Hard Drive Performance Monitoring

[itsofmi]

I am looking for the best way to monitor hard drive performance on a variety of different drives. One host may have a SATA drive and another may have a 10 year old EIDE. So what is the best method for checking this?

All of the methods I've tried involved a data rate minimum, which could be different for each drive.

Is there a way to take the average performance of each host's drive over a period of time and check against that?

Thank you!

[swaterhouse]

One of the best ways to monitor disk performance is looking at the disk queue length or the read/write latency.

disk queue length is a count of the io operations waiting for the disk to complete its current operation. A good performing disk subsystem should have a queue length of no more than 2 times the number of disks in the "array". So a raid 5 array with 3 physical disks should have queue lengths of no more than 6.

read/write latency is the number of milliseconds it takes to perfom an operation. This may vary and is a little harder to pinpoint what is a "good" number to stay under.

Unfortunately neither of these counters are available for linux using the zabbix agent but they are both available in windows using performance monitors. Under linux you would need to write a custom script to get the numbers.

[itsofmi]

That seems to work perfectly. Thank you!

[RustedJeep]

This thread is ~10 years old, but still one of the top google results. I am struggling to get disk latency and IO stats from linux servers. Has there been any updates for disk metrics?