RESOLUTION: LogMiner Hanging

I found a resolution to my LogMiner hanging issue. I was using the data dictionary in my Logical Standby database for mining the logs since that database receives the Production archived logs. My Logical Standby database is configured to skip some of the Production schemas, so its data dictionary is not in sync with Production regarding those schemas. I guess that, occasionally, LogMiner encountered an inconsistency that it couldn't get past. I built a new dictionary flat file from Production and used that instead of the Logical Standby data dictionary and LogMiner was able to mine all of the archived logs successfully. I would have expected LogMiner to either just skip over the redo records that it couldn't handle or at least report an error rather than hang. I've reported this to Oracle Support.

Regards,
Mark Strickland

• Forwarded message ---------- From: Mark Strickland <strickland.mark_at_gmail.com> Date: Apr 20, 2007 8:17 AM Subject: LogMiner Hanging To: oracle-l_at_freelists.org

Oracle 10.1.0.5 RAC, DataGuard, RMAN; Solaris 9; Veritas, Hitachi SAN

I recently implemented LogMiner to use as a low-impact just-in-case auditing tool and to get exact transaction counts for month-end reporting. Every morning at 2:00 a.m., I mine the previous day's Production logs and my script produces an HTML report for each hour of the day showing all the transactions for that hour. Groovy, EXCEPT, over the last 19 days that I've been mining the logs, there have been 3 days where there was at least one log that caused LogMiner to hang, never to be heard from again. If I don't include SQL_REDO and SQL_UNDO in the SELECT...FROM V$LOGMNR_CONTENTS, the log can be read. If I include either of those columns, LogMiner hangs and I have to snipe the SQL*Minus session and its associated shadow process. At least I can get transaction counts, but I also want the SQL. I searched the usual haunts online and opened an SR. Oracle Support has been able to duplicate the problem using one of the logs that I sent them. Interesting thing is that those same logs are mined successfully by a Logical Standby database beforehand. Just wondering if anyone out that has had a similar experience and if there was a cause/resolution. Thx!

Regards,
Mark Strickland
Seattle, WA

--

http://www.freelists.org/webpage/oracle-l Received on Mon Apr 23 2007 - 11:48:53 CDT