How to pre calculate archive logging?

Question

Hello,
We try to initialize a BI workflow load which fills the table BPMWITSTP. The extractor uses insert and update statements. The table looks like this:
MANDT             CLNT     3     0     Mandant
SWW_WIID     NUMC     12     0     Workitem-ID
SWFRMETS     DEC     21     7     Workflow: timestamp methode-einde
Our Q&A machine runs ora 10.2.0.2.0 on Solaris. Our hardware hosting party created a ZFS unix mount for the archlog with a fixed size of 25GB. the log files are written into this directory and eventually are written on tier 4 storage (tapes) when the directory is almost full.
When I start the extractor, the table gets filled so is the archlog. After about 30min the ZFS mount is full and oracles crashes  caused by the limited space of the mount (no archlog could be written  because the tape backup run was too slow).
I do understand the problem with the archlog (no space available) but what i dont understand is how the extractor generates 25gb of logging with simple inserts into this tiny table? We dont have the problem with other extractors which generates a lot more data (setup table for 2LIS eg)
I might be able to get our hosting party to increase the ZFS mount size but first I have to know how much archlogging will be created.
Is there even a way to pre calculate the size of the logging created by an insert statement?
Kind regards,
Edited by: Matthias Gamsjager on Nov 24, 2009 9:43 AM
Edited by: Matthias Gamsjager on Nov 24, 2009 9:51 AM
Edited by: Matthias Gamsjager on Nov 24, 2009 9:52 AM

Answer 1

Hi,
1st: 25 Gbyte for archiving isn't much.
On a productive system it should be several hundreds of Gbytes .
How big are your online redo logs? Check recommendations by SAP
2nd:
if we have a system were  your INSERT process is mostly contributing to your workload we could roughly estimate the amount needed redo logs (be aware that a lot of other things can cause archiving to log files, so I would recommend to check
what else is going on).
What is going into the Archived logs is the redo infomation of your INSERT statement:
"INSERT (MANDT, SWW_WIID,SWFRMETS      )  INTO BPMWITSTP  values ( .... )"
What you should check also is this excessive amount of INSERTS going into the table.
The query below shows you the amount of Archived log space in GBYTE  for a time intervall.
You could correlate that from  the number of inserts done to inserts expected and have a rough estimate
(add to this 30% GBYTES more as a safety buffer).
select to_char(min(completion_time),'HH24:MI:SS DD.MM.YYYY')
comptimeMin,
       to_char(max(completion_time),'HH24:MI:SS DD.MM.YYYY')
comptimeMax,
count(recid) ArchNo,
round(( max(completion_time) - min(completion_time) ) * 24 * 60 ,2)
log_switch_comp_timelag,
round((( max(completion_time) - min(completion_time) ) * 24 * 60),2) / 
count(recid) log_switch_minutes,
round(sum(blocks * block_size) /1024/1024/1024 ,2) GBRedo
from
V$ARCHIVED_LOG
where completion_time between
to_date('24.11.2009 00:00:09','DD.MM.YYYY HH24:MI:SS' ) and
to_date('24.11.2009 23:59:59','DD.MM.YYYY HH24:MI:SS' )
bye
yk