topic Re: OS Problem in Operating System - HP-UX

OS Problem

Fuad_1 — Sun, 07 Mar 2004 03:38:33 GMT

I usually run billing job, which is terminated normaly. When I checked the log, it says there was not enough space to run this job. Therefore it is a problem of the OS that it did not gave an error message.

Can you please help, what is the OS problem? The OS is HPUX 11.11.

Re: OS Problem

Sridhar Bhaskarla — Sun, 07 Mar 2004 03:53:46 GMT

Hi,

It could be due to

1. Insufficient swap space. Run "swapinfo -t" command and see if the total is near 100%. If so, then you will need to add more swap or bring down unnecessary processes.

2. Insufficient filesystem space in the directory used by your billing job. A 'bdf /your_filesystem' will show you how much you have. If you don't know what that filesystem is then do 'bdf' and note all the filesystems that are above 90% and see if anyone of them are used by your billing job.

It's most unlikely due to any kernel parameter settings as I believe the same job was running before on this system.

-Sri

Re: OS Problem

Steven E. Protter — Sun, 07 Mar 2004 04:04:30 GMT

Could be no space on the filesystem as noted.

Could be not enough processes nproc,nfile

Quick fix on that is to up the maxusers parameter on the kernel.

You can check termination as follows.

at the prompt:

scriptname

rc=$?
echo $rc

I'm not sure you really got a code zero normal termination there. In the circumstances you describe, you should have gotten an error code.

SEP

Re: OS Problem

Fuad_1 — Sun, 07 Mar 2004 04:10:01 GMT

How to verify, that the OS generate error message? This problem occur a week ago.

Re: OS Problem

Steven E. Protter — Sun, 07 Mar 2004 04:44:23 GMT

Run the job again.

Unless you write a custom script to collect such data, nothing is recorded except.

If a filesystem was full and the system has not been booted there might be a record in /var/adm/syslog/syslog.log

SEP

Re: OS Problem

Sridhar Bhaskarla — Sun, 07 Mar 2004 04:52:53 GMT

Hi,

If you have measureware installed, you can pull the historical data using the command 'extract'. It's a self-guiding utility that can take you through step-by-step process. During extraction it will generate a default template file with all the metrics (most of the metrics carry meaningful names GBL_SWAP_SPACE_UTIL for global swap space utilization) and you will need to uncomment the metrics that you need to get report on. You can specify the interval during which your job failed so only the metrics during that interval will be extracted.

If you don't have measureware, you can't do much other than waiting for another unlucky opportunity as it didn't give you an error message. Kernel parameters like nfile, nproc usually generate syslog messages if overflown.

-Sri

Re: OS Problem

Steven E. Protter — Sun, 07 Mar 2004 05:37:45 GMT

Hi Sri,

Can't sleep? Me neither

Try this:

http://hpux.connect.org.uk/hppd/hpux/Sysadmin/tusc-7.5/

Its called tusc.

You can call it like this in your script

proc=$$

/usr/contrib/bin/tusc -f -o $TRACE_FILE -p $proc &

TRACE_FILE is a file you collect data in. -f traces forces procis the process id you picked up with the $$ command.

SEP