CPU has 100% Idle and no connection to Remedy anymore

[walter blug]

Hello,
Since two days our productive Remedy/ARS-Ticketing tool is online. Several times a day you can observe (command "top") that the CPU idle time becomes 100%. In this situation you can't work with the ticketing tool anymore. A new connection to the ARS tool is also impossible. After 2-3 minutes I observe that the idle time falls below 100%. Now you are able to work again. This behavior is valid for all users. They all have to wait. It' s like a oracle archiving stuck. The whole ARS stands for a while. Then everything is fine again. Has anyone an idea?
Kind regards

Walter Blug

[Hemmetter]

Hi Walter,

What ist the load of the machine at that times?
What is your swap utilization?
Do you get any messages in syslog.log? E.g. about nfs or other?
Are your oracle table statistics up to date?

Pls give more info.

rgds
Hemmetter

[Arunvijai_4]

Hi Walter,

Can you post the following,

# swapinfo -tam
# vmstat
Output of SEP's system performance script, http://www.hpux.ws/system.perf.sh

-Arun

[walter blug]

Some more info:

swapinfo -tam:
dev 4096 262 3834 6% 0 -dev/vg00/lvol2
reserve - 1568 -1568
memory 1504 738 766 49%
total 5600 2568 3032 46% - 0 -

CPU:
LOAD|USER|NICE|SYS|IDLE|BLOCK|SWAIT|INTR|SSYS
5.95|89.6%|0.0%|8.0%|2.4%|0.0%|0.0%|0.0%|0.0%

There are no current entries in Syslog.log.

[walter blug]

Some more info: I attach the summary of the script mentioned in one of the answers.
If neccessary I will attach the whole output of this performance script.

[Hemmetter]

Hi Walter,

A high load with idle CPU may be caused by an io-bottleneck.
Is there oracle running on your server? check your table-statistics. out-off-date statistics can cause your oracle to not use some indices but do "full table scans" leading to very high io-load.

Do you have "glance" on your server?
Run it to monitor Systemtables, io, etc.


rgds
HGH

[walter blug]

Every time the server "sleeps" all parameters of cpu states (command "top")are zero or near zero.