Ok...tons of responses. Everybody agrees on what a defunct process is:
"As I understand it, they exist because they were child processes which
were created, ended, but not waited on by their parents. They cannot
be killed since they are already dead. In order for them to go away,
their parent process needs to wait on them, which is apparently not
happening. The two ways I know how to make this happen would be
either to get their parent processes to wait on them, which is likely
impossible, or to kill the parent processes so that the children will
be orphaned and adopted by the system, which will then wait on them,
at which point they will go away. The only alternative is to reboot
the machine."
So I looked at the ps -ef|grep defunct, and did notice that all of the defunct processes had the same parent. It is one of the Xipc daemons (xipcisd to be precise if anyone out there uses Xipc). We haven't fixed it yet, but I can happily pass it on to Envoy (the most recent Xipc purchasers) for them to fix the problem.
Thanks to everyone who responded.
Jonathan Williams
Unix Systems Administrator
The Shubert Organization, Inc.
-----Original Message-----
From: Jonathan Williams [mailto:jonathw_at_shubertorg.com]
Sent: Wednesday, December 05, 2001 1:45 AM
To: Tru64 Mailing List
Subject: Defunct processes
Ok...we have 3 production servers--Alpha ES40 running Tru64 5.1 Patch Kit 3.
We have noticed a ton of defunct processes running on the server that is the
"online" server. These processes don't appear in near the same numbers on
the backup servers--but if we bring a backup server online, the defunct
processes start appearing. I was curious last night and ran the following:
ps -A |grep defunct | wc -l and there were right around 10400 defunct
processes. Now I have no idea what a defunct process is. Also I'm not sure
how bad they are, because they've been there for a long time (months and
months and months) but just today the big-boss-man decided that they need
to be looked into. So my very general questions are: what are these
processes, are they normal or is there something wrong with our production
application (big surprise there), and what do we do about them? The only
things these systems are running is: TU 5.1, Informix (IDS 2000, client
SDK, and SQL), Passport, and Xipc. I know it's a very general
question...but I just need to know where to start investigating.
Received on Wed Dec 05 2001 - 22:02:56 NZDT