--H1spWtNR+x+ondvy
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

* kdeems AT parker DOT com <kdeems AT parker DOT com> [2003-10-30 09:41]:
>=20
> Hi,
> I am having the same problem in an AIX environment.  Did anybody test the
> changes suggested in tech note 237665?  My system is running much slower
> since the FP5 upgrade, not to mention the whole system actually locks up.

The problem is stale Java connections between the client and the server.
It has NOTHING to do with settings on the clientside Java. I talked to
Veritas support to resolve the issue of cleaning up defunct processes on
the master, and the response was to identify the parent java process to
the defunct process and kill it. Most of our defunct processes are from
clients that have been left running for long periods of time and have
either gone stale on their own, or have been cut off be networking
parameters to disconnect idle connections. Here is an example:

     UID   PID  PPID  C    STIME TTY      TIME CMD

    root  7069  7068  0 07:08:52 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root  7070  7068  0 07:08:53 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root  7352  7068  0                   0:00 <defunct>
    root  7347  7068  0                   0:00 <defunct>
    root  7355  7068  0                   0:00 <defunct>
    root  7346  7068  0 07:14:16 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root  7084  7068  0 07:08:56 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root  7068     1  0 07:08:52 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root  7108  7068  0                   0:01 <defunct>
    root 14401  7068  0 07:45:13 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root 14888  7068  0 07:56:39 ?        0:07 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root  7079  7068  0 07:08:55 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf
    root  8970  7068  0                   0:06 <defunct>
    root  7077  7068  0 07:08:55 ?        0:00 /usr/openv/netbackup/bin/bpj=
ava-susvc root 0 -1 en_US /usr/openv/java/auth.conf

These are all the processes that have PPID 7068. You will notice=20
that a number of them are defunct. The PID 7068 is bpjava-susvc ( the
Java client connection ). If you kill That process, all the other
related clide processes will die also.

Here is the korn shell script I am using to clean up these processes.
Please keep in mind that this is on a Solaris 8 system, so your exact
process may be a little different. The reason for the logic at the end
is so you don't kill the parent process UNLESS it is bpjava-susvc. This
works well for me, but PLEASE test this on your own system before
putting it into production. =20

#!/bin/ksh
#
# cleanbpjava.ksh
#
# script to clean up defunct NetBackup java client connections
#

GREP=3D/usr/xpg4/bin/grep

for p in `ps -eaf | $GREP defunct | $GREP -v grep | awk '{ print $3 }' | so=
rt -u`
do
ps -eaf | /usr/local/bin/gawk -v PID=3D$p 'PID=3D=3D$2{print}' | $GREP -q b=
pjava-susvc

if [ "$?" -eq "0" ]
then
    kill -9 $p
fi
done
# End of script

--=20
oavid Rock
david AT graniteweb DOT com

--H1spWtNR+x+ondvy
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)

iD8DBQE/oUOaMrO4/Yb/xwYRAr/dAKC6W92TFSIrGZy57oCamR0yL+GctACfWozU
P/04JmFp2Q9DEk+yjVpr6Uk=
=+Vbs
-----END PGP SIGNATURE-----

--H1spWtNR+x+ondvy--