Two possibilities Thomas.

Is /tmp symlinked to /var/tmp?  The problem may not be where you think
it is.
I have known for AIX filesystems to fill up but there be no file or
files big enough to cause the problem.  In Unix, if a file that is open
is deleted, the directory entry pointing to the file is deleted, but the
file itself does not disappear until it is closed.  Use lsof to see what
is open.

Regards

Steve

Steven Harris
AIX and TSM Admin, Brisbane Australia


Thomas Denier wrote:

-----Bill Kelly wrote: -----

Is it not possible that the /var filesystem *sometimes* has too
little space available? After all, didn't you say that this only
fails roughly every one to two weeks? Perhaps the failures coincide
with some other activity (possibly unrelated to TSM) that
temporarily eats up most/all of the available space in /var?
Dumps?  Logs?  Something else?

That scenario is essentially impossible to disprove conclusively,
but it doesn't look very promising as an explanation. The Linux
system is dedicated to running the TSM server and some monitoring
and automation functions for the server. We have been able to
check the free space in /var within an hour of most of the
occurances of the write failure. Dumps or log files that filled
up /var would probably have still been around when we checked
free space. There are no cron jobs that clean up files in /var,
and all the people who could have removed such files know I aminvestigating
the write failures.