ia64_corehw zombie

[Malte Neumann]

We have two Itanium2 based servers (rx2600 and rx5670) running under HP-UX 11.23.
Both machines show more or less directly after reboot one zombie process. It is owned by root and called ia64_corehw.
There is also a 'normal' process ia64_corehw running.

Is this permanent zombie a known 'feature' of HP-UX 11.23 or is our system somehow mis-configered.

Thanks for any advice.

Malte Neumann

[Nico Schockaert · Accepted solution]

Hello,
Indeed, this is currently a feature of HP-UX 11.23. You will see that the processes will be cleaned-up within 30 sec, and eventually a new process will appear. This is reported and will be fixed in the near future.
Regards,
Nico

[Andrew Merritt_2]

Nico is partly correct; it is not a permanent zombie process, but a new one being spawned every minute, and then dying.

This is known behaviour, and is not optimal, but it is not incorrect behaviour. It is not having any effect on the system beyond the overhead of forking the child process and dealing with its cleanup each time.

There are currently no plans to change this.

If it is causing a problem beyond being unexpected, please explain either here or via HP support.

Andrew Merritt

[Ivajlo Yanakiev]

This is normail :)
Why HP do this I do not know :(

[Andrew Merritt_2]

I'm sorry, Ivajlo, I'm not sure what your point is.

Andrew

[Ivajlo Yanakiev]

ia64_corehw create defunc and this is not real problem. HP told that !!!
But users like me do not like zombi and
defunc proces in there ps -ef
Get it ?
I can't get why HP make software ia64_corehw which make defunc.

Normaly if I have zombi, there is software that not work correct. I'm I right ?

[Andrew Merritt_2]

A defunct process is one that has died, but for which its parent has not yet issued a wait() command. It is not using any system resources, apart from a slot in the process table.

In this particular case, the parent ia64_corehw forks a child to process a log file every minute. When there is not much data to process, the child usually completes in a couple of seconds. The parent issues the wait() at the end of the minute, if the child has finished. This does have the effect that the child shows as for most of the time, and a new child is created every minute (unless there is sufficient data in the log file to keep the child busy for more than a minute).

I agree that this could be done more efficiently, and that the process is noticeable in the ps listings. In fact there are some plans to make changes to the monitor that would remove the child ia64_corehw in a future release, but nothing more definite than that currently.

However, it looks like an incomplete understanding of how Unix works to simply view this as an error. Can you point to any documentation that says it's an error for a process to be listed?

Andrew

[Ivajlo Yanakiev]

This is my experience, that defuncs refers to software that can not complete some stage.
/You write this on your msg/ :)

Did you see 10 or more defunc on one server ???
Is this normal ?????
I know that defunc not use any resources.
But also i know that 1000 mine serveres haven't defunc when they work whitout problem.
If you ask me :) if you see 10 or more defunc on your PS , take care about your software :))
Yes I'm not a programmer , but I saw a lot of software runing on UNIX based server.

I share what I know and my experience.
If you think that I'm wrong :))
Ok

I do not represent HP this is my personal opinion.

P.S. My customer ask me for this zombi process. Also he isn't a programmer.
do you get it ???

[Bryan Strait]

Is it correct behavior to have more than 2 ia64_corehw processes should be running at a time?

I'm running HP-UX 11.23 (Sept 2004) on an rx4640-4 and the system is on a pretty heavy constant load with three of the four processors locked by a process using pset_bind. At times, I've seen multiple ia64_corehw processes (more than 2) running at a time. Along with this, we've seen some issues reported with EMS not reporting expected events. I can only expect it's the multiple ia64_corehw processes but I haven't been able to find out how extra processes are getting started.

Hope this right place for this post. Thanks in advance for the help.

[Andrew Merritt_2]

Bryan,

Yes, you shouldn't see more than two copies of ia64_corehw at once time. What version of the OnlineDiags are you running? Is that Sept 2004 too? If so, you need to upgrade to a current version. There were some problems with ia64_corehw in the past, but they are fixed in the current version. http://www.docs.hp.com/en/diag/stm/stm_upd.htm#table

Andrew