vmstat and iostat

[Satish Y]

Hi Experts,

Greetings for the day!

Can you please explain how u analyse vnmstat and iostat output, I mean what values u treat as normal and what r exceptional and what actions needs to be taken if any value reaches its abnormal limit?...
(I am aware of what each column represents)

I found lot of books but they don't explain about situations.

I know it is not possible to explain in brief, but it's OK even each of u answer one situation.

Thanks in advance...

Cheers...
Satish.

[Bernie Vande Griend · Accepted solution]

Your budget probably won't allow it, but Measureware/glance is good for this as you can set up alerts to go off at certain points.

Otherwise, a couple vmstat items to look at:
If PO ever goes above 0 I'd want to know.
If free memory reached a certain point, (depends on amount on the server) I'd want to know. Maybe 25 MB or so.
If CPU Idle remained at >5 I'd also want to know.

[Sridhar Bhaskarla]

Satish,



I don't really look at iostat figures when I think that I have a disk problem. I look at sar -d and sar -b reports. However, you can find the busiest disk in detail using iostat by
looking at bps and sps. You may get a good feel of the response time by looking at msps. I have not seen msps beyond 1ms (1ms is the default). So, I feel anything above 1ms on msps is a bottleneck. However, I would look at sar -d along with %wio with sar -u and then confirm it.

And with vmstat, I would look at "b" (procs blocked for resources), so (swap out), po (page outs). I wouldn't like to see non-zero values there. Also I would look at free. I would be concerned to see a low value there. About cpu, us - user mode, sy - system mode and id- idle idle ~ 0 means cpu constraint. You can get the same info from sar -u.

vmstat -s |grep swap

If there are non-negative values, look at your memory usage.

This is only my 2c.

-Sri

[Sanjay_6]

Hi Satich,

please find some threads below which might help you.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,1150,0x4284dfe5920fd5118fef0090279cd0f9,00.html

http://us-support.external.hp.com/cki/bin/doc.pl/sid=236a449608be7b73d4/screen=ckiDisplayDocument?docId=200000049436468

http://us-support.external.hp.com/cki/bin/doc.pl/sid=236a449608be7b73d4/screen=ckiDisplayDocument?docId=200000050018417

Hope this helps.

Thanks

[Deshpande Prashant]

Hi Satish
The first column (r) in vmstat output gives you CPU run queue. Higher number suggests a CPU bottleneck. This is similar to #sar -q

Also higher number of "po" (page out) and "sr" (scan rate) less "free" memory
suggest memory bottleneck/near bottleneck. But this could be because of higher number of buffers also.

Prashant.