Severe library cache latch contention with DB Control job
Date: Tue, 1 Dec 2009 10:01:06 -0600 (CST)
Message-ID: <63185ad93e657d99c71dfdaba8348803.squirrel_at_society.servebeer.com>
Hey all,
I've noticed a condition twice in 10 days with our 10.1.0.5.0 DB on AIX where library cache latch contention was severe. Each incident I now believe was caused by the EMD_MAINTENANCE.EXECUTE_EM_DBMS_JOB_PROCS job -- a leftover from Database Control that I inherited but replaced with Grid Control over two years ago. Apparently, I was also supposed to kill this job, which I now have done. But two things still bug me.
First, according to Grid Control, this job was launched by SYSMAN but somehow changed it's session owner to *my* username. I haven't experimented with this other than the technique used in exp/imp scripts -- thoughts? Why me?
Second, what could this job have done to cause severe library cache latch contention? It was bad enough the first time to somehow crash the DBMS_STATS job that was running with an "ORA-20000: Unable to analyze TABLE "MYSCHEMA"."MYTABLE", insufficient privileges or does not exist" error.
The job will never run again, but I'm trying to learn as much as I can about it for future reference.
Thanks!
Rich
p.s. No references to this job in the list archives, a Google search turned up a mixed bag of half-complete ideas and a now-non-existent MOS (nee Metalink) article, and of course not much on MOS, either.
-- http://www.freelists.org/webpage/oracle-lReceived on Tue Dec 01 2009 - 10:01:06 CST