1748053 Members
4907 Online
108758 Solutions
New Discussion

Disk util

 
laub1216
New Member

Disk util

Hi,

 

In one hp-ux box disk utilization is showing consistenly 100%(both average and current) using glance command.

 

Using sar -d 5 5, I have found some of the disks which are having higher value for avser and avwait.

 

Now I am not sure how to solve the problem of disk util after finding disks with higher avser and avwait.

 

I mean do we send this this output of storage admin or are there any commands to look at other system parameters to further diagnose/solve the problem.

 

 

 

 

1 REPLY 1
Ken Grabowski
Respected Contributor

Re: Disk util

 

The Glance application is giving you the counter GBL_DISK_UTIL_PEAK which is closer aligned with the sar -d column %busy. The avser is the average time in milliseconds it takes to service a data transfer from the disk and the avwait is the average time a request has to wait to be services, again in milliseconds. See man sar.

 

If your average service is not much higher than other disks and your average wait is 0 or near 0, then you may not have a problem. Even if 100% busy, if you are not experiencing long service times of any wait time then your requests are being successfully serviced.  Fully utilized is not always a sign of a problem.

 

I would suggest you first use glance to find the process that has high "Disk IO Rates". Select that process, select OpenFiles. Match those files up with the disk under them to match your glance/sar disks with heavy usage. If you have the X-Window Glance you can drill down to the individual disk activity that way too.

 

Now that you have the application and disk matched up you can check with the application owner to see if there is an issue on their end. Runway processes, loop processes, poorly designed database queries, and just very busy databases can all cause this kind of activity on the disk. Do they think they have a problem?

 

If this is a SAN based disk, you can also pass that information to your storage team to determine if they have any indications of high usage or problems in the fabric or on their array(s).