At 11:37 PM 30/11/2005 +0000, you wrote:

>Some extra information we got today. Looking at a Linux log there are
>segfaults at the same time as the client read connection errors. The
>segfault entries look like
>
>Nov 30 09:49:35 dbserver kernel: fb_inet_server[14083]: segfault at
>000000004847c008 rip 00000000556469bb rsp 00000000ffffc4cc error 4
>
>Does this indicate the server is responding to a lost connection or
>are the segfaults causing the lost connection?

The segfault is crashing the server process, AFAICT. Googling this error
seems to indicate that it's a widespread 2.6.14.2 and earlier kernel bug on
dual Opterons, that seems to have got patched at some point since...

I suggest posting this piece of the problem in firebird-devel, since the
fb_inet_server process shouldn't be segfaulting *at all*. It's not enough
to say you're running Linux - repeat your hardware setup, mention exactly
which Linux (distro and kernel version) you are running and try to get as
precise as you can as to what the request might be that made this happen,
in case it is relevant...

One thing I would definitely want you to look at is what build of Fb 1.5.2
is actually installed there. Do an rpm query to find out the name of the
package. If it is the AMD (x64) build then it is strongly NOT
recommended. That build was withdrawn because it corrupted databases.

./heLen