Advanced search

Message boards : Graphics cards (GPUs) : Not getting jobs - Linux Boinc 7.2.42

Author Message
Paracelsus
Send message
Joined: 11 Aug 10
Posts: 11
Credit: 21,424,870
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 46172 - Posted: 14 Jan 2017 | 17:07:07 UTC
Last modified: 14 Jan 2017 | 17:18:00 UTC

Hi fellow Grid Denizens,

My client has not received jobs over several days, though the server status indicates work units are available. It's been about a year since I last ran GPUGrid jobs,but would like to start contributing cycles to the project again.

Client is running Rosetta CPU jobs fine, GPUGrid is my only GPU project.

I've reset the project, looked at project preferences, etc. but no joy so far.

Newer builds than 7.2.42 aren't readily available for Suse, though I might be able to jab in the Fedora builds for 7.6.22, or build from source or a spec (though I think that was a bit of a pain last time I tried). And I'm not sure the issue is with the client version.

I'd appreciate any assistance to get things running.

Cheers,
Pete


Linux (OpenSuse 13.2)
Nvidia driver 375.26
GTX650 Ti Boost

rpm -qa | grep -i boinc
libboinc7-7.2.42-2.1.5.x86_64
boinc-manager-7.2.42-2.1.5.x86_64
boinc-client-lang-7.2.42-2.1.5.noarch
boinc-manager-lang-7.2.42-2.1.5.noarch
boinc-client-7.2.42-2.1.5.x86_64

No apparent dependency issues, nor runtime errors to std err

Event Log repeat entries of:

Sat 14 Jan 2017 09:31:39 AM EST | GPUGRID | Sending scheduler request: Requested by project.
Sat 14 Jan 2017 09:31:39 AM EST | GPUGRID | Requesting new tasks for CPU
Sat 14 Jan 2017 09:31:42 AM EST | GPUGRID | Scheduler request completed: got 0 new tasks
Sat 14 Jan 2017 09:31:42 AM EST | GPUGRID | No tasks sent
Sat 14 Jan 2017 10:31:44 AM EST | GPUGRID | Sending scheduler request: Requested by project.
Sat 14 Jan 2017 10:31:44 AM EST | GPUGRID | Requesting new tasks for CPU
Sat 14 Jan 2017 10:31:46 AM EST | GPUGRID | Scheduler request completed: got 0

Any suggestions on what might be awry?

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2343
Credit: 16,201,255,749
RAC: 7,520
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 46173 - Posted: 14 Jan 2017 | 17:24:50 UTC - in response to Message 46172.

If you look closely you'll find the reason for it as your host asks GPUGrid for CPU tasks not for GPU tasks:

Sat 14 Jan 2017 09:31:39 AM EST | GPUGRID | Requesting new tasks for CPU ... Sat 14 Jan 2017 10:31:44 AM EST | GPUGRID | Requesting new tasks for CPU

Perhaps your BOINC manager did not detect your GPU, please check the first lines in the BOINC manager at its startup.

Paracelsus
Send message
Joined: 11 Aug 10
Posts: 11
Credit: 21,424,870
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 46178 - Posted: 14 Jan 2017 | 20:57:03 UTC

Thank you for pointing out the "Requesting CPU tasks" (not GPU), and indeed the first lines of event log included "GPU detection failed. error code 512"

After modprobe nvidia_uvm the GPU was still not being detected.

I realized as well I had been starting the client bdirectly via /usr/bin/boinc, and while that allowed Roesetta jobs to run, starting via systemd was needed for GPU detection - which makes sense, I just wasn't thinking of it.

When started via systemd GPU was detected and info logged to /var/lib/boinc/coproc_info.xml and of course the GPU then showed in the initial lines of event log.

Thanks for the help! Chomping away on work units once again...

Post to thread

Message boards : Graphics cards (GPUs) : Not getting jobs - Linux Boinc 7.2.42

//