Thanks to Jeffrey Beckstrom's sharing of a similar experience, we appear
to have identified a bug and a workaround. I turned off Dead Connection
Detection and the problem automagically went away. The psychotic thread
never showed up in v$process or v$bgprocess, so I must conclude that it
was a DCD thread. For some reason, applying patch #3820881 on a
FailSafe 3.1.2 cluster with Oracle is a bad combination.

Thanks Jeffrey!

Mark Strickland
Seattle, WA

-----Original Message-----
From: Mark Strickland=20
Sent: Tuesday, October 12, 2004 10:02 AM
To: 'oracle-l_at_freelists.org'
Subject: RE: High oracle cpu on Windows 2000 w/ FailSafe 3.1.2 after
patch applied for Security Alert #64

I installed qslice on the server and I can see which thread within the
oracle process is consuming the cpu. However, I can't match it to an
Oracle session within the database. That one thread is consuming 24% of
the cpu right now and the consumption will grow over the next couple of
days until the server is hosed. I searched MetaLink. Here's the query
I'm using:

select to_char(p.spid,'XXXX') "Thread ID",

b.name "Background Process",
s.username "User Name",
s.osuser "OS User",
s.status "STATUS",=20
s.sid "Session ID",=20
s.serial# "Serial No.",=20
s.program "OS Program" =20
from v$process p,=20
v$bgprocess b,=20
v$session s =20
where s.paddr =3D p.addr and=20
b.paddr(+) =3D p.addr;

Any ideas? Perhaps it is a thread related to FailSafe? We may backout
the patch to see if things return to normal. I have an open TAR and the
analyst hasn't reported any hairballs with regards to applying the patch
in a FailSafe environment.



-----Original Message-----
From: Mark Strickland=20
Sent: Friday, October 08, 2004 4:12 PM
To: oracle-l_at_freelists.org
Subject: High oracle cpu on Windows 2000 w/ FailSafe 3.1.2 after patch
applied for Security Alert #64

I'm trying to figure out why the oracle process on a Windows 2000 server
is consuming 98% of the cpu. Patch #3820881 for Oracle Security Alert
#68 was recently applied. This is a 2-node FailSafe 3.1.2 cluster
running Oracle Wait events are all "idle" events for the
Oracle background processes and "SQL*Net message from client" (with one
"to client") for the user sessions. Unfortunately, timed_statistics is
set to false in this database so I can't see which sessions are
consuming cpu. Once I get the server to respond so I can issue a query
in SQL*Plus, query results come back quickly. Navigating within Windows
between SQL*Plus and, say, Task Manager, is glacially slow. Ideas?

Mark Strickland
Seattle, WA


Search Discussions

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
grouporacle-l @
postedOct 15, '04 at 3:52p
activeOct 15, '04 at 3:52p

1 user in discussion

Mark Strickland: 1 post



site design / logo © 2021 Grokbase