DR render problem

[blackmug]

ok.. hello!
i am using SP2 and i am having some troubles with heavy scenes when i use DR... scene is full of displ, and proxies..
on the main machine i have i7 with 12 GB of ddr3, and on klusters (slaves) i have q6600 with 8 GB of DDR2.. i am using 64 bit XP.

so when i start the render i animate the cam (6 views for example) and i precalculate LC on the main comp, then i copy LC map on the slaves (in the root folder).. BF is primary and LC is sec..

they render first I say 2 pictures very good (2400 pixels) and then on the 3 picture some of the slaves render all till the end of render and the bucket of last 4 processors want render... but the render is 99 % done exept this last 4 buckets of one of the slaves...

thats pretty complicated because i have to stop the render and render it as a region on the main comp, but we r doing animation right now and every 5 or 6 frames that happened..

the last 4 buckets can stay there for hours .. nothing happened (the processor is like 50% of usage on that slave)

should i install SP3 becausei saw that there is restart option on DR.. and what does the restart option do? restart slaves after every render or?

please respond to me..

Thanks

[RErender]

I'm not sure if SP3 will restart the slave after every frame, I think it is after every job.

Best bet is to not use DR for animations, use backburner. If you run the math DR doesn't make sense unless you have less frames than slaves or the slaves are much slower than the main box

[simmsimaging]

Usually when I've had that problem the buckets eventually clear - but it can take a while.

In any case: I believe the DR will restart the spawners only on completion or cancellation of the render, so I don't think it will help you with those hanging buckets. As RErender said, the advice I have heard is not to use DR for animation anyway, so that might avoid the issue.

b

[vlado]

they render first I say 2 pictures very good (2400 pixels) and then on the 3 picture some of the slaves render all till the end of render and the bucket of last 4 processors want render... but the render is 99 % done exept this last 4 buckets of one of the slaves...

Can you get me the scene for this to vlado@chaosgroup.com? It will be very helpful to find the problem. Otherwise, for the time being, you can use backburner to render each frame separately.

Best regards,
Vlado

[VI_MOVE]

Having the same problem here right now. I'm about to render an animation and the deadline is already cancelled because of this strange behavior.

Could this also be a problem of windowsxp x64? All our Workstations/Slaves have xp x64 on but i noticed when a bucket stays for a while (some frames take 2-3 min. some take up to 40 min. within a flythrough just because of one last bucket) the cpu load jumps like crazy inbetween the cores and sometimes even splits between them. Therefor might this also be because of the bad smt-programming of the windows os?

I've put the job now on backburner hopefully that works better...

[VI_MOVE]

The problem was still there on backburner. But I had a transparent shellac material in this scene and after I changed that to a normal vraymtl it worked without any problem. No more stuck buckets...