topic Re: sam message in Operating System - HP-UX

sam message

mark_238 — Wed, 26 Mar 2003 01:40:26 GMT

09:47:56 RED CPU Bottleneck probability= 100.00%
09:48:17 END End of CPU Bottleneck Alert
09:50:19 RED CPU Bottleneck probability= 100.00%
10:00:06 RED CPU Bottleneck probability= 100.00%
10:03:20 END End of CPU hi, i was in sam, typed shift-H and got the below messages..

Bottleneck Alert
10:08:48 RED CPU Bottleneck probability= 100.00%
10:08:48 RED Network Bottleneck probability= 95.00%
10:08:54 END End of Network Bottleneck Alert

Re: sam message

Sridhar Bhaskarla — Wed, 26 Mar 2003 01:45:40 GMT

Hi,

These message are coming from glance. Try running '/opt/perf/bin/glance' to find more details.

If the file '/var/opt/perf/alarmdef' is incorrectlly configured, it may give you false alarms. When you run glance, it will display cpu, mem, disk and network bars in the main window. Look at them and see if they really are maxing out. Network alarm definitions have a good chance to be incorrect as by default, alarmdef is configured taking into account only one 10BT interface.

-Sri

Re: sam message

mark_238 — Wed, 26 Mar 2003 01:55:47 GMT

right, are there any recommended or standard values for parameters in /var/opt/perf/alarmdef?

Re: sam message

Michael Tully — Wed, 26 Mar 2003 02:10:40 GMT

Hi,

Have a look at these two postings.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x717cef70e827d711abdc0090277a778c,00.html

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xb72618276953d61190040090279cd0f9,00.html

Regards
Michael

Re: sam message

Sridhar Bhaskarla — Wed, 26 Mar 2003 02:24:27 GMT

Hi Mark,

If you go through the alarmdef file, you will find it very interesting. You will need to configure it to your needs applying that logic.

For ex., in the following simple memory probably definitions

symptom Memory_Bottleneck type=MEMORY
rule GBL_MEM_UTIL > 95 prob 80

rule GBL_MEM_PAGEOUT_RATE > 5 prob 5
rule GBL_MEM_PAGEOUT_RATE > 10 prob 5
rule
GBL_MEM_PAGEOUT_RATE > 25 prob 10

With the above definitions, I define my alarms like below

alarm Memory_Bottleneck > 80 for 5 minutes
type = "Memory"
start
if Memory_Bottleneck > 90 then
red alert "Memory Bottleneck probability= ", Memory_Bottleneck, "%"
else
yellow alert "Memory Bottleneck probability= ", Memory_Bottleneck, "%"
repeat every 180 minutes
if Memory_Bottleneck > 90 then
red alert "Memory Bottleneck probability= ", Memory_Bottleneck, "%"
else
yellow alert "Memory Bottleneck probability= ", Memory_Bottleneck, "%"
end
reset alert "End of Memory Bottleneck Alert"

In the above, as long as the memory probability is 80-90%, then it is going to be yellow. The memory probability = 80 means, the system is using more than 95%. Anything above will add the other probabilities. At memory probability 100, it is not only the indication of 95% or more of memory utilization but it is doing atleast 25 pageouts in that interval.

-Sri

Re: sam message

Steven E. Protter — Wed, 26 Mar 2003 04:56:38 GMT

You were running your systmem hard for a while. Not a big deal at all.

Here is a script for collecting data on peformance in the background, based on your parameter settigns. Its production.

Its attached.

SEP