error message

[Rahul]

I am getting the following error message while I try to render a 3ds max scene file and the message is as follows :  Insufficient system resources exist to complete the requested service.

I have 6 workers nodes all having 8 GB of ram so this is a out of memory error? I would also like to know is there any memory optimizations I could do before I submit my job to the queue.

Any help would be greatly appreciated.

Thanks,
Rahul

[Rahul]

I would also like to know is there any automatic retry option for failed frames?

[shinya]

Hi Rahul,

Do you get the "insufficient system resources" error on all 6 nodes?
Does it improve (for a while at least), if you restart those machines?
It could be the case that you have some program running in the
background that's hogging resources (such as memory). Please check
for that.

Yes, you can set the "retry frame/work" field to a positive integer N
so that failed frames are retried automatically up to N times.

-shinya.

[Rahul]

Yes, I do it get the error on all nodes and it does improve after a while but still some frames/agenda would be failed which has to retried to complete the job.

btw, is this retry frame/work field is in the job submit window?

Also another query is, I doubt whether this happen due to my subjob process value, I don't really understand how  should I divide the job among the nodes. As of now I just set it to 50 just like the authorized dealer showed me..

[shinya]

Hi Rahul,

I think you may be running too many instances of the renderer at once on the
workers, and exhausting the memory on the nodes.  By default, workers are
configured to have as many "job slots" as there are cores. In other words,
for example, an 8-core worker will, by default, run up to 8 subjobs (instances)
at once.

So, you may want to try setting the job's "reservations" to "host.processors=1+".
 ("1+" means "reserve all cores/slots") when you submit, so that each worker
node will run only 1 subjob (thus 1 instance of 3dsmax).  Since the renderer is
multi-threaded, it should still utilize the multiple cores on the workers.

So, if you have 6 workers, then set the "subjobs" count to 6, but also set the "reservations" field to "host.processors=1+".  This will run 1 instance of the
renderer on each node.

[Rahul]

Hi shinya,
               Thanks a lot, that seems to make things much faster and easier.

Rahul