Logfile switching

[pando]

true indeed... thinking about checkpoint because karmst mentioned the log interval parameter and the log is not full :D

karmst when u query v$log which column are you looking at to determine that it has switched? Log files doesnt switch unless it´s full or you do it manually.

I would check alert log to see if it´s switching every 3 seconds

[marist89]

No multiplexing will not make a difference.

The rate at which the logs get filled is directly proportional to the number of transactions that have taken place. The only way a logfile will not get filled is if someone issues an "alter system switch logfile".

If you look at v$loghist and post the results of:
select thread#, sequence#, first_change#, to_char(first_time,'mm/dd/yyyy hh24:mi:ss') first_time, switch_change#
from v$loghist
where first_time > to_date('01/02/2001','mm/dd/yyyy')

If you see that the logs are indeed switching every 2-3 seconds, I would then look in my alert.log file. If you are getting a bunch of "Checkpoint not complete" messages, your logs are not big enough and you probably don't have enough of them.

[jmodic]

Keith (karmst),

Obviously there is something extremely odd going on with your instance ;-). Since online redo files are allways the same size, the only way (AFAIK) to be shure how much information is written in the log when log switch occur is to look at archived redo logs. Archived logs are not simply a coppy of online logs - their size reflects the amount of data actually written in the redo log when log switch occured. That is why I asked you if you have archiving turned on.

If you run in noarchive mode then just for testing purposes set the database in archivelog mode and switch automatic archiving on. Then observe the sizes of archived logs. Are they of uniform size? Equal in size to online redo logs or smaller?

BTW, I just remembered in some previous versions there were problems when LOG_CHECKPOINT_INTERVAL vas set to 0. It doesn't behaved like it shud, i.e. it didn't disabled time-based log switches. In fact it had the oppositeeffect - it caused extremely frequent checkpoints (checkpoints, not log switches). If you have this parameter set to 0 try it with a very high value instead.

[st2000]

My db is in no archive log mode and yet consistent log switch happens every 14 minutes..
My init.ora parameters related to log..
log_archive_start boolean FALSE
log_archive_trace integer 0
log_buffer integer 163840
log_checkpoint_interval integer 999999999
log_checkpoint_timeout integer 0
log_checkpoints_to_alert boolean FALSE
I tried posing the scenario once last month and I got answers that could not solve this..
looks like this is a better time..
My db is 8.1.6
Thanks

[goodhealth]

Maybe there is a running script that is forcing log switch like hot backup script etc. Can you verify? Also verify if your hot backup is running, provided ARCHIVELOG mode is on, by doing select on v$backup and check for ACTIVE status. Backup also generate stuff for redo logs.

[Edited by goodhealth on 01-03-2001 at 10:23 PM]

[st2000]

Nothing of that sort. (No backups available) Its a new database and is recently furnished..
Thx..

[rsuri]

Are you running Management Server that happens to have its Repository on that database (or any other such repository) ? If yes then that generates a lot of transactions depending upon what the refresh rate of the monitored jobs is.

You may want to look into the v$sessions view at different times to see what sessions (unusual sessions) are there and what are they doing.

In case of Management server, you may want to increase the refresh interval to reduce transaction information.

- Rajeev