Re: Database talking to itself?

Take a few snapshots of v$filestat shortly after startup and tell us if there is anything significant about the tablespaces that
Oracle is doing its I/O to.

(The 'obvious' guess is that tha some very large objects with lots of extents have been dropped and smon is going crazy trying to coalesce thousands of free extents)./

--

Jonathan Lewis
Yet another Oracle-related web site:  http://www.jlcomp.demon.co.uk

Jim Kuschill wrote in message <8ak658$7ak$1_at_slb1.atl.mindspring.net>...


>We've got a pretty good sized DB (400GB) that seems to have gone senile.

>

>Three days ago the SMON process started up and starting using large amounts

>of CPU and doing some I/O. Soon afterward the LGWR process started doing

 its


>thing. Both ran and ran and ran and are still running. We did a shutdown

>abort, fired it back up and off it went again. It generates about 150MB of

>rollback information every 3 minutes. It appears to be getting and updating

>huge numbers of data blocks but I/O counts at the OS level don't seem to

>verify the numbers in the DB (maybe its updating the same block over and

>over?).

>

>We can execute most queries against the database and direct sqlldr loads

>work. Some queries and (we think) all updates block on something.

>

>Oracle claims it must be doing a huge rollback but we are unaware of any

>transaction that might have been running at the time that would have been

 of


>any size. There are no unusual messages being generated during the

>processing. Oracle claim's we should let it go but haven't been able to

>provide us any concrete way to identify that it is making any progress.

>

>We're running Oracle 8i 8.1.5.0.0 (I think) under HP-UX 11.00.

>

>I'm about to hit the tapes ... anybody have a clue what might be going on

 or


>how we might be able to tell what is going on (and ideally when it might

>finish)?

>

>tia

>Jim Kuschill

>VP, MIS

>Frequency Marketing, Inc

>

>