| nibalizer | slangasek: its running in valgrind now | 00:30 |
|---|---|---|
| nibalizer | in screen | 00:30 |
| nibalizer | and appending the valgrind log to a file | 00:30 |
| nibalizer | i think im a terrible person | 00:30 |
| === mbarnett` is now known as mbarnett | ||
| bkerensa | slangasek: So how would I request approval to upload raring doc? | 01:13 |
| bkerensa | ;) | 01:13 |
| bkerensa | or get it sponsored even | 01:13 |
| bkerensa | Doc Team has just finished changes to doc and we handed it off to translations | 01:13 |
| bkerensa | running behind | 01:13 |
| slangasek | nibalizer: so in case you're not familiar with valgrind, let me point out that you don't want to run it to OOM, because then valgrind won't be able to report on the leaks... probably running for an hour or two and then ^C'ing it is sufficient | 01:33 |
| slangasek | bkerensa: approval to upload raring docs> talk to the DMB about PPU | 01:33 |
| bkerensa | slangasek: no no :) not upload rights... Uhh exception for uploading raring doc after string freeze | 01:53 |
| bkerensa | slangasek: the doc series was made after the string freeze | 01:53 |
| bkerensa | slangasek: raring is shipping old docs right now ;) | 01:56 |
| bkerensa | Version: 12.10.3 | 01:57 |
| nibalizer | slangasek: kk | 03:25 |
| c_smith | hmmmmm....... activity on this channel, that's kinda rare (or seems so to me, but I'm not on IRC 24/7) | 03:27 |
| c_smith | anyone know if there'll be a release party for 13.04 thrown by the LoCo team? | 03:28 |
| QIII | This would be the first time I have seen a live person on here after joining the team -- and I haven't been able to make any events because they are not scheduled when I can make it down from Portland. | 03:32 |
| c_smith | QIII, dang, would love to see you at the Ubuntu Hours, the days can still be tweaked if needbe for people. | 03:34 |
| QIII | Oh, it would probably be 7:30 or 8:00 before I could make it to Salem on a week night. Any chance anyone brings a laptop with Skype? | 03:36 |
| nibalizer | slangasek: okya have have some datase | 03:36 |
| nibalizer | http://i.imgur.com/mzymxK1.png | 03:36 |
| nibalizer | https://raw.github.com/nibalizer/gisst/valgrind.log-22200/valgrind.log | 03:36 |
| nibalizer | on first inspection, the box lost a ton of memory, but valgrand doesn't seem to think so | 03:37 |
| c_smith | I usually have a laptop, a Lenovo IdeaPad U400, and I'll have to see if Arch has a package of Skype available. | 03:37 |
| slangasek | bkerensa: per https://wiki.ubuntu.com/DocumentationStringFreeze, if you want to upload the docs after string freeze you should get a consensus with the translators | 07:16 |
| slangasek | nibalizer: so, are you sure the memory usage isn't due to caches or the like? maybe it would be worth having a look at those pmaps of yours | 07:17 |
| slangasek | nibalizer: what *exactly* is being graphed in those various graphs? | 07:18 |
| nibalizer | slangasek: Its collectd running one of its collectors | 07:21 |
| slangasek | ok, but what is it collecting? | 07:22 |
| nibalizer | blkperl: what exactly are we graphing with collectd.*.memory.memory-* | 07:22 |
| blkperl | slangasek: uh... used-memory vs free-memory | 07:26 |
| * nibalizer imagines steve wants to know how thts calculated | 07:27 | |
| nibalizer | like does it shell out to free -m or what | 07:27 |
| slangasek | well, for what definition of "used"? is the kernel page cache part of your "used" count or your "free" count? | 07:27 |
| blkperl | slangasek: https://github.com/collectd/collectd/blob/master/src/memory.c | 07:27 |
| slangasek | s/page cache/disk cache/ | 07:27 |
| blkperl | https://collectd.org/wiki/index.php/Plugin:Memory | 07:28 |
| nibalizer | slangasek: when you run `htop` it reports slapd as consuming most of the memory | 07:28 |
| slangasek | ok | 07:28 |
| slangasek | so in that case, yes, would like to see some pmaps | 07:28 |
| slangasek | fwiw this may be a bdb tuning issue | 07:29 |
| nibalizer | kk | 07:32 |
| nibalizer | https://raw.github.com/nibalizer/gisst/pmap.2013-04-05-18206/pmap.2013-04-05 | 07:35 |
| nibalizer | https://raw.github.com/nibalizer/gisst/pmap.2013-04-09-3394/pmap.2013-04-09 | 07:35 |
| nibalizer | the earlier one is right after a restart so is 'healthy' | 07:35 |
| nibalizer | the other one is when consuming massive rseoruces | 07:35 |
| slangasek | hmm. everything points to it being legitimate memory usage that's not leaked | 07:39 |
| slangasek | just apparently... not garbage-collected until exit | 07:40 |
| slangasek | maybe there's per-thread memory usage and you have an excessive number of threads configured? | 07:40 |
| slangasek | (combined with threads are going stale and not being reused correctly?) | 07:40 |
| nibalizer | looking at it makes it look like only 4 threads run | 07:47 |
| nibalizer | which makes sense because the box is a 2 core with HT | 07:47 |
| nibalizer | oshi! | 07:48 |
| nibalizer | you may be on to something | 07:48 |
| nibalizer | how do i see threads? do they show in the process table? | 07:48 |
| nibalizer | pgrep only shows one slapd | 07:48 |
| nibalizer | but htop shows about 12 slapds all with 9% of the memory and differt pids | 07:49 |
| nibalizer | here we go | 07:57 |
| nibalizer | root@citadelofadun:~# ps -e -T | grep slap | wc -l | 07:57 |
| nibalizer | 14 | 07:57 |
| slangasek | nibalizer: right - is that number going up over time? | 14:45 |
| blkperl | slangasek: its 18 now | 16:00 |
| slangasek | blkperl: so I'm not finding, in a quick doc search, any info about configuring the size of slapd's thread pool; but if you see the thread count continuing to rise, I'd look at something being amiss there | 16:05 |
| bkerensa | mmm | 18:54 |
| nibalizer | holding at 18 | 20:47 |
| === slangase` is now known as slangasek | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
