Renders freezing

[tigerX]

Hi
I'm having a problem with long renders getting stuck on a frame.
Its happened a few times now, usually about 2 days into an animation render, different scene's, different rendering situations. The buckets just stop moving round the frame, the machine appears to be rendering, but the bucket doesnt move.

Im rendering on BOXX workstations, through backburner.

So far the only way I've been able to solve it is to kill the offending machine as a server, then restart it as a server & off it goes again.

Thought it might be an overheating problem, but BOXX says the machines would do much weirder things if they were too hot!

I know you gyus know about these things, any ideas...all will be greatly appreciated

Dougie

[gilpo]

This happens to us from time to time. It seems totally random. We've had jobs go for days and days with not a single stuck frame. Then other jobs seem to stick all the time. One solution is to set a task timeout to an amount that will still let the frame render, but will end the task if it goes too long on a single computer.

[vlado]

What is the configuration for these machines? E.g. are they dual/dual-core/hyperthreaded? Also, is there any dynamic geometry in your scene (V-Ray proxies, displacement, fur)?

Best regards,
Vlado

[tigerX]

Vlado, they are dual core, and yes there have been quite a few proxies (trees) & large scale displacement (profiled roofing) in the scenes I've been doing recently.

[Candelero]

Vlado, they are dual core, and yes there have been quite a few proxies (trees) & large scale displacement (profiled roofing) in the scenes I've been doing recently.

I remember an old thread with a < Grass - carpet > animation, where
from 100 frames, all rendered normally for 2-3 minutes, and only frames 77, 78 (in memory) stucked for hours.

Another user-problem was having an extruded spline with very tine fillets representing a house stair (steps).
All that had been displaced with a carpet map and it also stucked.
The issue was that the extrussion was done with only 1 segment, so the
filleted cells, were very long and only some tiny units wide ( X, Y dimensions). Raising the extrussion number (making the topology more dense - and the cells more quad-like) fixed the hanging of the rendering.

So depending on the viewing angles, displacements and other geometry which need dynamic memory can often lead to render-stuck, when the geometry isn't well prepared for them (displacements, furs )

Always keep those in mind, and consider your mesh-topology properly !



Best regards,
nikki Candelero

[gilpo]

If it helps to know, ours are dual-core AMD Opterons. Almost everything in our scene is a vray proxy. No displacement here, though.

[vergefio]

i'm having from time to time the same problem here (dual dual core AMD opteron)... scenes are different and different rendering situation but sometimes they just stop to render, it seems that it's rendering but it doesn't... proxy, or proxy + displace (many and heavy), or even nothing and i never use dynamic memory.

[Morbid Angel]

hehe...well I never had it stop render. But I did have it render for a long time. What you guys are seeing is probably very complex scenes rendering for very long. I suggest you let a render go for over night and see if it changes next morning, if it doesnt then there is defenently a problem

[gilpo]

Morbid,
I can't speak for the others, but that's exactly what's been happening. I'll be rendering an animation from a pre-calced IM and LC and every frame is rendering in 60 secs but all of a sudden, one or 2 frames out of 1000 will just get stuck. The CPU will still be at 100% but it will not move for 10 hours. I'll reset the task in backburner and another computer will pick it up and render it in 60 secs.

Not a big problem here, but an annoyance when that one stuck computer happens to be your fastest machine on the farm.

[Morbid Angel]

i see, it sounds like a problem which some one posted a while ago. they had a dispalced roll of grass unrolling into a flat line; the problem was fixed by switching the types of antialiasing, from adaptive subdivision to adaptive qmc.

[vlado]

If you are willing to test out a new build, you can email me at vlado@chaosgroup.com; there were some memory management problems that were fixed so they may have been the reason for these issues.

Best regards,
Vlado

[tigerX]

nikki & morbid, thanks for the input. I got a feeling in my scene that it is actually the grass, loads of it, heavily displaced with a sharp disp map...something to keep in mind.

Vlado, thanks, mail sent.

Dougie

[Candelero]

nikki & morbid, thanks for the input. I got a feeling in my scene that it is actually the grass, loads of it, heavily displaced with a sharp disp map...something to keep in mind.

Vlado, thanks, mail sent.

Dougie

Can we see the WIRE of the grass objects ? Thanks!

Best regards,
nikki Candelero

[tigerX]

nikki,
the grass is just a fairly simple mesh, over a largish area- tho, thinking about it, its probably not sub-divided enough. I'm not *too* bothered about resolving that particular scene, 'cause its all rendered out now & being compiled...my heads already on the next one!!-
I'm fairly happy its some sort of memory issue, which hopefully will be resolved with the newer build, meanwhile i'll check future geometry for the issues that you mentioned earlier.
Thanks for the help
Dougie

[vergefio]

thx Vlado, going to sent you an email, i can send you even the scene but actually it's pretty heavy (over 100mb), and i'm not sure cleaning it would help finding the problem though