Disk Usage

[Sreekanth Makam_1]

Hi Folks,
There is 100% Utilization in my disks. Iam using EVA 5000. I am getting Phy IO of 5000.
Now my questions are,
1. What may the problems?
2. What to check?
3. WHat is the exact meaning of PHY_IO and PHY_IO_RATE that was shown by perfView?
4. How to find the current block size in system?

Iam new to disk related problems...Please guide me.


Regards,
Sree

[Con O'Kelly]

Hi Sree

The first step is to identify the processes responsible for the high disk utilisation.

If you have glance installed you can easily identify the filesystems causing the high disk util. Run
# glance -i

Also with gpm you can sort processes by disk utilisation to identify the processes causing the high disk util.

The following command will show the block size (bsize) for the filesystem:
# mkfs -F vxfs -m /dev/vgXX/lvolXX
(Enusre you use -m option!!)

You can check the definition of all glance and perview metrics using the help menu in gpm.

Cheers
Con

[Sridhar Bhaskarla]

Hi Sreekanth,

How big is the disk (LUN)?. In the same glance window where you are seeing 100%, look at the QLEN. Ideally it should be 0. If it more than 0, then that disk is not responding enough or it may be due to large number of requests queued up on it.

Go to the 'logical volume' window and look at the LVs that are heavily used. If you have two or more LVs that are heavily used in this PV, then try moving some of them into another PV. If this is a too big disk, you may need to increase the 'queue_depth' parameter. Look at 'man scsictl' for more information on it.

PHYS_IO are the number of physical IOs during the last interval. A physical IO is an IO from the buffer cache or process to the disk subsystem. PHYS_IO_RATE is the the number of PHYS_IO / interval.

'mkfs -m' or 'fstyp -v' will show you the block size. It varies depending on the filesystem size. 1K is default and it can go upto a max of 8k for VXFS filesystems.

-Sri