Re: [Savannah-hackers-public] excessively high load avg, again (daily?)
This high load may be due to the nightly rsync you see below: Looks more like a runaway python to me. PID USER PR NI VIRT RES SHR S %CPU %MEMTIME+ COMMAND 1927 www-data 39 19 1090m 740m 3596 S0 12.1 425:43.54 python 2691 nobody20 0 109m 58m 33m S0 1.0 0:02.24 git 10840 root 20 0 107m 48m 1556 D1 0.8 1:19.83 rsync (BTW, not to be teaching my betters, but I suggest top c to show command lines and not just names.) But anyway ... Michael, what is the rsync job doing? It is easy to make it consume less resources in exchange for more time, --bwlimit=100 (Kbytes/sec) or whatever number turns out to be good. It should probably also be running with nice -19. Best, Karl
Re: [Savannah-hackers-public] excessively high load avg, again (daily?)
On 12/09/2011 07:49 PM, Karl Berry wrote: This high load may be due to the nightly rsync you see below: Looks more like a runaway python to me. PID USER PR NI VIRT RES SHR S %CPU %MEMTIME+ COMMAND 1927 www-data 39 19 1090m 740m 3596 S0 12.1 425:43.54 python 2691 nobody20 0 109m 58m 33m S0 1.0 0:02.24 git 10840 root 20 0 107m 48m 1556 D1 0.8 1:19.83 rsync (BTW, not to be teaching my betters, but I suggest top c to show command lines and not just names.) But anyway ... Michael, what is the rsync job doing? It is easy to make it consume less resources in exchange for more time, --bwlimit=100 (Kbytes/sec) or whatever number turns out to be good. It should probably also be running with nice -19. I nice'd rsync to 19, The python process, loggerhead, is already nice'd to 19 as well.
Re: [Savannah-hackers-public] excessively high load avg, again (daily?)
On Thu, Dec 08, 2011 at 01:18:13PM +0100, Jim Meyering wrote: This high load may be due to the nightly rsync you see below: The rsync job (nightly backups) is likely the cause of the high IO wait (22% in your top output). Still, I don't think that that would account for all of the high load average. Thanks, Ward. -- Ward Vandewege | CTO, Free Software Foundation GPG Key: 25F774AB | http://identi.ca/cure | http://fsf.org/blogs/RSS Do you use free software? Donate to join the FSF and support freedom at http://www.fsf.org/register_form?referrer=859
Re: [Savannah-hackers-public] excessively high load avg, again (daily?)
On 12/8/11 7:18 AM, Jim Meyering wrote: Since there were some git-daemon processes dating back to November and since we have a limit on those (at least I think that's what Michael said), I've just killed those November git-daemon processes. I just added a timeout for them, just like we do for bzr. Perhaps related, there are lots of these in dmesg. We get from 1 to ~6 per hour: [2572324.224898] cgit.cgi[29777]: segfault at 863d000 ip b768c810 sp bfe9da2c error 6 in libc-2.11.2.so[b7619000+14] I'll investigate this further later today... In the near future I'll upgrade the version of git we're using, along with upgrading cgit.cgi. -- Michael J. Flickinger
Re: [Savannah-hackers-public] excessively high load avg, again (daily?)
Michael J. Flickinger wrote: On 12/8/11 7:18 AM, Jim Meyering wrote: Since there were some git-daemon processes dating back to November and since we have a limit on those (at least I think that's what Michael said), I've just killed those November git-daemon processes. I just added a timeout for them, just like we do for bzr. Perhaps related, there are lots of these in dmesg. We get from 1 to ~6 per hour: [2572324.224898] cgit.cgi[29777]: segfault at 863d000 ip b768c810 sp bfe9da2c error 6 in libc-2.11.2.so[b7619000+14] I'll investigate this further later today... In the near future I'll upgrade the version of git we're using, along with upgrading cgit.cgi. Great. Thanks for pursuing it.
Re: [Savannah-hackers-public] excessively high load avg, again (daily?)
On 12/8/11 3:30 PM, Jim Meyering wrote: Michael J. Flickinger wrote: On 12/8/11 7:18 AM, Jim Meyering wrote: Since there were some git-daemon processes dating back to November and since we have a limit on those (at least I think that's what Michael said), I've just killed those November git-daemon processes. I just added a timeout for them, just like we do for bzr. Perhaps related, there are lots of these in dmesg. We get from 1 to ~6 per hour: [2572324.224898] cgit.cgi[29777]: segfault at 863d000 ip b768c810 sp bfe9da2c error 6 in libc-2.11.2.so[b7619000+14] I'll investigate this further later today... In the near future I'll upgrade the version of git we're using, along with upgrading cgit.cgi. Great. Thanks for pursuing it. Upgraded git: git version 1.7.2.5 Upgraded to latest version of cgit: 0.9.0.2 I'm not really sure I like the idea of C cgi programs. Hopefully the segfaults will stop. If they don't, I'll debug it when I have some time... -- Michael J. Flickinger