hi
These tests are run on many platforms, linux(suse, redhat) hp-ux, unixware, aix, solaris, tru64.
When the test for cpu idle fails I do a
ps -e -o user,stime,pcpu,pid,ppid,time,tty,args
to try to find out what process is hogging the cpu, at best the output does not clearly show which process is affecting it or at worst does not return anything other than 0.0 values for %CPU for all processes.
eg - vmstat returned 0 cpu idle but ps output was
========================================================
========================================================
I know that ps -e -o pcpu returns a percentage value - the CPU time divided by the time the process has been running and it will never add up to 100% but I would expect something useful.
top won't work on all platforms.
Does anyone have an ideas? Thanks in advance
ps -ef | cut -c42-80 | sort -nr | head -20
will give you the last 20 processes that took much cpu time in order. The output should look like this :
Thanks for your command Sysgate it seems I get a lot of old processes that are hanging around such as
00:04:25 root Feb28 0.0 6245 1 ? /opt/IBM/db2/V8.1/bin/db2fmcd
00:03:55 root Feb28 0.0 55 1 ? [kswapd0]
00:01:46 root Feb28 0.0 5834 1 ? /usr/sbin/nscd
00:00:27 root Feb28 0.0 2185 3 ? [reiserfs/0]
00:00:12 root Mar29 0.0 28598 1 ? /home/hub/staff/ss/cb/bin/mfs2
00:00:12 hub 13:44 0.1 32049 30759 pts/0 top
00:00:04 root Feb28 0.0 1 0 ? init [5]
00:00:00 wwwrun Feb28 0.0 13669 6122 ? /usr/sbin/httpd2-prefork -f /etc/apache2/httpd.conf
00:00:00 wwwrun Feb28 0.0 6127 6122 ? /usr/sbin/httpd2-prefork -f /etc/apache2/httpd.conf
00:00:00 wwwrun Feb28 0.0 6126 6122 ? /usr/sbin/httpd2-prefork -f /etc/apache2/httpd.conf
00:00:00 wwwrun Feb28 0.0 6125 6122 ? /usr/sbin/httpd2-prefork -f /etc/apache2/httpd.conf
00:00:00 wwwrun Feb28 0.0 6124 6122 ? /usr/sbin/httpd2-prefork -f /etc/apache2/httpd.conf
00:00:00 wwwrun Feb28 0.0 6123 6122 ? /usr/sbin/httpd2-prefork -f /etc/apache2/httpd.conf
00:00:00 root Mar29 0.0 28451 1 ? /home/hub/staff/ss/cb/bin/ejl32
00:00:00 root Mar20 0.0 19441 19430 ? /opt/kde3/bin/kdm_greet
00:00:00 root Mar20 0.0 19430 5878 ? -:0
00:00:00 root Mar20 0.0 19429 5878 ? /usr/X11R6/bin/X -nolisten tcp -br vt7 -auth /var/lib/xdm/authdir/authfiles/A:0-jcbx9Z
00:00:00 root Feb28 0.0 6250 1 ? ./mflm_manager
00:00:00 root Feb28 0.0 6244 1 tty6 /sbin/mingetty tty6
00:00:00 root Feb28 0.0 6243 1 tty5 /sbin/mingetty tty5
They don't seem to have been processes that would have kicked in and changed the cpu Idle of the vmstat report. Am I missing something?
offsihr:
hi
These tests are run on many platforms, linux(suse, redhat) hp-ux, unixware, aix, solaris, tru64.
When the test for cpu idle fails I do a
ps -e -o user,stime,pcpu,pid,ppid,time,tty,args
to try to find out what process is hogging the cpu, at best the output does not clearly show which process is affecting it or at worst does not return anything other than 0.0 values for %CPU for all processes.
eg - vmstat returned 0 cpu idle but ps output was
========================================================
========================================================
I know that ps -e -o pcpu returns a percentage value - the CPU time divided by the time the process has been running and it will never add up to 100% but I would expect something useful.
top won't work on all platforms.
Does anyone have an ideas? Thanks in advance
First sort based on the CPU% in the reverse order, something like this:
Now couple of things (just throwing in my 2 cents):
vmstat shows the summary of activity since the system was booted last. Well although the output changes with the kernel state changes, the old processes would still show up if they accounting to a higher CPU.
Do you want to use mpstat instead of vmstat. Hmm not sure if that would really change the output much.
deal_nodeal:
First sort based on the CPU% in the reverse order, something like this:
Now couple of things (just throwing in my 2 cents):
vmstat shows the summary of activity since the system was booted last. Well although the output changes with the kernel state changes, the old processes would still show up if they accounting to a higher CPU.
Do you want to use mpstat instead of vmstat. Hmm not sure if that would really change the output much.
thanks for your reply Deal_NoDeal.
The problem I'm getting is that I get 0.0 returned for all my %CPUs. AARGH!
When I use vmstat I don't take the first output for the reason you state. I get the latest val output after 2 seconds
#gets the column that contains the cpu idle val as diff col on diff sys
idColNo=`/usr/bin/vmstat 1 2 | grep id | awk -v idleVal=id -f cpuIdle.awk`
#gets val that corresponds to column
cpuIdleTime=`/usr/bin/vmstat 1 2 | tail -1 | awk -v idColNo=$idColNo '{ split($0, vals); print vals[idColNo]}'`
Do you have ideas?
offsihr:
thanks for your reply Deal_NoDeal.
The problem I'm getting is that I get 0.0 returned for all my %CPUs. AARGH!
When I use vmstat I don't take the first output for the reason you state. I get the latest val output after 2 seconds
#gets the column that contains the cpu idle val as diff col on diff sys
idColNo=`/usr/bin/vmstat 1 2 | grep id | awk -v idleVal=id -f cpuIdle.awk`
#gets val that corresponds to column
cpuIdleTime=`/usr/bin/vmstat 1 2 | tail -1 | awk -v idColNo=$idColNo '{ split($0, vals); print vals[idColNo]}'`
Do you have ideas?
I will try to find out more on this one. Nothing useful hits me now.