Re: Performance Monitoring strategies?

A common mistake with rollback segments is to have a few really large ones and then try to place long running transactions on them. The real need is to have the rollback segments preserve as much transaction history as possible. It is the exhaustion of undo history that normally "blows up", with the ubiquitous ORA-1555, right?

So, the dba needs to peek at the high water mark for the rollback segments, size accordingly and make sure the are all the same size. If you have enough of them, and they are equally sized, the database takes care of load balancing pretty nicely and the 1555 thing generally disappears.

• ricky

Ryan Gaffuri wrote:
>
> We often have problems with rollback segments blowing up. Our DBA then
> needs to go in and essentially increase their size. I know enough
> about being a DBA(which is not much) to know that this probably isnt
> the best way to go about things.
>
> We are not using any GUI tools(such as the OEM)... In your experience
> how have you monitored these kinds of things? My assumption is to have
> jobs run every 2 hours? or so that hit the data dictionary and then I
> would assume insert the results into some tables. Then have some
> queries hit those tables analyzing them for possible issues that may
> arise soon. Then have an alert sent out if they pass some limit.
>
> Am I going in the right direction?
>
> Ryan
Received on Thu Jun 13 2002 - 20:26:30 CDT