feature request: have sgfxi shut down BOINC
Hello,
Given that BOINC uses the NVIDIA driver to do work units on the GPU, it would be nice if sgfxi checked for a running BOINC client and stopped it before doing an install and then started it up again before starting the desktop. If I don't do a reboot or restart BOINC manually after an upgrade, work units that use the GPU think the graphics card is not there. I don't know if ATI folks would benefit from the same sort of thing. Thank you for your attention, Myrmidon Back to top |
You'd have to provide me with some more concrete details, that is, explicit steps etc, and how sgfxi would know boinc is running.
Since this is the first time I've ever heard of this boinc thing, I'm not really in a position to handle it without having a better grasp of what it is exactly. Back to top |
Hello,
BOINC is a platform for donating unused computing resources to scientific projects. (see boinc.berkeley.edu/ ) Once the user has signed up with the different projects to donate computing time, the BOINC client will download work units that either use the cpu or the gpu depending upon their preferences. It is the the work units that utilize the gpu that will stall after installing or updating an nvidia driver. Milkyway@home and Einstein@home are two projects that deliver gpu work units. Under Debian, Boinc can be installed :: Code :: aptitude install boinc This grabs the client and the manager. After apitiude full-upgrade delivers a new kernel from say liquorix, I reboot to the new kernel and am eventually delivered to a terminal. Before that happens the init script in /etc/init.d/boinc-client will have run and it will not be able to utilize the graphics card (no driver). So I then run sgfxi and the new kernel is installed and select start desktop and am delivered to my desktop. Opening up boinc-manager will show under the tasks tab any gpu work unit complaining that "GPU is missing" So I restart boinc with :: Code :: /etc/init.d/boinc-client restart As to how sgfxi can detect boinc running, I don't know how sgfxi works or how any script works besides reading it and getting a mild headache ;) , but I do know that in a terminal as root :: Code :: /etc/init.d/boinc-client status returns either something like :: Code :: Status of BOINC core client: running.
Scheduling of BOINC core client: 3101. pid 3101's current scheduling policy: SCHED_OTHER pid 3101's current scheduling priority: 0 Scheduling of BOINC core client's children: 4971. pid 4971's current scheduling policy: SCHED_BATCH pid 4971's current scheduling priority: 0 OOM killer status for BOINC core client:. PID 3101: adj 0, score 1 PID 4971: adj 0, score 0 or :: Code :: Status of BOINC core client: stopped.
start , stop and restart are available. I hope that I have answered your questions. I don't post often anywhere. Starting the client manually is really no great hardship and anybody who did not check would have it running properly on their next boot. I just thought it would be a nice addition if practical or possible. Thanks again, myrmidon Back to top |
that seems possible, however it would be a debian only feature in sgfxi since trying to track such specifics is a big headache. Maybe also ubuntu, but ubuntu is moving away from the init script system, or has already moved away from it, not sure.
However a small routine could check that, if running, stop it, set boolean global, and if global set, restart, or try to restart, at completion. Not sure about fglrx, I have my doubts that it works, but don't know. Back to top |
That probably OK for some that used the Debian packaged client, but most people I know run the client in user space i.e. out of a boinc directory in their home.
The Boinc Manager also runs in the panel so it is not too much of a problem to shut it down if you use it I've been doing seti@home since 1999 although I no longer run a "farm". I have to admit it was a great source of heat in my office in the winter and also a great form of geek competition with the rankings. Back to top |
All times are GMT - 8 Hours |