Author Topic: Qube Sup issue  (Read 2712 times)

uppi

  • Full Member
  • ***
  • Posts: 10
Qube Sup issue
« on: September 06, 2011, 03:37:31 PM »
Hi,

i am new to this forum,below problem i am facing at present ,


1)The Qube supervisor system net work port we changed to another port after this supervisor system rebooted and  ,regular MAYA  jobs  submitted .
Job  received some workers , after some time  status showing as a failed ??

2)Frame range like 1-100 for this ,subjob processes I given 2 and I selected 2 hosts only
one frame one worker 8 process is not utilizing fully   if I want use 2 process one frame what I have to fallow,???

3)For MAYA job 
Range Execution  :individual is deflate for this I want to give chunk frames how can I  ????

4)Qube  worker & client & supervisor   all logs Dir same if it the same any problem ????


in advance
Thanks

BrianK

  • Hero Member
  • *****
  • Posts: 107
Re: Qube Sup issue
« Reply #1 on: September 22, 2011, 06:10:11 PM »
1.  If you change the port on the supervisor, you must also change the port on the workers, lest they have no idea how to connect to the supervisor.

2. I'm not sure if I understand your question.  I believe you want to run more than one frame on a host & you want to do that on two hosts.  If this is the case, there are three settings to look at:
 - Subjob process: This is the number of frames from this job that will run across your *entire farm* at any one time.  Subjob processes of 2 means only 2 frames will ever render at one time, regardless of how many hosts or cores you have.
 - RenderThreads: You should almost always leave this at -1.  You can explicitly set it to a number > 0, but leaving it at -1 will make Maya render with the same number of threads as the next setting...
 - Reservations:  Click "Browse" then change the number of processors you'd like a frame to use on any one machine.  If you leave RenderThreads at -1, and set host.processors =4, then each frame rendered will use 4 cores on the render node.

So, if, for example, you have two 12 core render nodes, you can fully subscribe each machine with the following settings:
Subjob processes=4
RenderThreads=-1
reservation: host.processors=6
... then select the 2 hosts.

That will allow 4 frames to render across the 2 hosts, each frame reserving 6 cores, and the -1 for render threads signifies that Maya will render with 6 threads - one for each core reserved.  Each frame will use 6 cores, there are 12 cores on a node, so each node will accept 2 frames.  2 hosts * 2 frames = 4 frames at one time.

Note: there is no need to specify hosts.  If no hosts are given, Qube will give the frame(s) to the next available host(s).

3.  For Maya Job type (not the Maya BatchRender, but Maya Render), there is no need to render in chunks.  The only reason one would want to render in chunks is to save the overhead of opening Maya and loading the scene for each frame (you would do it once per chunk rather than once per frame).  However, the Maya Job type loads Maya in prompt mode (on each computer that's running the job), loads the scene file, and leaves it open until all frames are rendered.  This automatically load balances the farm and makes chunking frames irrelevant.

4.  They should all be the same.  Also note the worker_logmode setting.