Re: automatic undo mysteries

See the following metalink documents. Do you have a high transaction rate?

10630.1
40689.1
62005.1

http://www.ixora.com.au/tips/admin/ora-1555.htm Also search for ora-1555 on asktom.oracle.com for some in depth discursion.

What exactly do you mean just after starting. Starting the db? SMON might still be rolling back. Starting the query? delayed block cleanout or overwritten transaction slot or a small rollback segment or the OPTIMAL is set and just shot you in the foot.

Remember, the snapshot is needed because there are different views of the database for each transaction - so it could be the other one that "caused" the problem, not the one in the alert log. Could even be something killed - or not killed - "yesterday." (I have a lightly used db with about 7 hours retention that has a nightly kill of sessions that people forget to log out. One night I had to turn the kill off to do something else overnight, and by 10AM the next day, ORA-1555's).

Also check out the undo retention advisor in OEM if you have it. 36000 isn't too terrible, but it depends on your db activity.

jg

-- 
@home.com is bogus.
As always, it depends. http://www.depend.com/