Another suggestion, as a first look is to run sysrec (based on Brendan's
sysperfstat) which
records system utilization / saturation for cpu, mem, disk and net. I find this
very useful
and simple to use:
http://www.systemdatarecorder.org/recording/recorders.html
You can try 0.70 which adds supp
> was a default install. What got me into my own direct
> kstat/procfs programming
> was the pain of installing the SE toolkit everywhere just
> to pull out very
> basic stats - like network utilization!
>From many customer installations, a similar message: they kind of backout
>installing yet
Easy
formatting to various reporting systems: RRDTool, Graphite would be a bonus.
Thx,
stefan
stefan
Stefan Parvu
http://www.nbl.fi/stefan.parvu
>
> Brendan
>
> --
> Brendan Gregg, Sun Microsystems Fishworks.
> http://blogs.sun.com/brendan
>
> >>4518644 I/O wait statistic is still misleading
> and should be dropped
+1.
> It really depends on the definition of idl. Is idl defined
> to be the
> rounded idle time or 100 - (sum of the rounded sys +
> rounded usr).
> They will be different (as in the example below). Which is
> corre
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl
0 84 00 392 192 706 101001 3414 10 3 0 86
05 00 308 108 2309000 4221 1 0 97
00 00 308 108 2329000 4161
>
>if (Kernel + User > 100) {
>User = 100 - Kernel;
>}
>Idle = 100 - Kernel - User;
>
> If you want to be smart, you might want to round small
> numbers up, and large numbers down (e.g. perhaps any
> non-zero amount should always show as at least 1%?)
I am working now to man
> Just grab the nsec values. As I recall, the others are just
> lower precision
> numbers derived from them.
using cpu::sys:cpu_nsec_user,kernel,idle
does not fix the problem. I think this is more
an issue about printing the output in Perl using
print or printf and rounding the values.
Example u
> Just grab the nsec values. As I recall, the others are just
> lower precision
> numbers derived from them.
Cheers. These are most likely the goodies from S10+ regarding the new
microstate accounting interface. I will try to experiment and see what
I get.
Thanks for pointers.
stefan
__
Hi,
Thanks for tip. Are you saying I should have fetched in my script
cpu::sys:cpu_ticks_kernel,user,idle as mpstat does ?
I had a version some time ago but with similar results. I checked it
and it seems it reports same bogus information from time to time.
Or are you saying that I should forget
Hi,
I wrote a simple recorder script, similar with mpstat, done in Perl,
using Sun::Solaris::Kstat. The utility is found:
http://www.nbl.fi/~nbl97/solaris/perf/sdr_bin/cpurec062. Based on
Brendan's K9Toolkit, the script uses kstat to extract for each cpu
found on the system the following informat
Hi Dave,
> Yes, but note they're computed values, while the time
> values are
> measured. The measured values for resource usages are
> traditionally
> in units like KB/S or IOPS, and they're hard to turn
> into utilization
> numbers. That leaves the poor sysadmin faced with coming
> up
> for a v
> http://www.teamquest.com/resources/webinars/display/30/index.htm
> Without the www, it goes nowhere (;-))
Many thanks. Interesting.
stefan
___
perf-discuss mailing list
perf-discuss@opensolaris.org
> corestat isn't integrated into Solaris because it
> hasn't been productized and has some issues (eg. must be
> run as root, prevents anyone else from using the CPU
> performance counters while it is running, only works for
> Niagara processors, may not be sufficient for doing capacity
> planning
Hi David,
> The classic tools for doing capacity planning measure
> operation/transaction rate and time taken, rather
> than resource usage.
>
> Once you've *done* the planning, then you correlate
> that with the amount of resources you used and use
> that to predict when you need to add more r
Hi all,
Solaris has a rich number of tools/APIs in order to help to put together a
Capacity Planning model of your servers/site:
- /proc, process statistics
- Kernel Statistics: used by a majority of userland tools to obtain:
1. CPU/Cores utilization (Adrian Cockcroft might disagree with this [
Hey,
Your Java application is a standalone application or does it run on top of an
application
server ?
Couple of things to consider here:
- are you running on 32 or 64 bit JVM ?
- which version of Java are you using: 5 or 6 ?
- you are saying that your application should reduce to maximum th
> 9 zones, currently 15 jvms.
enough big number of jvms. Is your server based on 8 cores with 16 or
32GB ? I would say you should have a big enough swap as well for all
these JVMs and the needed reservations.
We were in a T2000, 8 cores, 32GB RAM with 4 zones, 20-25 jvms. It is
somehow diffic
>> As an experiment you could try running one or more of
>> your Perl scripts with
>>ppgsz -o heap=8K /path/to/script
>> ormance problem disappears. If this performs
>> better, then your problem is very likely what I've
> This fixes it!
>
> No more high smtx values (even w
>- how many processes are you running on this box vs number of CPU ?
Read this: vs. number of cores :=)
We had a kind of similar problem at our office where we simple had too many
java app server instances running on one T2000 box. Hard things comes to
size properly a T2000, how one should do tha
Hi,
when you run your Perl scripts what 'vmstat 1' shows ? As well what exactly are
those perl scripts, are you monitoring your site using them or ?
Couple of other questions:
- are you using ZFS ?
- what disks are you using on T2000 ?
- the problem occurs only when you start your monitoring P
> in the code, and it indeed derived from runq depth (a running, decayed
> average).
as supplement, a good reading for weekend is the article of Gunther
about load average:
http://www.teamquest.com/resources/gunther/display/5/index.htm
Cheers,
Stefan
_
Hey,
> Wow, I thought you were about to encourage something like:
> - use of the SE Toolkit
> - making libproc a stable interface
> - adding a Sun::Solaris::Proc to Perl
+1
Sun::Solaris::Proc might be a nice thing to have in Perl.
> A decision you may have already made, would b
Hi,
Im not running GNOME/JDS but some things to consider:
- nvidia support should be good in Solaris. Make sure you are running the
latest drivers from Nvidia. Check what version are you running.
- bootup ~50sec - my machine boots in 30secs or so, I think even less. Do you
have same number o
Hi,
I also noticed that the RAM usage was excessive, the performance monitor
reports 426MB of RAM, when using just Firefox and Gaim, and another 41MB in
swap space.
You can check yourself the memory utilisation from mdb ::memstat.
You have already reported the figures.
> ::memstat
Page Summ
Hi,
It might be useful first to estimate if your current server can sustain the
workload you plan to do.
You cant simple add applications and then expect things will work the way you
want... so you gotta be careful.
Example: how many processes this machine has and how many syscall/s comparing
Hi,
One thing to consider is to see how other web software works under T2k. What
about replacing Apache with Sun's web server. Redo that for Ubuntu too then see
how things are flying.
rgds,
stefan
This message posted from opensolaris.org
___
perf-
ris vs BSD, and how
> they differ in terms of performance?
You might like to read the following article:
http://www.opensolaris.org/os/article/2005-10-14_a_comparison_of_solaris__linux__and_freebsd_kernels/
Stefan Parvu
Blog: http://stefanparvu.blogspot.com
__
Cool, thanks a lot.
stefan
This message posted from opensolaris.org
___
perf-discuss mailing list
perf-discuss@opensolaris.org
Thanks Eric for information.
It would be nice to have some information about Nevada and the mentioned bug
6336786:
in what build the problem is fixed.
Stefan
This message posted from opensolaris.org
___
perf-discuss mailing list
perf-discuss@opensolari
29 matches
Mail list logo
