Re: Do I really need more than 1 rollback segment?

In <f45d9b0.0112060306.71f78349_at_posting.google.com> andrew_webby_at_hotmail.com (andrew_webby at hotmail) writes:

>You need more than two segments. Your problem is that you need a space
>for the allocated segment to extend, but you've already filled the
>space in the datafile by the other rollback segment. And at a huge
>size which will probably never be used fully.

Ah, I think I see my mistake there.

>No information on the number of concurrent users, average rollback
>size etc, so I'd recommend that *as a starting point* you leave your
>1gb datafile, and drop both rollback segments.

Macimum 2 users, the loging user, and the routine processing user, Not certain how to get information about average rollback size. How can I do this?

>Then, assuming 10 users, create 5 rollback segments with
>5m/5m/minextents 20 and see how that works. Monitor wraps/extends in
>the rollback segments during peak time and drop/recreate at a more
>appropriate initial/nextextent if necessary.

That would be the numbers for initial, and optiomum, right? And the minextnts is a bare number (number of extenst) not size in M, correct?

>Rule of thumb: rollback segments should be as small as you can make
>them, with lots of extents. You don't want them too small that they're
>constantly extending however. Oh, and set an optimal size for them as
>well.

>nb: I know most docs say check v$resource_limit for peak transactions
>and create one rbs each. In my situation, I've found that number to be
>around 180, yet I get by quite happily with only 8 segments with
>2m/2m/20 as the settings. YMMV.

SQL> describe v$resource_limit
Object does not exist.

OK, I clearly don't undertand how to access this. How do I do that?

-- 
"They that would give up essential liberty for temporary safety deserve
neither liberty nor safety."
						-- Benjamin Franklin