Where all the redo/archive log volume come from?

Question

Hi,
I have a database with a size of around 50G. Thats not very much, but the crazy thing is that this database has a redo log volume per day that is nearly equal the size of the database:
-- Daily View
SQL> SELECT TRUNC(completion_time) AS time, ROUND(SUM(blocks * block_size)/1024/1024,2) AS size_mb, ROUND(SUM(blocks * block_size)/1024/1024/1024,2) AS size_gb FROM v$archived_log GROUP BY TRUNC(completion_time) ORDER BY 1 DESC;
TIME          SIZE_MB    SIZE_GB
2013-02-03    49599.7      48.44
2013-02-02   50104.63      48.93
2013-02-01   54466.36      53.19
2013-01-31   52935.41      51.69
2013-01-30   51596.85      50.39
-- Hourly View
SQL> SELECT TRUNC(completion_time,'HH24') AS time, ROUND(SUM(blocks * block_size)/1024/1024,2) AS size_mb, ROUND(SUM(blocks * block_size)/1024/1024/1024,2) AS size_gb FROM v$archived_log GROUP BY TRUNC(completion_time,'HH24') ORDER BY 1 DESC;
TIME          SIZE_MB    SIZE_GB
2013-02-03 23     160.68        .16
2013-02-03 22     163.73        .16
2013-02-03 21     195.65        .19
2013-02-03 20    4492.44       4.39
2013-02-03 19      176.3        .17
2013-02-03 18     4259.8       4.16
2013-02-03 17     226.99        .22
2013-02-03 16    5465.64       5.34
2013-02-03 15     166.47        .16
2013-02-03 14     165.05        .16
2013-02-03 13     253.33        .25
2013-02-03 12    9568.93       9.34
2013-02-03 11     391.14        .38
2013-02-03 10    9585.36       9.36
2013-02-03 09     160.02        .16
2013-02-03 08      200.1         .2
2013-02-03 07     166.66        .16
2013-02-03 06     267.45        .26
2013-02-03 05     309.26         .3
2013-02-03 04    1486.87       1.45
2013-02-03 03   11106.77      10.85
2013-02-03 02      318.1        .31
2013-02-03 01     147.47        .14
2013-02-03 00     165.47        .16
2013-02-02 23     160.17        .16
2013-02-02 22     159.25        .16
2013-02-02 21      205.9         .2
2013-02-02 20    4473.27       4.37
2013-02-02 19     124.64        .12
2013-02-02 18    4364.69       4.26
2013-02-02 17     295.02        .29
2013-02-02 16    5415.06       5.29
2013-02-02 15     166.22        .16
2013-02-02 14     166.83        .16
2013-02-02 13      84.62        .08
2013-02-02 12    9905.55       9.67
2013-02-02 11     418.31        .41
2013-02-02 10    9548.78       9.32
2013-02-02 09     166.51        .16
2013-02-02 08     204.88         .2
2013-02-02 07     167.37        .16I know that every database change produces redo, but why that much?
Is there a way to investigate further where all the redo information comes from?
Every 2 hours redo is backed up to tape.
There are no significant batches, shaking and shuffling the data around all the time.
Nearly all changes in this database is happening through user inputs.
Database is 11gR2 on Redhat 5.6.
Thanks
941743
Edited by: 941743 on 04.02.2013 15:32
Edited by: 941743 on 04.02.2013 15:33

Answer 1

I am afraid there is no easy and direct way to pull the answer from the data dictionary.
But with some work, you can get to the bottom of this.
This query shows you how much redo is used by the sessions in your system currently:
select a.sid , a.value  from v$sesstat a,  v$statname b
where a.statistic# = b.statistic#
and b.name = 'redo size'Unfortunately, I am not aware of a AWR view (license required!) that corresponds to v$sesstat, so it would not be easy to look back and see which session generated the redo.
One way is to troubleshoot the problem online, using a utility such as snapper- http://blog.tanelpoder.com/2010/03/22/oracle-session-snapper-v3-10/ . See which sessions generate most redo and see what SQL they are running at that time.
The other way is to create your own “repository”. Sample V$SESSION, v$sesstat , v$statname and possibly V$SQL frequently (every few seconds) and record how much redo and what SQLs are running by each session. Once you gather enough data, look which sessions had the biggest “redo size” delta and what SQL were they running at that time.
Iordan Iotzov
http://iiotzov.wordpress.com/