Re: Strange Swapping Issues(?)
Hello, On 04/15/10 20:41, Peter Jeremy wrote: > The out-of-swap hander will kill the largest process so one of your > problems is probably "throwto003". I can't offer any suggestion as > to why the swap_pager_getswapspace() errors continued afterwards. Okay, it was my fault. After huge processes, like "throwto003" are killed, swap usage falls below 3-4% in a second. (I will need to consult the author of that processes.) For your amusement, here is a log excerpt of the situation (it can by easily reproduced on my system). Columns are as follows: - Time - Swap Used (KB) [swapinfo] - Swap Free (KB) [swapinfo] - Number of Processes [ps] - Active Virtual Pages (KB) [vmstat] - Size of Free List (KB) [vmstat] Messages are pounding the logs at 17:16:37: "kernel: swap_pager_getswapspace(12): failed", and at 17:16:39 "throwto003" is finally killed: "kernel: pid 117 (throwto003), uid 1001, was killed: out of swap space", no further fails. 17:15:3081720 4112584 155 2033636 71464 17:15:3181720 4112584 155 2777084 9732 17:15:41439384 3754920 155 2953212 10104 17:15:42493804 3700500 155 3121148 10152 17:15:44730560 3463744 155 3280812 10208 17:15:47817608 3376696 155 3459068 9980 17:15:49988420 3205884 155 3625008 10268 17:15:511182748 3011556 155 3799036 10036 17:15:531326872 2867432 155 3963900 8280 17:15:541445332 2748972 155 3969020 9632 17:15:551457348 2736956 155 4093928 40304 17:15:571598944 2595360 155 4139004 10568 17:15:581646460 2547844 155 4276220 8360 17:15:591689776 2504528 155 4365308 8476 17:16:001882924 2311380 155 4453372 89048 17:16:042049676 2144628 155 4567036 9860 17:16:062219004 1975300 155 4733948 10264 17:16:082389028 1805276 155 4903932 10012 17:16:102558516 1635788 155 5047396 99916 17:16:122735864 1458440 157 5254428 10532 17:16:152917096 1277208 158 5475788 100944 17:16:173103632 1090672 159 5716856 98876 17:16:203231660 962644 159 5929908 40144 17:16:243498256 696048 159 6168504 68192 17:16:283675324 518980 159 6238136 10700 17:16:303861972 332332 159 6414172 8500 17:16:324032132 162172 157 6556588 10324 17:16:334054828 139476 155 6718460 7932 17:16:344176548 17756 155 6719536 9584 17:16:384193604 700 155 6799908 71096 17:16:39100580 4093724 151 1132240 1640856 17:16:4096336 4097968 151 1123620 1637372 17:16:4196332 4097972 153 1189608 1613748 17:16:4296328 4097976 156 1269164 1529760 17:16:4396272 4098032 154 1147628 1602084 Cheers, :g ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"
Re: Strange Swapping Issues(?)
On 2010-Apr-14 06:44:58 +0200, Gabor PALI wrote: >Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed >Apr 14 05:26:45 xxx kernel: pid 7388 (throwto003), uid 1001, was >killed: out of swap space >Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed The out-of-swap hander will kill the largest process so one of your problems is probably "throwto003". I can't offer any suggestion as to why the swap_pager_getswapspace() errors continued afterwards. -- Peter Jeremy pgp0nNZofvEHu.pgp Description: PGP signature
Re: Strange Swapping Issues(?)
Hi Jeremy, On Wed, Apr 14, 2010 at 7:26 AM, Jeremy Chadwick wrote: > The swapinfo command you ran was not run at 05:26 in the morning. It was run a few minutes after. I accidentally got it live :) Well, I was expecting that because I have seen similar message previously in the logs. I think it is unlikely that things suddenly fall below 3% after the kernel has complained about the lack of swap space. Please, correct me, if I am wrong here. > You should probably set up a small script, run via cronjob, that logs > swapinfo -h output to a file somewhere (rotate it if you want via > newsyslog.conf). Great idea, will do it. > You may have something running on the system that spirals out of > control, such as a web board script being pounded to death, or something > that's forking excessively. It is called parallel nightly build of the Glasgow Haskell Compiler :D According to its official documentation, compilation and testing is very intensive, indeed. I am trying to launch the builders in different times in order to distribute the load. > I'd also recommend having the script output "top -b -o res 100", which > will give you the top 100 processes on the machine sorted by RSS > [..] So I'm making the assumption RSS will be large. We will see soon... Thanks for the quick help! :g ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"
Re: Strange Swapping Issues(?)
On Wed, Apr 14, 2010 at 06:44:58AM +0200, Gabor PALI wrote: > Hello there, > > I am running a FreeBSD/amd64 8-STABLE with GENERIC kernel as of > February 17 on my box (quad core, 2 GB RAM), and recently I spot some > interesting problems in my logs. My machine runs two instances of a > client in two separate chroot environments in parallel with 32-bit and > 64-bit userlands respectively, doing a nightly building and testing. > According to the logs it puts a nice load on my system, though it > still seems to be working fine. > > Except one thing: it produces strange error messages on swap space > without an apparent reason (at least to me): > > xxx# tail -f /var/log/messages > Apr 14 05:26:44 xxx kernel: swap_pager_getswapspace(4): failed > Apr 14 05:26:44 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:44 xxx kernel: > Apr 14 05:26:44 xxx kernel: swap_pager_getswapspace(3): failed > Apr 14 05:26:44 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:44 xxx kernel: swap_pager_getswapspace(3): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(12): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(2): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: pid 7388 (throwto003), uid 1001, was > killed: out of swap space > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(8): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(8): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(12): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(9): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(9): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(3): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(16): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(9): failed > Apr 14 05:26:45 xxx kernel: swap_pager_getswapspace(3): failed > ^C > xxx# swapinfo -h > Device 1K-blocks UsedAvail Capacity > /dev/ad0s1b 4194304 112M 3.9G 3% The swapinfo command you ran was not run at 05:26 in the morning. You should probably set up a small script, run via cronjob, that logs swapinfo -h output to a file somewhere (rotate it if you want via newsyslog.conf). You may have something running on the system that spirals out of control, such as a web board script being pounded to death, or something that's forking excessively. I'd also recommend having the script output "top -b -o res 100", which will give you the top 100 processes on the machine sorted by RSS (non-shared) memory usage. I don't know of a way to show "the amount of swap used by process N, or all processes", since it's transparently handled by the VM. So I'm making the assumption RSS will be large. -- | Jeremy Chadwick j...@parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB | ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"