3
$\begingroup$

I'm running Blender 2.73 with Network Render using VCS setup with Git.

My jobs just get stuck in a queued mode. I've left it for hours and nothing helps.

All my slaves are on a remote cluster. Everything (client, master, slaves) are all on Linux. client&master are on the same Ubuntu 14.04 box, and slaves are all on a CentOS 6 cluster, three nodes including the headnode.

My setup is, my client runs in the GUI on the same box as the master running master.blend. The slaves all open tunnels back to the master and run blender in CLI mode using slave.blend.

My file, ozone.blend is in git. On the client I specify everything requested

VCS Settings on Client

My slaves nodes all connect as you can see in that image, I also see them in the master console at http://localhost:8000

I go to my slaves which operate out of a clone of my repo, and I pull the newest changes. These files are shared across the cluster, but /tmp/ is unique to each node.

But my jobs simply stay in my queue Jobs stuck in queue

Why do they get stuck? Am I using the VCS tools poorly? The manual instructions aren't much help, and while I see other people with issues , I don't see any solutions.

Should I not have clones of my repo on my slaves? Should the clones be in separate directories?

I've tried restarting everything, including the master. When doing so I delete everything related to them out of /tmp, etc.

Note, without VCS, the jobs run.

$\endgroup$
5
  • $\begingroup$ I have never gotten netrender to work using VCS or plain Blender. Everything connects to the master but nothing ever comes off the queue. Everything successfully connects to the master. Same version of Blender everywhere. Same version of MacOS everywhere. The only clue is the a message on the master status area that says "Version check". The web monitors work, I can check job status. Nothing ever renders. Started reading the code yesterday and will add my own diagnostics. No idea where to look for support. :-( $\endgroup$
    – Chuck
    Jan 27, 2015 at 22:21
  • $\begingroup$ That's not encouraging. :) I was hoping that because I'm using a cluster, I could avoid uploading everything N times over, but I guess not. Thanks for the feedback. Is there any way you know of that I might avoid multiplying the upload time? $\endgroup$
    – Matt
    Jan 27, 2015 at 22:43
  • 1
    $\begingroup$ I might give up on netrender for a while if this solution works... blender.stackexchange.com/a/5263/9651 $\endgroup$
    – Chuck
    Jan 27, 2015 at 23:33
  • $\begingroup$ Nice, I had come accross something similar ( blender.stackexchange.com/a/7221/10933 ) before, thinking of also giving it a try. Netrender hasn't been living up to what I imagined it would be.. $\endgroup$
    – Matt
    Jan 28, 2015 at 4:24
  • 1
    $\begingroup$ I think my next step will be to get up to speed on Brenda, which allows you to render on AWS. Check out github.com/jamesyonan/brenda . $\endgroup$
    – Chuck
    Jan 28, 2015 at 5:19

0

Browse other questions tagged .