The fork function failed. Too many processes already exist.

[Jose Mosquera]

I'am havving this trouble with a couple of users in my HP-UX 11.0, rest of users looks fine. I suppouse that nproc and maxuprc are involved. Trying to discard nproc influence, I have test it with sar -v 2 10, obtaining the following output:
13:21:28 text-sz ov proc-sz ov inod-sz ov file-sz ov
13:21:30 N/A N/A 1430/2420 0 4759/37680 0 7536/65300 0
As you see, limit of nproc it's bigger that wide-system procs number.

By other hand, maxuprc looks lower (integer value 300). I'm thinking that this parameter needs grow up. Some tip in that quantity should grow?

Pls find attached a file with the kernel configuration. Our Physical Mem is 4 Gb.


Rgds.

[melvyn burnard]

well if you think the nproc parameter is ok, then change the maxuproc arameter and monitor what happens to your users.
You may find you have to increase BOTH parameters in the end to get the desired result.
This will be a matter of change/test/monitor until you feel it is all ok.

[Steve Steel]

Hi


For certain users it must be maxuprc.

If they are just hitting limit then add 50%

Steve Steel

[Pete Randall]

If you have Glance (or can install the free trial version), you can check the system tables report. This will help narrow down which parameter needs to be increased.

Pete

[Fragon]

Even though swapmen_on should be set to 1 normally because your physical memory is 4GB, your can make a test to set it to 0, if your applications is based on Oracle!
Just an idea,need to check!

[Anonymous]

you may want to have a look at
http://docs.hp.com//hpux/onlinedocs/os/KCparams.OverviewAll.html

Overview of Process-Management Parameters
http://docs.hp.com//hpux/onlinedocs/os/KCparam.ProcessParmsOverview.html

Usually I tend to double/half the parameter to be modified (unless it meets some limit like maxuprc <= nproc-4)

pls be so nice to indicate to what extend our answers helped you
by assigning points.

thx, Tom.

[doug hosking]

Maxuprc is one place to look, but another is
whether one of those users has something like
infinite recursion in a script. If so, then
nothing you can do with system tunables will
help. I would be tempted to look for any
correlation between the error message and
specific activity by those users, then check
the applications that seem to be associated
with the error, to see if this type of
problem has happened.

$ echo "/sbin/sh ./a" > a; /sbin/sh ./a
./a: The fork function failed. Too many processes already exist.