Setting oom_adj on linux?
in [PgSql]
|
Prev: [HACKERS] Setting oom_adj on linux?
Next: [HACKERS] What's the current theory about derived files in VPATH builds?
From: Tom Lane on 8 Jan 2010 12:56 Robert Haas <robertmhaas(a)gmail.com> writes: > There could well be moving parts if the user wants to adjust the value > being written to oom_adj, and can't because it's compiled in. I don't > see why we can't just add a GUC for this and be done with it. The number of users who will want to do that might be different from epsilon, but not by enough to justify a GUC. Furthermore, we don't have any reasonable infrastructure for supporting platform-specific GUCs, which means that the amount of effort you're proposing is extremely large. regards, tom lane -- Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
From: Alex Hunsaker on 8 Jan 2010 14:30 On Fri, Jan 8, 2010 at 10:24, Stephen Frost <sfrost(a)snowman.net> wrote: > As I recall, oom_adj wasn't visible in the container because you > explicitly set what proc entries processes can have access to when using > VServers, and OpenSSH didn't handle that cleanly. Â Guess what I'm just > saying is "don't just assume everything is as it would be on a 'normal' > system when dealing with /proc and friends", and, of course, test, test, > test when you're talking about back-porting things. Sure this was openssh? I just looked through the entire cvs history for opensshp and found 0 references to 'oom' let alone 'oom_adj'. Maybe something distro specific? [ for the curious here is what I tried ] $ git clone git://git.infradead.org/openssh.git $ cd openssh $ git grep oom_adj $ git grep 'oom' $ git grep oom | grep -vi loomis | grep -v room $ git log -p | grep oom | grep -vi loomis | grep -v room | grep -v tsoome -- Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
From: Tom Lane on 8 Jan 2010 14:45 Alex Hunsaker <badalex(a)gmail.com> writes: > Sure this was openssh? I just looked through the entire cvs history > for opensshp and found 0 references to 'oom' let alone 'oom_adj'. > Maybe something distro specific? FWIW, I see no evidence that sshd on Fedora does anything to change its oom score --- the oom_adj file reads as zero for both the parent daemon and its children. Kinda scary to realize the OOM killer could easily lock me out of boxes I run headless. regards, tom lane -- Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
From: Alex Hunsaker on 8 Jan 2010 15:22 On Fri, Jan 8, 2010 at 12:45, Tom Lane <tgl(a)sss.pgh.pa.us> wrote: > Alex Hunsaker <badalex(a)gmail.com> writes: >> Sure this was openssh? I just looked through the entire cvs history >> for opensshp and found 0 references to 'oom' let alone 'oom_adj'. >> Maybe something distro specific? > > FWIW, I see no evidence that sshd on Fedora does anything to change its > oom score --- the oom_adj file reads as zero for both the parent daemon > and its children. Â Kinda scary to realize the OOM killer could easily > lock me out of boxes I run headless. [ OT, CC trimmed] Yeah, for me sshd has a score of 24 and has the 13th lowest oom_score on my box. While postgres with 0 connections and shared_buffers = 28MB has a score of 26558 and has the 5th highest