Re: Ora 1595 followed by Ora 1575

From: Badri_at_cris.com (BADRI) >>

• In second post on prob. Ora-01575, time-out waiting on space management resource (not exact wording) Badri says >> I do observe that almost every time I get the 1575, I get the 1595 - error freeing extent (nnn) of rollback segment <<

We have suffered the 1575 error with 7.0.15, 7.1.3, and 7.1.6 and I think once or twice with 7.2.3 but we have never been able to tie it to a specific process. The Orace support paper on the subject pretty much point to temp segment management. Since your problem appears to be related to rollback segment management may I suggest that you consider re-allocating your rollback segments. Take a look at how many concurrent updaters you have and I think Oracle recommends allocating one rollback segment for every four concurrent users. I suggest that you set the initial and next extent equal and minextents to a number between 5 and 10. Use the optimal parameter and set it to minextents * extent size. I have never seen any estimates in writing about how much total space should be permanently allocated to rollback segments, but I suggest at least 5% of the total redo generated in one day to minimize 'snapshot too old' from reasonably well written processes.

The 1595 error could be some type of side effect from the 1575, but checking them out will not hurt and may help. Finding the process creating the problem would probably help, but may not be practical. Have you discussed taking an Oracle system state dump when the problem first occurs to see if Oracle support could find the offending process for you? It may be too late once the problem occurs to find the cause via the dump, but it may be a question worth asking.

I hope these ideas help some.

Mark Powell -- Oracle 7 Certified DBA
- The only advice that counts is the advice that you follow so follow your own advice - Received on Fri Apr 10 1998 - 13:42:20 CDT