- Download aixperf.sh
- Edit
aixperf.sh
then change SCRIPT_SPAN
to
greater than or equal to the duration of the test. Also ensure
JAVACORE_INTERVAL
is changed so that
SCRIPT_SPAN
is evenly divisible by it. For example:
SCRIPT_SPAN=43200
JAVACORE_INTERVAL=60
- Find the process IDs (PIDs) of each application server:
ps -elf | grep java
- Start aixperf.sh as root:
# su
# cd /var/tmp/
# nohup ./aixperf.sh ${WASPIDS} &
- Reproduce the problem
- The
aixperf.sh
script will gather a tprof sample just
once when it first starts. Also gather a tprof sample at the peak of the
problem:
# su
# cd /var/tmp/
# LDR_CNTRL=MAXDATA=0x80000000 tprof -Rskex sleep 60 >> tprof.out 2>&1
- Once the reproduction is complete:
- If
aixperf.sh
hasn't completed, you can manually stop
it:
ps -elf | grep $(ps -elf | grep aixperf | grep -v grep) | awk '{print $4}' | xargs kill
- Gather any
javacore*txt
files produced in each WAS
JVM's profile directory.
- Gather
/var/tmp/aixperf/*
and
/var/tmp/tprof/*
Previous Section (AIX vmstat Recipe) |
Next Section (Troubleshooting Windows Recipes) |
Back to Table of Contents