oom_score. Only chromium and xmonad are above it. With 5 connections it just about doubles its score (to 47238). Had I been headless postgres would certainly die on oom. But even something like alsamixer or bash is higher than sshd *shrug*. For the curious here below is the raw data and how i generated it [ yes its far from perfect... ]: (for file in /proc/*/; do echo `cat $file/oom_score` `cat $file/cmdline` $file; done) | sort -n 6 /sbin/agetty-838400tty1linux /proc/2204/ 6 /sbin/agetty-838400tty2linux /proc/1977/ 6 /sbin/agetty-838400tty3linux /proc/1978/ 6 /usr/sbin/crond /proc/1954/ 6 /usr/sbin/uptimed /proc/1971/ 10 /usr/sbin/irqbalance /proc/1951/ 12 /usr/sbin/ntpd-s /proc/1967/ 12 /usr/sbin/smartd /proc/2128/ 13 hald-addon-input: Listening on /dev/input/event2 /dev/input/event1 /proc/2044/ 13 hald-addon-storage: polling /dev/sr0 (every 2 sec) /proc/2088/ 13 /usr/lib/hal/hald-addon-cpufreq /proc/2092/ 19 /usr/sbin/syslog-ng /proc/1639/ 24 /usr/sbin/sshd /proc/1943/ 26 /usr/lib/sa/sadc-F-L-SDISK6006- /proc/2716/ 27 supervising syslog-ng /proc/1638/ 38 hald-runner /proc/1992/ 53 cat/proc/self//cmdline /proc/self/ 64 /usr/lib/postfix/master /proc/2089/ 90 /usr/bin/X-nolistentcp /proc/2173/ 102 -bash /proc/10997/ 108 /home/alex/.cabal/bin/xmobar-x1 /proc/2190/ 140 /usr/bin/dbus-daemon--fork--print-pid5--print-address7--session /proc/2256/ 140 /usr/bin/dbus-daemon--system /proc/1983/ 191 /usr/lib/hal/hald-addon-acpi /proc/2093/ 194 dbus-launch--autolaunch004d7f457c373938f22d796a4ae05b60--binary-syntax--close-stderr /proc/2255/ 199 /usr/sbin/ntpd-s /proc/1960/ 206 ssh-agentxmonad /proc/2188/ 287 tail-f/var/log/httpd/error.log /proc/8688/ 310 firefox /proc/18571/ 339 xbindkeys /proc/2192/ 341 -bash /proc/10589/ 354 /usr/local/bin/cmus /proc/2205/ 400 /bin/sh/usr/bin/startx /proc/2155/ 456 sort-n /proc/10998/ 487 /usr/sbin/hald /proc/1991/ 525 qmgr-l-tfifo-u /proc/2097/ 639 /bin/sh/home/alex/.xinitrc /proc/2180/ 736 /usr/lib/GConf/gconfd-2 /proc/18578/ 861 tail-f/var/log/httpd/error.log /proc/1621/ 1122 urxvtd-q-f-o /proc/2189/ 1151 /usr/lib/chromium/chromium /proc/2220/ 1172 -bash /proc/22276/ 1351 mutt-y /proc/27637/ 1497 alsamixer /proc/3528/ 1528 ssh192.168.0.15 /proc/1523/ 1534 -bash /proc/2863/ 1534 -bash /proc/2881/ 1534 -bash /proc/2891/ 1793 /usr/lib/chromium/chromium /proc/2219/ 1828 ssh70.98.186.4 /proc/2860/ 2066 pickup-l-tfifo-u /proc/26254/ 2195 postgres: stats collector process /proc/10602/ 2281 -bash /proc/3521/ 2299 -bash /proc/1516/ 2447 -bash /proc/23990/ 2573 -bash /proc/1538/ 3082 /usr/lib/chromium/chromium --channel=2219.aa9eb00.196295212 --type=renderer --lang=en-US --force-fieldtest=AsyncSlowStart/_AsyncSlowStart_off/CacheSize/CacheSizeGroup_6/DnsImpact/_max_500ms_queue_prefetch/GlobalSdch/_global_disable_sdch/SocketLateBinding/_enable_late_binding/ /proc/2392/ 3305 -bash /proc/20490/ 3796 /usr/sbin/httpd-kstart /proc/1553/ 4697 /usr/bin/knotify4 /proc/31084/ 6117 /usr/lib/chromium/chromium --channel=2219.afcb450.251437212 --type=renderer --lang=en-US --force-fieldtest=AsyncSlowStart/_AsyncSlowStart_off/CacheSize/CacheSizeGroup_6/DnsImpact/_max_500ms_queue_prefetch/GlobalSdch/_global_disable_sdch/SocketLateBinding/_enable_late_binding/ /proc/18311/ 6564 /proc/self/exe--channel=2219.a684ce8.115295409--type=extension--lang=en-US--force-fieldtest=AsyncSlowStart/_AsyncSlowStart_off/DnsImpact/_max_500ms_queue_prefetch/GlobalSdch/_global_disable_sdch/SocketLateBinding/_enable_late_binding/ /proc/2242/ 6921 /proc/self/exe--channel=2219.a6862e0.261342415--type=extension--lang=en-US--force-fieldtest=AsyncSlowStart/_AsyncSlowStart_off/DnsImpact/_max_500ms_queue_prefetch/GlobalSdch/_global_disable_sdch/SocketLateBinding/_enable_late_binding/ /proc/2245/ 7453 /proc/self/exe--type=plugin--plugin-path=/usr/lib/mozilla/plugins/libflashplayer.so--lang=en-US--plugin-data-dir=/home/alex/.config/chromium/Default--channel=2219.afd16a58.340779570 /proc/18316/ 10132 -bash /proc/2195/ 10154 postgres: wal writer process /proc/10600/ 10154 postgres: writer process /proc/10599/ 10299 postgres: autovacuum launcher process /proc/10601/ 11019 init [3] /proc/1/ 17332 /usr/sbin/httpd-kstart /proc/1616/ 17365 /usr/sbin/httpd-kstart /proc/1611/ 17366 /usr/sbin/httpd-kstart /proc/1612/ 17398 /usr/sbin/httpd-kstart /proc/1613/ 17567 /usr/sbin/httpd-kstart /proc/1614/ 19954 xinit/home/alex/.xinitrc--/etc/X11/xinit/xserverrc:0-auth/tmp/serverauth.NHLLZS74xg /proc/2172/ 26558 bin/postgres-Dblah /proc/10597/ 26882 /proc/self/exe--channel=2219.b1692d18.1483224477--type=extension--lang=en-US--force-fieldtest=AsyncSlowStart/_AsyncSlowStart_off/CacheSize/CacheSizeGroup_6/DnsImpact/_max_500ms_queue_prefetch/GlobalSdch/_global_disable_sdch/SocketLateBinding/_enable_late_binding/ /proc/1597/ 52201 /usr/lib/chromium/chromium--type=zygote /proc/2221/ 53839 /home/alex/.xmonad/xmonad-i386-linux /proc/2187/ 61530 /usr/lib/chromium/chromium --channel=2219.aa993d0.1197730980 --type=renderer --lang=en-US --force-fieldtest=AsyncSlowStart/_AsyncSlowStart_off/CacheSize/CacheSizeGroup_6/DnsImpact/_max_500ms_queue_prefetch/GlobalSdch/_global_disable_sdch/SocketLateBinding/_enable_late_binding/ /proc/1527/ -- Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
From: Stephen Frost on 8 Jan 2010 17:24
* Tom Lane (tgl(a)sss.pgh.pa.us) wrote: > Alex Hunsaker <badalex(a)gmail.com> writes: > > Sure this was openssh? I just looked through the entire cvs history > > for opensshp and found 0 references to 'oom' let alone 'oom_adj'. > > Maybe something distro specific? > > FWIW, I see no evidence that sshd on Fedora does anything to change its > oom score --- the oom_adj file reads as zero for both the parent daemon > and its children. Kinda scary to realize the OOM killer could easily > lock me out of boxes I run headless. There were a few issues, as it turns out, the particularly annoying one was in the init script which caused upgrades to fail due to sshd not being restarted, bug report here: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=473573 The other issue was with a Debian-specific patch which was applied to OpenSSH which basically just created noise in the log file, bug report here: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=487325 In the end, the problem was with errors being returned from attempts to modify oom_adj. As long as we can just ignore those hopefully there won't be any issues. Thanks, Stephen |