> However, I'm a bit doubtful that Firebird is writing any files to a file or
> directory named /tmp/core. The /tmp directory is globally configured, for
> use by apps and the OS. Firebird's temp files have dynamically generated
> names with numbers in them which are, by default (i.e. if you didn't
> configure space specifically) written into /tmp. Is "core" the name of the
> file or the name of a directory? On RH 7.2 I see Firebird temp files but
> they aren't named "core"... It looks a bit like a core dump...
>
> heLen
>

I believe it is a core dump.
1. It is binary.
2. It is HUGE - 150 MByte
3. The name is probably trying to tell us something ;)

The only thing I have found relating IB/FB to core is something I found somewhere on the web that said an old IB bug could cause a core dump on a drop trigger. This is not the issue here.

What I find significant is that these core files appear to have a similar datestamp to interbase.log entry pairs that look like:
/opt/interbase/bin/ibguard: bin/ibserver terminated abnormally (-1)
/opt/interbase/bin/ibguard: guardian starting bin/ibserver

Just so everyone knows, we are looking at a Linux box that is only used to run FB, though there is an Apache server active. Number of users is no more than 5, and there is only one table being accessed.

I have two Linux boxes that are set up identically, one serving as an IB server, the other as the company www server. The only one that gets this core dump is the one running FB. I have even switched machines, and the error follows FB.

There is plenty of partition space for /tmp, so I don't think that is the problem. However, I have learned a lot about configuration from all of the suggestions so far, and will put the temp directory for FB in its own partition.