Interpretting iostat

[Red Squirrel]

I'm interested in seeing how my raid performs but I'm really not sure how I should be interpretting any of this data. I have an idea of what the values mean. Like read/sec, but overall just not sure what is considered good, and what is not. Anyone know of a good resources to know this, or perhaps a better program I should be using?

Here's a sample I just took from my server:

Device:         rrqm/s   wrqm/s     r/s     w/s   rsec/s   wsec/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.96     1.34    1.77    2.99   106.43    34.67    29.63     0.02    4.96   1.37   0.65
sda1              0.00     0.00    0.00    0.00     0.00     0.00    11.97     0.00    7.94   7.81   0.00
sda2              0.67     0.22    0.49    0.27     9.23     3.86    17.34     0.01    7.02   3.26   0.25
sda3              0.29     1.12    1.28    2.72    97.20    30.81    31.95     0.02    4.57   1.13   0.45
sdb              36.08   254.74   26.17   35.49  1649.77  2349.18    64.86     1.71   27.03   1.94  11.96
sdc              36.41   254.12   26.16   35.48  1652.21  2344.97    64.85     1.78   28.17   1.96  12.11
sdd              36.85   249.54   25.89   34.38  1653.57  2298.01    65.56     1.82   29.41   2.00  12.03
sde              35.08   256.07   25.93   29.62  1639.66  2316.24    71.21     1.93   69.72   2.44  13.53
sdf              35.73   254.82   26.16   29.61  1646.61  2305.78    70.87     0.21    2.84   2.53  14.13
md0               0.00     0.00   95.99 1068.48  6515.42  8547.81    12.94     0.00    0.00   0.00   0.00
sdg               0.01   226.35    0.27    4.55     2.23  1848.06   384.08     1.73  352.38  26.38  12.71
sdg1              0.01   226.35    0.27    4.55     2.23  1848.06   384.08     1.73  352.38  26.38  12.71
sdi               0.00     0.00    0.00    0.00     0.00     0.00    11.02     0.00    2.98   1.90   0.00
sdi1              0.00     0.00    0.00    0.00     0.00     0.00     9.86     0.00    1.89   0.96   0.00

 

[Relativistic]

It can be used to to find out if a device is bottlenecking the entire array. The numbers to look for are await and svctm. In your case everything looks normal.

Type man iostat for more information.

If you really want to benchmark the array you have to run a dd or bonnie++ test.

 

[Red Squirrel]

What type of numbers are considered bad for await? The sdg and sdg1 kinda catch my attention, but I'm really not sure if those numbers are bad or not.

 

[niomosy]

await
The average time (in milliseconds) for I/O requests
issued to the device to be served. This includes the time
spent by the requests in queue and the time spent servic-
ing them.


I'd say that's up to you or others to determine. If things seem slow or you're getting complaints that the system/application is "slow" then it's a problem. No complaints? No problem.