[ Date Index ][
Thread Index ]
[ <= Previous by date / thread ] [ Next by date / thread => ]
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Wednesday 05 Feb 2003 12:51 am, Simon Waters wrote: > Neil Williams wrote: > > what would cause a previously working system to try to run > THIRTY copies of > > cron and thirty copies of python -S /var/lib/mailman/qrunner, > leading the > > load average to hit 47! > > Okay cron spawns cron to do things, so 30 copies with 30 copies > of qrunner is probably qrunner not finishing fast enough. So that leads to a feedback loop - one doesn't finish because of overload, so start another and add to the load !??!! > > It all started with that Mozilla spell-check error and ever > since, whether > > mozilla is running or not, I suddenly find that the machine > won't respond to > > the mouse, keyboard or network for upto TWENTY minutes at a time. > > What error? http://www.dclug.org.uk/archive/msg00894.html > > Sounds like typical excess load, I assume you could drop > priority of cron jobs, perhaps limit resources to any > problematic programs - users. I'm back online with it now, but by using a pre-existing telnet connection and su and top, I found that KPPP was suddenly writing 900MB of swap data. I had to load diskdrake and unmount and format both swap partitions (2 hard drives, one swap on each). When I reloaded KPPP using the suspect connection, the swap data re-appeared and continued increasing as before. Only by using Ctrl-Alt-Backspace was I (eventually) able to restore control. (Cron is currently stopped). > I get nasty problems if Demon serve up too much spam all at > once, as it spools too many Python jobs to sort the wheat from > the chaff. Can you wrap qrunner not to run if it is already running? > > i.e. Something like - and I haven't tested it. > mv qrunner qrunner.real > > vi qrunner > RUNNING =`ps -ef | grep qrunner.real | grep -v grep | wc -l` > if [[ RUNNING -eq 0 ]] > then > qrunner.real > fi Thanks Simon, I'll try that. At the moment I'm intent on using top to see what is writing so much swap data and working from there. Quite where the original error has been introduced I don't think I'll ever really know. The fact that it happens just as badly when cron has been stopped since login and that it never happens unless a user IS logged in under X, leads me to think it's a software config error. > > > I've tried stopping crond before starting any work, but now > even KPPP causes > > the system to loop. I'm beginning to think it's the linmodem > driver. > > Possible hardware issues as well - but the cron thing sounds > just like overwork. > > Did you change any hardware or kernel settings? No. I just had that error from Mozilla. (I will investigate KPPP though - I did change those settings.) > > BTW. Does anyone know what 'No carrier' really means in terms > of the level of > > line noise? > Too much noise to work - how much more detail do you need ;-( > > BT can measure line gain issues automatically from the exchange > - some voodoo in modern exchanges. If it is intermittent I'd > start with cabling your end, and try a different modem (or check > if you have too many phones/answer phones plugged in - that kind > of stuff). This No Carrier result happens even when the PC is connected straight into the BT socket (as the sole connected item). Two different machines, 3 different modems, 5 different distros, 3 different ISP's. > > BT can be good on modems, but I think you are only garanteed > 9600 baud under the USO. But if I could only get 9600, it would still connect wouldn't it? The fact that it won't connect at all, at any speed, is a problem with the line? No? - -- Neil Williams ============= http://www.codehelp.co.uk http://www.dclug.org.uk http://www.wewantbroadband.co.uk/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.7 (GNU/Linux) iD8DBQE+QWu0iAEJSii8s+MRAgl3AKCDnSWtMaRWAakSZ+5CIxrcC/StjQCgkiwo coV4KQYHpkJIGLa7FoTyYEE= =m54f -----END PGP SIGNATURE----- -- The Mailing List for the Devon & Cornwall LUG Mail majordomo@xxxxxxxxxxxx with "unsubscribe list" in the message body to unsubscribe.