Huge Volume of Redo generation

[orakan]

Hi Frnds,
I have an interesting issue. One of my prod database archive log frequency had reduced from 2Hrs to 3 min step by step avg since nov. Redo size is 100MB. I am sure my database is not that much hot. I observed the amount of changed blocks for 1hr (as per a metalink doc). It totals to around 50MB only. No direct load etc. Any guesses whatz going into redo log? My last attempt is about to explore log using Logminer. Before going into that, what are the likely causes?
No new index created/rebuilded since nov.

Thanks for you support.

[pareshjavkar]

Have you set archive_lag_target to some value (in your case 2 minutes) ?

Please check.

[orakan]

Hi,
archive_lag_target=0. I removed auditing and AQ

AQ_TM_Processes =0
Transaction_Auditing=FALSE

It helped me little bit. CUrrent Redo frequency is avg 20mins. Still exploring other possible cause.

Thanks

[orakan]

Oracle version is 9.2.0.6

[abhaysk]

do u do lots of index rebuilds?

[orakan]

Hi,
There was no new index creation/rebuild since nov ( I mentioned in the top of the thread).

[abhaysk]

try this?..

select ss.sid, sn.name, ss.value
from v$sesstat ss,
v$statname sn
where sn.statistic# = ss.statistic#
and sn.name in ('redo entries','redo size') ... see for fast increasing values..

[abhaysk]

u say 9.2.0.6.. but anyways look at this.. Bug 2186174 - Excessive REDO generation from parallel rollback

[orakan]

Hi All,
Abhaysk, Thanks for the support. Finally i got the hint from dba_waiters joining v$session and v$sql. It was series of update query updating same table, which was blocking all the sessions in application flow. It has been re-written and now the redo generation back to normal. 3 Hrs/switch average. Logminer too shown the same statement which helped me to narrow down to the issue.

Thanks