I am having a problem with the system freezing when the system is very busy. I have found the entry oom-killer in our messages log. I would like to confirm that the proper way of dealing with this is to set the sysctl parameter below. Also, the kernel parameter for shmmax and shmall are not correct configured, but another system that appears to be configured correctly has the settings below, do they look right?
Linux Kernel
2.6.18-53.el5
Is this the fix for this problem?
sysctl -w vm.overcommit_memory=2
Proposed ( 16 gigs )
kernel.shmmax=17179869184
kernel.shmall=3774873 x (4096) = 15461879808 .89 = 15461879808/17179869184
Currently ( 16 gigs )
kernel.shmmax = 68719476736
kernel.shmall = 4294967296 x (4096)
vi /var/log/messages
Sep 16 00:11:43 pgprd kernel: postgres invoked oom-killer: gfp_mask=0x201d2, order=0, oomkilladj=0
Sep 16 00:11:43 pgprd kernel:
Sep 16 00:11:43 pgprd kernel: Call Trace:
Sep 16 00:11:43 pgprd kernel: [<ffffffff800bed05>] out_of_memory+0x8e/0x2f5
Sep 16 00:11:43 pgprd kernel: [<ffffffff8000f071>] __alloc_pages+0x22b/0x2b4
Sep 16 00:11:43 pgprd kernel: [<ffffffff80012720>] __do_page_cache_readahead+0x95/0x1d9
Sep 16 00:11:43 pgprd kernel: [<ffffffff800618e1>] __wait_on_bit_lock+0x5b/0x66
Sep 16 00:11:43 pgprd kernel: [<ffffffff8003f11a>] __lock_page+0x5e/0x64
Sep 16 00:11:43 pgprd kernel: [<ffffffff800130ab>] filemap_nopage+0x148/0x322
Sep 16 00:11:43 pgprd kernel: [<ffffffff800087ed>] __handle_mm_fault+0x1f8/0xdf4
Sep 16 00:11:43 pgprd kernel: [<ffffffff80064a6a>] do_page_fault+0x4b8/0x81d
Sep 16 00:11:43 pgprd kernel: [<ffffffff8005bde9>] error_exit+0x0/0x84
Create a cool, new character for your Windows Live™ Messenger.
Check it out
