I only had a couple of replies on this one. See original querie below.
Stan Horwitz suggested using fuser to find out what or who is creating the
files. I tried this and it came back as root. Kind of a dead end there.
Derk Tegeler suggested ensuring that if we use tmp as part of the production
system that it either be a partition in its own right or at least not be on
root. We do in fact run tmp as an advfs fileset on the same disks as the var
domain. I have moved it to its own disk but this hasn't solved the problem
by removing the cause.
George Gallen suggested that Universe does create a large number of temp
files but none with the prefix we are getting. He also suggested that I take
this to the universe list. One of my colleagues here is on that list and
will be taking the question there.
In the meantime we have created a cron job that runs every ten minutes and
removes the temp* files older than an hour where ever they are on the
system. (We later found there were similar file in the /var/spool/lpd##
directories. This at least pointed back to the way that universe creates the
print jobs.
Thanks to those who offered their help. If we get anything out of Informix
or the Universe mail list I will pass it on.
Wayne Blom
Systems Specialist
F H Faulding & Co Limited.
-----Original Message-----
From: Blom, Wayne [mailto:Wayne.Blom_at_au.faulding.com]
Sent: Tuesday, 17 July 2001 10:54
To: True64 Unix Managers (E-mail)
Subject: Files in the tmp directory
Hi,
Just wondering if anyone out there has come across this particular problem.
We are running 4.0F on DS10 clusters. The application is running on an
Informix Universe database and prints via jetforms and LPRng. Quite a
volatile mix I agree.
The problem is that files are being created in the tmp directory during the
printing process. All the files are of the form temp00xx####. So far we have
eliminated jetform as the culprit as one of the clusters doesn't run it but
still creates the temp files.
Under normal circumstances I would say this is not a problem. However the
printing throughput of these systems results in the /tmp directory being
filled if we do not clear it out on a nightly basis. When the /tmp system
fills the machine goes into a hung condition requiring a hard reset to
recover.
As I said at the beginning, if anyone has seen or solved a similar situation
your input would be greatly appreciated.
Wayne Blom
Systems Specialist
IT_at_Healthcare
F H Faulding & Co Limited
Received on Sat Aug 04 2001 - 19:48:48 NZST