Disk utilization on 9i oracle.

[Jean-Luc Oudart]

It seems there is a lot of datafiles and each oracle checkpoint will update the header :
"Depending on the number of datafiles in a database, a checkpoint can be a
highly resource intensive operation, since all datafile headers are frozen
during the checkpoint. There is a performance trade-off regarding frequency
of checkpoints. "

cf. Metalink doc id 147468.1

what value do you have for (oracle parameters) :
- CHECKPOINT_PROCESS
- LOG_CHECKPOINT_INTERVAL
- LOG_CHECKPOINT_TIMEOUT
- LOG_CHECKPOINTS_TO_ALERT

Regards
Jean-Luc

[Jean-Luc Oudart]

Charlie

as you are running Oracle with EVA backend storage you may want to read this doc
http://h71028.www7.hp.com/ERC/downloads/5983-0322EN.pdf

Regards
Jean-Luc

[Alzhy]

You do have queuing per your sar -d post.

If you're scsi_max_qdepth is 8 or 16, increase it to 32 or even 64 since you are using few LUNs.

scsi_max_qdepth is a dynamic kernel tunable so you need not reboot your box.

[Charles Temple]

what kind of damage or issues can occur by changing that. Right now it is set to 8.

Thanks again for all responses!!!

Charlie

[Alzhy]

There should be none.

[Charles Temple]

CHECKPOINT_PROCESS
- LOG_CHECKPOINT_INTERVAL 100000
- LOG_CHECKPOINT_TIMEOUT 1200
- LOG_CHECKPOINTS_TO_ALERT TRUE

[Alzhy]

Charles,

That explains your queuing and evidently slow performance your app is draining from your EVA. I suppose your lone filesystem is also on just one big EVA LUN or a couple of big EVA LUNs. You really need to bump this value up.

How many environments are hooked up to your EVA? If it has say 8 servers hooked up and each has dual paths.. I can safely recommend bumping up your current 8 to 32 on this server where you are having a problem.

[Charles Temple]

Thanks for quick response. Now you saying the values of Checkpoint? I bumped it up on one of the servers to 32 and disk utilization is still high. We have 6 servers all connecting to the EVA. Yes the filesystems are one BIG LUN. Setup before I got here.
Think I should go higher say 64?

Thanks
Charlie

[Alzhy]

You can certainly try. Each HSV controller can hold up to 1024 queued SCSI operations.

[Eric Antunes]

Nelson:

I have a SureStore E Disk Sys SC10 Factory Rack.

Here are my stats:

09:10:09 device %busy avque r+w/s blks/s avwait avserv
Average c4t8d0 9.67 0.53 14 416 4.94 9.59
Average c4t9d0 6.23 0.50 9 143 5.11 7.83
Average c5t8d0 6.43 0.54 9 329 4.95 11.08
Average c5t9d0 2.03 0.50 2 40 4.93 11.12
Average c1t2d0 1.30 0.50 2 18 3.17 12.23
Average c2t2d0 0.58 0.50 1 14 3.43 7.96

Charles:

You may want to try to decrease your LOG_CHECKPOINT_INTERVAL to 20000, increase the LOG_CHECKPOINT_TIMEOUT to 3600 and switch off LOG_CHECKPOINT_TO_ALERT. Another parameter that influences the checkpoint process is FAST_START_MTTR_TARGET: what is your value for this parameter??

Best Regards,

Eric