Archivelog mode - performance problems anyone??

[Axr2]

Has anyone out there experienced any performance problems solely by putting a database in an archivelog mode? I'd like you to share your thoughts and suggestions on what to watch out for. (buffer sizes and the like..)

I have a 800G database that has 25 log groups (200M each, 1 member/group). During peak periods, it switches about 25-30 redo logs per minute. (Yes, I've seen it happen). I'm not sure if adding more members will entirely help the cause..not that I have a whole lot of disk to spare.

Yes - I'm aware of the database hanging if there isn't enough space in the archlog dir and the fact that archivers sometimes can't cope with the speed of log switches. The latter causes a hang as well. Besides these 2 obvious things, please post if you have anything more to share.

[marist89]

During peak periods, my 128M logs switch every 2-3 minutes. I have multiple archivers setup to a very fast filesystem. I constantly monitor the log_archive_dest and do an archivelog backup & purge when it gets to 80% full or every X hours, whichever comes first. I have never had a performance problem with archivelog mode.

[Axr2]

Had a typo in there..the logfiles aren't 20M, but 200M!

[DaPi]

Let's get this straight - you're generating 100MB of redo PER SECOND ?!

[Axr2]

Jeff - Thanks. Yeah, the LOG_ARCHIVE_MAX_PROCESSES is currently set to 2. I assume Oracle will automatically spawn more ARCn processes if the load on the default ARC0,ARC1 processes increases dramatically. Correct?

Given a choice would you place your logfiles or your archlogfiles to the fast file system. I would assume the former. In your case, I guess you have multiple fast disks.

[Axr2]

DaPi - yes. I have a new incarnation of the database, else wd've posted contents from v$log_history.

[Dilippatel]

Performance is not a problem for Machine with multiple CPU and Faster Disks. Benifit of Archivelog is known to all. Please put Online redo logs and Archivelog on separate disks to minimize I/O contention.

[rajabalachandra]

Originally posted by Axr2
I have a 800G database that has 25 log groups (200M each, 1 member/group). During peak periods, it switches about 25-30 redo logs per minute. (Yes, I've seen it happen). I'm not sure if adding more members will entirely help the cause..not that I have a whole lot of disk to spare.

You have 1 log member per group.
Moreover, why don't you multiplex your log files?

-Raja

[tamilselvan]

Are the disks in which archived files are stored stripped?

The file system is on 8-Way stripped disks using 4 MB Meta volume at the hardware level.
My redo log file size is 500m, and I can able to write 1.5 GB ( 3 archived files) in 1 Min. The entire disk array size is 27 TB. And the array has 32 GB cache.

Tamil

[nagarjuna]

axr,

man, you are enjoying.. with such a huge environment. I wish I could work there..

increasing the log file size may help you. We have done it. One of our oracle apps env. used to switch 3-5 log files of size 30m each every minute during peak hours.. we increased the logfile size to 100m and it helped us.

one more tip you can follow. make sure that your arcive log destination does not contain any other files as this can avoid fragmentation of disk and help the performance.