RE: HP-UX 11 kernal parameters for Oracle

From: Jesse, Rich <Rich.Jesse_at_quadtechworld.com>
Date: Thu, 5 Aug 2004 12:21:17 -0500
Message-ID: <FBE1FCA40ECAD41180400050DA2BC54004E93C10@qtiexch2.qgraph.com>

"It depends". First, I highly recommend you get a license from HP for "GlancePlus". It's top, sar, vmstat, and a whole lot more rolled into one easy-to-use package. Of course if you've already bumped into a resource problem, you're probably not going to be able to fire it up to see what it is. In the meantime, have you looked at /var/adm/syslog/syslog and the output from "dmesg" lately? These might give you a better idea of what you're up against.

Also, check to make sure the value of kernel parm "nproc" is high enough -- that's a lotta Oracle processes for a 4GB machine!

Here's a list of our kernel changes for an 8GB HPUX 11.11 box running 2 9i instances and 2 8i instances totaling about 400 Oracle processes. Note that some of the parms are purposely high, as there's other stuff running on this box. Some are low, as we went with the SAM recommended value instead of the Oracle recommended value, solely based on our experience with our server load. Sorry if the formatting is icky -- I haven't figured out how to send HTML tables via MS Lookout! yet...

KERNEL          ORIG            NEW
fs_async        0               MUST BE 0 FOR ORACLE!
ksi_alloc_max   2208            32928
max_thread_proc 64              256
maxdsiz         268435456       1073741824
maxdsiz_64      1073741824      2147483648
maxssiz         8388608         134217728
maxssiz_64bit   1073741824      1073741824 (N/C)
maxswapchunks   16384           16384 (N/C)
maxuprc         ?               3704
msgmap          42              42 (N/C)
msgmni          50              4116
msgseg          2048            32767
msgtql          40              4116
ncallout        515             7235
ncsize          7876            12196 (LOW)
nfile           7750            14662 (LOW)
nflocks         200             4096
ninode          2756            7076 (LOW)
nkthread        499             7219
nproc           276             4116
semmap          66              4098
semmni          64              4096
semmns          128             4096
semmnu          30              30
semvmx          32767           32768
shmmax          4294967296      4294967296 (N/C)
shmmni          200             512
shmseg          120             120 (N/C)
vps_ceiling     16              64

Don't forget to RTFIM (Installation Manual)! All the above are specified in there for running Oracle.

HTH! GL! Rich

Rich Jesse                        System/Database Administrator
rich.jesse_at_quadtechworld.com      QuadTech, Sussex, WI USA

-----Original Message-----

Sent: Thursday, August 05, 2004 11:49 AM Subject: HP-UX 11 kernal parameters for Oracle

We have a development box that is HP-UX B.11.11 9000/800. It has 4gb of memory. It is running anywhere from 5 to 7 8i instances and 3-5 9i instances on any particular day. Once every week it will "lock up". vmstat will show blocked processes and the %wio will go to the high 90s. Nothing crashes or dies, it just appears to have to many processes to run. I have used TOP to look at the processes running. I see nothing unusual. Can anybody share their kernal parameters for HP that they have running 8i and 9i on? Particularly the shmmax, shmmni, ... and so on. Any help will be greatly appreciated.

David Ehresmann PerotSystems

Please see the official ORACLE-L FAQ: http://www.orafaq.com

To unsubscribe send email to: oracle-l-request_at_freelists.org put 'unsubscribe' in the subject line.

--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html


-----------------------------------------------------------------

Received on Thu Aug 05 2004 - 12:17:30 CDT