JFS 3.3 /Online JFS and memory issues.

[Anand Murugesan]

I have a strange scenario. We loaded JFS 3.3 and OnlineJFS on one of the servers. The OS level is 11.00, Since we loaded it, the server seem to use memory rapidly to the point it start to get in to thrashing within 24 hours and eventually just turn non-responsive. Has anyone come across a similar scenario? Thanks.

[harry d brown jr]

And you have the machine patched up?

http://www1.itrc.hp.com/service/patch/releasePage.do?BC=patch.breadcrumb.main|patch.breadcrumb.releaseIndexPage|&releaseId=5013-2091

What does

kmtune | grep dbc

return?

and

swapinfo -at

returns?

live free or die
harry d brown jr

[Steven E. Protter]

I've never run into this scenario, though my install was 11i v1. I'm pretty current on patches and would suggest you being so as well, especially ones related to Online JFS.

Also December or Marches quarterly is a good idea.

SEP

[harry d brown jr]

The last bundle is march 04 for 11.00

live free or die
harry d brown jr

[Anand Murugesan]

# kmtune | grep dbc
dbc_max_pct 50
dbc_min_pct 5



# swapinfo -at
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 2621440 0 2621440 0% 0 - 1 /dev/vg00/lvol2
dev 2097152 0 2097152 0% 0 - 0 /dev/vg01/lvol1
reserve - 162960 -162960
memory 1980496 993312 987184 50%
total 6699088 1156272 5542816 17% - 0 -


It is patched up to the jan release.

[A. Clay Stephenson]

The first thing you should do is decrease dbc_max_pct to something like 15%. On 11.00 buffer cache should be no bigger than about 800MB and that is quite generous. I actually prefer a fixed buffer cache enabled by setting bufpages to a non-zero value indicating the number of 4K pages.

[A. Clay Stephenson]

Also, have you installed a Quality Patch Set after you installed OnlineJFS. The behavior you describe was typical of unpatched installs on 11.0. You should still reduce your buffer cache.

[Anand Murugesan]

There seem tobe a known memory leak, the patch required is..... PHKL_32575