[CentOS] oom-killer.

Wed Dec 14 15:19:27 UTC 2005

I use to launch 'vmstat 5' after booting, at the bottom of this lines you have
the anomalie the time i think it happened the oom-killer trap, don't know if it
could be usefull for this, you can see the time the oom-killer 'works' i think.

Regarding to linux configuration, it's a standard intall from RH Centos 4.2
release with updates from CentOS-Base.repo, dag.repo and rpmforge.repo.

It's not a server layout but a workstation in a lan running normal user apps
except rosetta and seti nice tools from BOINC. IT's true that's the first time i
have this issue, not be4 using this tools. Don't know if it's related to the
load caused in this tools because it's not the first time the system is almost
down due to swapping, but it's the first time it calls oom-killer.

No swappiness to zero preset yet, i'm specting a new fault to see if it happends
again, but thx to W.Warren for the hint, im not expert tunning, where can i find
a good reference doc/book/info about tunning Linux/RH EL kernel? I mean,
different schemas to afford different loads?

And yes, i can afford buying 512 MB more, just was thinking that it was almost
correct 512 MB by normal user. Yeah, the bigger, the better.

TIA

Jose

procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in    cs us sy id wa
 0 27 1048568   2364    196  13936    0    0  4831    14 1225   476  0 43  0 57
 0 18 1048368   3980    544  25128   19    0  4638    14 1252   394  1 25  0 74
 0 17 1048568    864    716  25056   64   45  3577   110 1263   538  7 10  0 84
 0 17 1048564   1164    764  22264   44    2  2418   119 1236   523  8  8  0 84
 0 14 1048568    780    344  18912   32    6  3104    41 1297   537 10 14  0 76
 0 16 1048568   2180    320  16424   26    2  3264    55 1370   612  4 20  0 76
 1 15 1048568   2720    280  15620    0    0  4114     2 1298   549 41 15  0 44
 0 15 1048568   1868    224  15840    0    0  4437    13 1334   585 51 21  0 28
 0 24 1048568    824    192  14488   13    1  4936    10 1229   591  0 48  0 52
 2 23 1048568    740    176  14520    6    1  5382     4 1340   637  0 44  0 56
 1 29 1048568    888    152  13024   64   13  5228    20 1347   661  0 56  0 44
 4 32 1048568    880    136  12836    6    2  5538     6 1322   710  0 94  0  6
 8 22 1048568   1224    148  12208    0    0  5903     4 1405   734  0 98  0  2
11 28 1048568    804    140  12620    0    0  6625     3 2044  1043  0 98  0  1
 2  3 904064 235124   3192  53020 3710 12147 211036 13431 93669 43865  9 89  0 2
 1  1 871552 229960   3520  56792 1270    0  2102   176 1263   538 94  6  0  0
 1  1 792312 232580   3528  56792 2540    0  2540    23 1270   531 96  4  0  0
 1  1 713256 234628   3536  56792 2377    0  2377     3 1206   519 97  3  0  0
 1  1 616188 244316   3536  57036 2068    0  2110     0 1231   521 96  4  0  0
 1  1 518356 260900   3544  57036 2140    0  2140    21 1282   491 97  3  0  0
 1  1 417372 272712   3576  57372 1882    0  1938    27 1257   470 97  3  0  0
 3  8 401424 267872   3680  62168 1149    0  2270    18 1285   651 97  3  0  0
 1  3 392584 257272   3844  70820  790    0  2522   140 1290  1112 93  7  0  0
 2  2 392564 246808   4604  74880 1021    0  1978    51 1262   841 94  6  0  0
 1  2 382044 232192   5684  88348  876    0  3783    66 1217   899 95  5  0  0
 1  1 374388 235788   6044  90340  202    0   666    79 1239  2705 91  9  0  0
 4  0 374388 224236   6652  93172   89    0   735    65 1185  1092 94  6  0  0
 1  0 371824 226028   6696  93700  371    0   426    86 1158   873 97  3  0  0
 1  0 371824 225836   6708  93836    2    0    30     5 1212  1035 96  4  0  0
 2  1 371756 225532   6716  93968   43    0    46    16 1264   962 96  4  0  0

William A. Mahaffey III wrote:
> William Warren wrote:
> 
>> set swappiness to zero.  That will force the machine to use all
>> physical ram before swapping for the most part.  Also i have heard
>> that xorg has a memory leak.  I do not know if that has been addressed
>> yet and that could be the cause of your issue.
> 
> 
> 
> Could you elaborate on that (swappiness) just a bit :-) ? Is this a
> generic Linux setting, or RH/CentOS specific ? TIA
> 
> A memory leak *WOULD* cause this kind of thing. SGI had such a problem a
> few years back in some version or other of their (released) IRIX OS, &
> X.org just got back into the game, so that's fairly plausible as well.
> 

-- 
Linux smarteyebox.stigmatedbrain.net 2.6.9-22.0.1.EL i686 GNU/Linux
15:45:02 up 1 day, 15:56, 11 users, load average: 1.38, 1.39, 1.60
---------------------------------------------------------
The best things in life go on sale sooner or later.