How to analyze the load on disk subsystems?

A

abbaerro2012-10-28 20:36:00

linux

abbaerro, 2012-10-28 20:36:00

There is a zoo of servers on win, linux performing different functions (DB Server, RDP Serer, Router, telephony, etc.).
There is a Zabbix monitoring system.

The question is how to correctly analyze the load on the disk subsystems of servers (with logging in zabbix)?
The goal is to understand in a timely manner that there are not enough server resources for the task and plugging in the disk (for the CPU there is utilization and average load, for RAM - free / busy).
But with the disk everything is complicated (

Reply

Answer the question

In order to leave comments, you need to log in

4 answer(s)

L

lnx, 2012-10-28
@abbaerro

for linux-machines I can offer such an option .

E

EvilMan, 2012-10-28
@EvilMan

On Linux, see the iowait value in CPU usage statistics. The vmstat, iostat, and iotop utilities will also be useful.

V

vsespb, 2012-10-28
@vsespb

collectd, rrdtool habrahabr.ru/post/93205/

A

Alexander Demchenko, 2012-10-29
@braindamagedman

Again, I'm sorry, but what is the maximum disk performance?
I am not strong in evaluating indicators, but as far as I know, the number of maximum iops, as well as the number of maximum megabits, is highly dependent on blocks / sizes, as well as parallel operations.
Therefore, you cannot run a benchmark, see that raid 10 from sas gives out 300 megabits for writing and 500 iops and say that in a real situation the server will rest on this data.
...

Have you looked here habrahabr.ru/post/154235/ ?