Glance and raw disk space

[Bob England]

Hello,

I have a question about how Glance reports disk queues for raw disk space. We have an Informix database residing on a raw EVA disk LUN. The database is heavily used and Glance shows extended periods where disk utilization is above 90% but the disk queue is 0. Is this because the raw disk space is managed by Informix and so the disk queue is not used? I have a hard time believing that nothing is being qued up when the disk is that busy.

Any thoughts would be appreciated.

[A. Clay Stephenson]

Glance can be very misleading when trying to apply metrics to array LUN's. Glance doesn't have a clue that this is not, in fact, a single conventional disk so high disk utilization is not surprising. The good news is that the requests are being satisfied very quickly so the queue is very small. I assume that you are seeing high physical i/o rates. The real distinction in using raw character (as opposed to block devices) is that the buffer cache is completely bypassed so that every i/o is a physical i/o.

I have seen many cases where (because of high disk utilization from Glance's point of view) a few large LUN's were divided into many smaller LUN's and indeed, Glance's Disk Utilization was "fixed" but the actual i/o rates were no better.

[Mustafa Gulercan]

hi bob;
#glance
and press
and than press

you can see the disks with device path.

#ioscan -funCdisk
by ioscan you can see the raw device path.


regards;
mustafa

[Bob England]

I found that Informix does indeed manage the io to its raw spaces. By doing an onstat -g ioq command you can look at the Informix statistcs that it keeps for it's internally managed queues.