Analyse vmstat and iostat
Hi,
I am looking at a set of vmstat data and iostat data collected
over a period of a month on a 2cpu + 2 gb server running
iplanet web server. My questions are:
1. The 4th column of vmstat is swap. To be honest I never
really was able to correlate what I understand in school
about swap and what I see in real life. What I am seeing is
that swap available can drop to as little as 328 KB and
move back up to 2 GB and then drop back to few hundred
Ks again. During this period, free mem remains constant
at about 200 MB. Appreciate any advise.
2.
[627 byte] By [
jon_jon888] at [2007-11-25 23:17:35]
# 1
Hi again,
For the second part of my previous unfinished mail,
2. I oftern look for io problems using svc_t column of iostat
looking out for more than 30 ms service time. However
I have often realised that root disks and their mirrors on
many system consistently display up to 100ms of svc_t.
My instinct tell me this has to be wrong but could never
find any documentation explaining why. Any advise is
appreciated.
# 2
Take a look at Memory System, sizing, Tools and Architecture (Sun document)
The 4th field of vmstat shows free and unreserved swap available in Kbytes.
The variations observed therefore depend on the swap freed and reservation. Those two elements are monitored with the pi,po,fr columns (respectively pages paged in, paged out, pages freed).
As for the memory being constant, Solaris use an extra abstraction layer called vnode to deal with memory resources. As a result, "Free memory" isn't reliable as a mean to evaluate memorty shortage.
To detect memory shortage, look the sr and po columns from vmstat. Excessive paging activty will result in constantly non-zero values.
