines did you have in that DR render altogether? How many CPU cores do they have?
Best regards,
Vlado
We have a constant number indicated in VRayDR.cfg file of 8 machines if i remember correct (not at the office now) , and some of them are 4 , some of them are 6 cores (all with HT)
Though we did not have this kind of issues before , i can try to render this scene again, with attempt to isolate problem.
This happened only on one spawner - the Xeon CPU , others are I7’s different models.
If you can reproduce it, it will be helpful, yes.
Best regards,
Vlado
This issue might be caused by a previous installation as well. Therefore could you please try to do a clean install? Chaos Docs
Make sure that after uninstall V-Ray you delete any remaining V-Ray related files from here:
C:\Program Files\Autodesk\3ds Max 20xx
C:\Program Files\Autodesk\3ds Max 20xx\plugins
and C:\Program Files\Chaos Group\V-Ray\3dsmax 20xx
Then install V-Ray and let us know if the issue persist.
I tried another job and looks like it is rendering fine now , as it was rendering fine before that job.
I will try that specific scene a bit later today.
Got this thing again. Scene is loaded in memory - about 5 GB of ram , the same node again, only its 3 buckets left, though they were not rendering at all , as CPU usage was 0. Stopped VRayDR SPawner there - at the same time another node picked that up and finished rendering.
VRay is started as usual - not a service , all network paths visible , connection to the node is stable. Happens not on all of the scenes and looks like only with this node. Last time it was exactly the same one.
The next time it happens, before killing the spawner on that machine, can you make a crash dump from the Task Manager for the 3dsmax.exe process and get it to me at vlado@chaosgroup.com ?
Best regards,
Vlado
I sent you that dump couple of days ago , you had any chance on checking that ? We had exactly the same issue again with the same node.
I got the .dmp file, yes. I’m still investigating what happened.
Best regards,
Vlado
Sorry for bumping again , but we are concerned about this , as if this happens - renders goes to infinite amount of time , have you got a chance to solve this thing ?
This thing got to a new level, got render stuck and i cannot identify server name on buckets 
I restarted all of the distributed servers rendering this scene but that had no effect - it is still stuck.
In resource monitor it is 17% CPU and 25.131.532K of memory , not moving at all in any direction.
We are still using 3.20.02.
Does that happen on every scene or just on particular one?
Is there a chance to send it for investigation?
A good idea is to update to the latest version 3.30.03 and see if it happens with it too.
Thanks for your answer, Svetlozar , i am already in progress of resolving this with support via e-mail.
This happens on random scenes , random nodes. Weird thing for me is that the text identifying the name of slave rendering looks like to be consisting of several names together. Quite strange ?
In about 2 hrs this scene was finished to render - looks like these 2 buckets belonged to the local host - not the nodes.
The overlapping names seems strange indeed, it might happen if the bucket has been sent to several different nodes which might be a memory issue.
Have you checked how much memory is occupied during the rendering? If the bucket requires more memory that available one on the slave it might crash and be forwarded to another machine.
It took about 25gb , with host machine 64gb and nodes 24gb-32gb of memory.
Quite weird thing is that one of the main servers (which was acting as host for DR nodes) died couple of days ago , not starting at all , with motherboard LED status reporting something related to CPU bridge.
So it might be a hardware related error , though cannot be sure about that. We changed main server to another machine and for now everything is smooth.
Also we plan to update to 3.3 this week , once we decide how to be with the different whitepoint (i know about script commands , though need to coordinate that with project manager) , so we will check how that works also.
It could be a hardware issue for sure but since the scene uses 25GB on the workstation and some of the machines has less memory that might be also the reason for dropping spawners.
You could easily find out if that is the case if you optimize the scene to use about 20GB and run the same test.
Yeah that’s the thing i want to try when we will encounter that next time. For now we updated to 3.30, for now it looks fine.
Glad to hear that with 3.30.03 it works fine so far.
Keep us posted if the issue occurs again.