Results 1 to 7 of 7

Thread: Iray server cpu only mode?

  1. #1
    Join Date
    Jan 2018
    Posts
    6

    Question Iray server cpu only mode?

    I'm setting up iray server to run on a VM on my nas. I just want it to run as the master node, handling the images, db, etc.

    Then my other machines with GPUs will run as slaves in the cluster.

    Problem is, when I start the server, it just shuts down after a lot of "waiting on workers"

    Is there a way to run the master without GPUs? Like a CPU only mode or something? Can't find anything in docs or googling

    Thanks!

  2. #2
    Join Date
    Dec 2017
    Location
    Melbourne, Australia
    Posts
    376

    Default

    This is definitely not a directly supported configuration, the thing likely to give you more issues is running inside a VM than anything else. I believe you can disable rendering on the master but I will need to confirm this tomorrow when I can test on our cluster. You likely will need to use TCP mode however for the cluster since I doubt your VM is going to handle multicast traffic very well. Is the network interface in the VM on the same subnet as your other machines?

    In terms of the shutting down, the best place to check for issues if it is not starting the workers would be the worker logs themselves, which are not part of the main logs, which you will find in the %ProgramData%\NVIDIA Corporation\Iray Server directory in the worker*.log files. It would be helpful to have your full logs to diagnose further, however as I say, running in a VM is in general not a supported configuration.

    You'll definitely have to exclude the NAS from rendering, while the clustering can work well with somewhat different machines, having one that under powered would probably cause issues and contribute almost nothing in terms of performance.

    I'll check further into what may or may not work though and come back.

  3. #3
    Join Date
    Jan 2018
    Posts
    6

    Default

    Yea the VM to my knowledge would work just fine. Same network, etc (ubuntu 16.04) It's just a celeron cpu qnap nas. Problem is I can't find any docs that show much in the way of command line info. Yea if I could just turn cpu and everything off and let it just be the master would be perfect. (I know you can do this once it's up and running at least there is an option there, but don't know how to do it before hand)

    I have 3 other machines with 10series gpus I'm going to use for the slaves.

    I also have a dual xeon machine i could use too, but it has no video either, was just going to use it's cpu power.

    The worker's directories don't seem to have log files in them. I know they do when it's running (have it running on google compute, I know what you are talking about on the logs) but I'm guessing if it doesn't start they don't log? Odd...

  4. #4
    Join Date
    Dec 2017
    Location
    Melbourne, Australia
    Posts
    376

    Default

    Yes, if the workers don't even try to start then you won't get any logs. Do you have the main iray_server.log file you could share?

  5. #5
    Join Date
    Jan 2018
    Posts
    6

    Default

    Ok, so now have 3 machines 1 master 2 slaves, no vm.

    They all link up, see them in the resources page.

    The config
    Master has two 960s and cpu
    1 slave has a 1080 and cpu
    2nd slave just has dual xeon cpus, no gpu

    both slaves seem to start and connect, slave 2 complains about not able to start cache manger, but then connects to the master.

    Nothing seems to happen on the slaves. Slave 1 will load up memory, like it's going to do something, but the cpu or 1080 never do any work.

    So it's like only the master is doing any work. The rest just kind of sit around doing nothing.

    What in the world am i doing wrong.

  6. #6
    Join Date
    Dec 2017
    Location
    Melbourne, Australia
    Posts
    376

    Default

    Do each of the nodes run Iray Server tests successfully (you can just use the benchmark on the resource page for that) when used individually and not clustered?

    You would need to review the server and worker logs on the nodes to establish what is happening and whether there are any errors, but as I say, you first want to be sure everything works on each node well since they won't do any better when clustered if they don't work individually.

  7. #7
    Edyth Guest

    Default

    I also have a dual xeon machine here
 i could use too, but it has no video either, was just going to use it's cpu power.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •