problem on sol 8 - 280R

[antonioelderprado]

Hi,
a machine has 8G of RAM and is very slow.
I have ran vmstat / sar
on vmstat the b column (procs) shows me a huge number = 234183 - this number keeps steady . on cpu column the machine is almost 100% idle.
from sar -a the only column with numbers is namei/s that has number up to 1000 the other two columns zeros.
the sar -u under %wio goes to 98% - 100% on waiting.
I did ran also vmstat -s and looked under cache hit ratio and shows me 99%.

any ideas ...

thank you all.

./antonio/.

./antonio elder prado/.
.\mountain view, ca\.
./bauru, sao paulo/.

 

[KenCunningham]

What is the output from top - any obvious candidates for hogging all the resource?

 

[daFranze]

or prstat output?

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years

 

[antonioelderprado]

DaFranze and Ken

I have only prstat, on prstat there is no process that takes more than 1% of usage. I believe the cpu is not being accessed by some i/o block. the %w/io is always 90-100% and the %usr, or %sys are always close to zero.
I was checking the system, and it has java 1.4.2 however out of 29 patches that solaris requires for java only 2 are in there. do you think this could cause all this mess ? do you know what kernel parameters I can work on it to get those blocked out ? thank you guys. I am in CAlifornia and in 1 hour I will be at office. I will keep you posted of what is going on.
thank you

Peace.

./antonio/.

./antonio elder prado/.
.\mountain view, ca\.
./bauru, sao paulo/.

 

[daFranze]

what do you mean by slow; what do you run to say, it's slow? Is there a "slow" application? Did you restart this? I'm thinking about an application claiming lot of ram (we used Unicenter some time ago and it's Solaris Agent was a mess), so look at the Size column of prstat.

Another idea is, maybe you have problems with the disks, so look at 'iostat -pn' if you see something "mad". The problem will be to find out what is "mad" when you don't have something to compare these values...

Check for Logfiles, some applications open logfiles with systemcalls using 32bit pointers [open()] instead of [open64()] which is 64bit arithmetic...

Are there some hot entries in messages?

hmmm, no further ideas... I don't think patches are a problem, I have seen lot's of problems but a sudden performance problem is usually not a patch problem...

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years