Find if the disk is asked for more than it can deliver

[labadie_1]

Hello

I would like to find a command similar to the Vms command

$ monitor disk/item=queue

See for the doc
http://www.pi-net.dyndns.org/docs/openvms0732/732final/6491/6491pro_009.html

This shows if there is a queue length, if when an I/O is done on a disk, the info is delivered immediately, or after a (short or long wait).

For example, a slow disk will be able to deliver 60 I/O (of 16 K) /second, when a Raid 5 of 5 fast disks will be able to deliver much more.

When this Vms command shows more than 1 for the queue length, I know that the users issuing the I/O will wait, as the disk is asked to do more than it can.

I have found that
iostat -x 10 10
, for example shows me the I/O on the "disks".

How do I find if there is a queue length on a disk ?

Thanks for any hint.

[Peter Godron]

Hi,
sar
or
sar -q ?

[Victor Semaska_3 · Accepted solution]

Hi,

In the 'iostat -x' output is the column avgqu-sz. According to the manpage:

avgqu-sz
The average queue length of the requests that were issued to the device.

Isn't that what you're looking for?

Vic

[labadie_1]

yes it seems that's it !

Thanks a lot.

[labadie_1]

In fact one of the 3
avgqu-sz await svctm
is want I want.

And all 3 are interesting.