How can I monitor disk io?

Question

I'd like to do some general disk io monitoring on a debian linux server. What are the tools I should know about that monitor disk io so I can see if a disk's performance is maxed out or spikes at certain time throughout the day?

Answer 1

For disk I/O trending there are a few options. My personal favorite in the sar command from sysstat. By default, it gives output like this:

09:25:01 AM     CPU     %user     %nice   %system   %iowait    %steal     %idle
09:35:01 AM     all      0.11      0.00      0.01      0.00      0.00     99.88
09:45:01 AM     all      0.12      0.00      0.01      0.00      0.00     99.86
09:55:01 AM     all      0.09      0.00      0.01      0.00      0.00     99.90
10:05:01 AM     all      0.10      0.00      0.01      0.02      0.01     99.86
Average:        all      0.19      0.00      0.02      0.00      0.01     99.78

The %iowait is the time spent waiting on I/O. Using the Debian package, you must enable the stat collector via the /etc/default/sysstat config file after package installation.

To see current utilization broken out by device, you can use the iostat command, also from the sysstat package:

$ iostat -x 1
Linux 3.5.2-x86_64-linode26 (linode)    11/08/2012      _x86_64_        (4 CPU)

avg-cpu:  %user   %nice %system %iowait  %steal   %idle
           0.84    0.00    0.08    1.22    0.07   97.80

Device:         rrqm/s   wrqm/s     r/s     w/s   rsec/s   wsec/s avgrq-sz avgqu-sz   await  svctm  %util
xvda              0.09     1.02    2.58    0.49   112.79    12.11    40.74     0.15   48.56   3.88   1.19
xvdb              1.39     0.43    4.03    1.82    43.33    18.43    10.56     0.66  112.73   1.93   1.13

Some other options that can show disk usage in trending graphs is munin and cacti.

Answer 2

Monitoring Disk Io can be done by multiple tools like the following.

1. Iostat
2. Iozone
3. SAR
4. vmstat

Also some important operating system concepts are very much necessary to comprehend them..read the Linux IO complete tutorial

Answer 3

Have a look at iotop.

Or iodump, if that's more down your way of thinking.

Note: This requires at least kernel 2.6.20 to work.

Answer 4

The best thing to do if you want to see performance on a long period is to use Centreon which is a powerful tools build on Nagios. Centreon make you able to monitor by drawing graphs of resource's performances and lot of other things.

Answer 5

I like dstat. It can show totals and statistics per disk and even md-devices (RAID), also can use colors for better overview:

$ dstat -tdD total,sda,sdb,sdc,md1 60 
----system---- -dsk/total----dsk/sda-----dsk/sdb-----dsk/sdc-----dsk/md1--
     time     | read  writ: read  writ: read  writ: read  writ: read  writ
08-11 22:08:17|3549k  277k: 144k   28k: 851k   62k: 852k   60k:  25k   82k
08-11 22:09:17|  60k  258k:1775B   15k:  13k   63k:  15k   60k:  68B   74k
08-11 22:10:17| 176k  499k:   0    14k:  41k  122k:  41k  125k: 273B  157k
08-11 22:11:17|  42k  230k:   0    14k:9830B   54k:  14k   51k:   0    70k
08-11 22:11:52|  28k  132k:   0  5032B:5266B   33k:9479B   28k:   0    37k

• -t for timestamps
• -d for disk statistics
• -D to specify the exact devices to report

• 60 to average over 60 seconds. The display is updated every second, but only once per 60 seconds a new line will be started.

• not used in this example, but -c can report wait IO percentage, which in most cases is related to the CPU waiting for data from the disks.

It is available for most Linux distributions, but sometimes needs to be installed from repositories.

Answer 6

Just from files on your computer (assuming a recent kernel) you can ask /sys/block/sda/stat or /proc/diskstats. It will need some translating, however. But it's nice for a quick and dirty check.