Database process/session spins on CPU with MMAP call on /dev/zero
- From: krishna sarabu <ksarabu@xxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx
- Date: Tue, 7 Dec 2010 12:40:24 -0800 (PST)
Hi,
We are experiencing this issue intermittently on random processes and when
it happens the database process basically spins on CPU for few minutes
(sometime it may last for 30+mins). From the strace, it appears that the
process was making MMAP calls on /dev/zero. From ASH, the session appears to be
on CPU, and no latch free or any other wait events involved.
Database version is 11.2.0.1 and the o.s/version is Linux/2.6.18-164.el5.
Please note that this database is using AMM (memory_max_target set to 32G).
Please see the below strace snippet:
--- strace started @ 14:01 ---
Process 1336 attached - interrupt to quit
14:06:42.034264 mmap(0x2ac2e3d86000, 1114112, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_FIXED, 15, 0) = 0x2ac2e3d86000
..
14:25:56.218359 mmap(NULL, 16777216, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_NORESERVE, 15, 0x1f04000) = 0x2ac2e3ef6000
14:25:56.218830 mmap(0x2ac2e3ef6000, 1114112, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_FIXED, 15, 0) = 0x2ac2e3ef6000
14:25:56.220922 mmap(0x2ac2e4006000, 1114112, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_FIXED, 15, 0) = 0x2ac2e4006000
And, from the pmap:
........
00002ac2e3d86000 1088K rwx-- /dev/zero
00002ac2e3ef6000 1088K rwx-- /dev/zero
00002ac2e4006000 1088K rwx-- /dev/zero
........
One option we are looking at to address this issue is to turn-off AMM and use
HugePages instead.
Can anyone please have a look and advice me why/what causing the process to
spin on CPU for MMAP call on /dev/zero in AMM mode and how we can address it?
Thanks for your help in advance.
Regards, Krishna
--
//www.freelists.org/webpage/oracle-l
Other related posts:
- » Database process/session spins on CPU with MMAP call on /dev/zero - krishna sarabu
