sar output gives 98% idle CPU

UNIX HP-UX
1

Dear All,
Our HPUX 8 GB 8CPU database server is behaving abnormally for the last 4+ weeks. I have generated a sar output and it is here-

11:46:52      %usr    %sys    %wio   %idle
11:46:53       1       1       6      92
11:46:54       0       1       0      99
11:46:55       0       1       0      99
11:46:56       1       2       2      95
11:46:57       0       1       1      98
11:46:58       3       3       1      93
11:46:59       1       1       0      98
11:47:00       2       2       6      90
11:47:01       1       3       3      93
11:47:02       0       1       0      99

During this time the entire system comes to a standstill and nothing is possible on the server. Logon to the database is very slow and the logged on users face a degraded performance. During normal period, the output is something like this...

11:47:35      %usr    %sys     %wio    %idle
11:47:36      55      45       0       0
11:47:37      55      45       0       0
11:47:38      50      50       0       0
11:47:39      50      50       0       0
11:47:40      53      46       0       1
11:47:41      61      39       0       0
11:47:42      45      55       0       0
11:47:43      57      42       0       1
11:47:44      47      49       0       4
11:47:45      55      42       1       2

Please help me.
Regrds
Ashrunil

2

Hmmm a heavy weight process which is causing a lot of CPU time to be consumed by the kernel as well.

Look at /usr/bin/top when it's happening. The responsible process(es) should become obvious.

-c

3

Did you install or upgrade software 4 weks ago?
Did you push a major change to your proprietary sfotware?

Can you tell what processes are eating the system? (craig's post) I don't think one process is doing that unless it is high priority - like realtime.

4

The original post seems to claim that the first sar (with a lot of idle time) is when the box is in trouble. The second sar (with little or no idle time) is when the box seems to be ok. This would be very odd and both responses have assumed that the situation is reversed. We need to clarify this situation. And the output from "vmstat 1 7" during both states might shed some light as well.