High average queue size

[José Enrique González]

Hello, sirs:

Attached file is a collect of sar -d (only average finals are showed) for a server connected to an EMC Symmetrix Array, running for about 1 hour. EMC PowerPath is installed. Almost all disk showed in the list are 71 GB.

Disks c5t5d5 and c5t5d6 are consistenly high in average queue size, even when they don't seem to be most I/O requested ones. From this, I have two questions I hope you can help me with:

1. May be it is because, for those disks, the writes are quite greater than reads, as the column in the reports (r+w/s) shows the sum of both values?

2. Could an adjustment to scsi_max_qdepth improve disk I/O in this case?

Any other detail you can discover in thisn report may be helpfull for me. Thanks in advance.

Jose Enrique

[Alzhy]

What do you have carved up on those 2 disks - t5 and 56? Are you using LVM or VxVM? Can you send a layout of the lvol or vol those disks are part of?

Your queue values are unacceptably very high indeed.

[José Enrique González]

I am using LVM. All disks in controllers c5 and c7 are Symmetrix disks hsoting raw devices for Oracle database.