- use topas or nmon instead of vmstat
- put alerts in place to monitor pagingspace - we even do it with a user
defined event in OEM
- I have a 'wacking' script on our develop box that shoots off processes if
they use more than xMb of paging when paging is almost
full - the cool thing is - it sends a wall (something like - KERNEL PANIC
killed process blah blah), so when this
happens, almost immediately we get a concerned developer on the line asking
if he broke something :-) always good for some fun ...
It's also handy as these processes get logged we can have a closer look at
it...
- switch to shared server / MTS if you have a lot of sessions
- reduce your AIX fs cache using vmtune (default is 80% of real memory I
think)
- in the end - add more paging ...
"DBA" <marcelmix_at_yahoo.com> wrote in message
news:3607af43.0404061235.67bb5bbf_at_posting.google.com...
> Last weekend I faced the problem of paging space filing up, wich
> caused the oracle database to be inaccessible, not only the database
> wasn't running my telnet service was down too. After rebooting the
> server everything went back to normal.
> I have about 10 instances running and those are most of all
> development instances so there is not much havy load goign on.
> The server has 8Gigs of Phisical RAM and 4Gigs of Swap space. I've
> been monitoring the server using VMSTS since but I don't see any major
> activity going on. Paging-In is at zero and the CPU is at 98% Unused
> on average.
> So what could cause the server to start paging and halt all the
> services? And how to prevent it from happenig again?
>
> Thank You
Received on Wed Apr 07 2004 - 06:18:23 CDT
