Help with Oracle-solaris performance

I am trying to figure out what is going on with our solaris oracle server. If anyone has ANY ideas about this, please send them to me (sjs_at_corp.portal.com)......Thanks.

-steve

last night I observed the following on our oracle server:

when the load hit a certain level, Oracle spent all its time doing something other than productive work. When I truss-ed an oracle shadow process (truss is the Solaris version of trace) - it was doing lots of shmop, sig<mumble>, getmsg, get/set itimer, times - but very little read/writes.

The IMs when trussed whould sleep for a response from Oracle for many seconds, sometimes 10s of seconds.

We lowered the number of client procs talking to the server from 48 from each of
the 3 client machines down to 16 per client machine. This made the system much more robust.

We also observed what appears to be a memory leak in oracle, though I'm not sure what is going on.

(_) ask around to see if Oracle and Solaris/MP have load vs. number

        of shadow process sensitive problems?

(_) ask around to see if there are known memory leaks in Oracle,

        especially around LONG RAW things. Received on Thu Dec 15 1994 - 23:06:23 CET