Defunct Processes and UniVerse Print Spooler

[Joanne Keegan]

Hi Everyone,

Has anyone had this problem, and if so, is there a way to fix things?

We are running UniVerse 9.6 on a couple of HP-UX 11.00 servers. It appears that each time a user prints a document a defunct process is generated and is not killed off until the user logs out. Over a period of a day this can result in many defunct or zombie processes.

I know that I could up some kernal parameters to cater for this, but I suspect that this shouldn't be happening at all. I would rather fix it than changing kernal parameters if possible.

Any help or suggestions will be appreciated and I do award points.

Regards,

Joanne

[Bill Hassell]

Can't fix it with the kernel. A zombie (defunct) exists because the parent did not properly wait for a child process to complete. If this is consistent (ie, every print job causes a zombie) then it is a defect in the UniVerse code.

A zombie occurs because the parent did not execute a wait, wait3, waitpid or waitid or does not have SIGCLD set to SIG_IGN, the calling process is transformed into a zombie process. A zombie process is a process that only occupies a slot in the process table. Since the process is not running or occupying any RAM, there is no way to kill them (remove them from the process table) except to reboot.

Another possibility is that users are not terminating UniVerse correctly (very common on PCs). Users may be rebooting or perhaps just X'ing the window without first closing down the application.

[Joanne Keegan]

Hi Bill,

Thanks for your response. The strange thing with these zombies is that we don't need to reboot to get rid of them - they go when the user logs off.

We experimented with this, and we physically logged out a user with 23 zombies against his ID (we went to his terminal and did the logging out ourselves), and the zombies disappeared.

Thanks for your help.

Regards,

Jo

[Frank Slootweg]

Which process is the parent of the defunct processes? I look in the ps(1) "PPID" column of the defunct processes and do a "ps -fp ..." for that PPID.

This process is probably the culprit, i.e. its children have terminated, but it (the parent) is not wait(2)-ing for its children.

[Steve Steel]

Hi

if you check out kmtune you will see parameters which can be altered while ths system is live.This may be your best bet if hitting soomething like maxuprc.

Log a second call with
http://www.informix.com/informix/products/servers/universe/index.html


Steve steel

[Paula J Frazer-Campbell]

Joanne

How is universe dealing with the print jobs ?

If they are from within a VU scrip and the are delt with as a phantom routine then this can occur.

If you do a ps -ef | grep defunct do you still see the PPID as the users - most likly yes.

Had you tried stopping and restarting the universe spooler daemon (usd)?

This is not a unix problem but is a Universe script problem.

Are these jobs being delivered to the spooler from a universe hold file?

Paula

[Mark Greene_1]

Are you using the default directory for the spool file (/var/spool/, IIRC) or have you changed that to something else? Either way, check that the permisions on the directory are 777, and owned by root (this requirement changes at the 10.0 release).

Next, check the driver script permissions. It has to be at least 755 and owned by root.

Also check the uv spool file that tracks the spooler number information (I don't recall the name) for correct permissions.

HTH
Mark

[Rodney Hills]

We are running universe 9.6.1.6 and hpux 11.11 and have not seen this problem.

Is it possible you have an unquoted & (ampersand) in your interface script that is starting a background process?

I've worked with universe print spooler for many years, if you attach a copy of your driver I will take a look.

-- Rod Hills

[Rodney Hills]

Is this a new problem (did it work ok at one time)?

Any system change that could have started the problem?

Does this occur for all users on any printer?

Are your interface scripts calling the unix printer spooler for printing, or sending directly to the printers?

How are users submitting print jobs; through standard universe commands (SPOOL, LPTR, etc) or is a shell call made to the "usp" program.

-----

If your users have to log off before the defunct jobs disappear, then I'm thinking less that it is a print driver interface (since it runs as a seperate process) and more it is in how the user submits the print job.

-- Rod Hills

[Joanne Keegan]

Thanks to those who have contributed so far...

I am looking into some of the suggestions and will post the fix (if I find it). I am on course next week (yay!), so it maybe a week or so before I award points.

Rodney, I'm particularly interested in your ideas, and will post the driver if I can't isolate the problem soon. Thanks for the offer of help.

I am still open to any further suggestions, so keep them coming.

Thanks for the help/support.

Regards,

Jo

[Joanne Keegan]

Attached are copies of the printer drivers. Generic is for any line printers that may still be in existence.

Regards,

Jo

[Joanne Keegan]

And now for the attachments ...

[Joanne Keegan]

The other attachment...