We are running Firebird 1.5.1.4481 (nptl build) on a dedicated
Mandrake 10.0 Linux server with 2GB ram (no gui - console only). We
have approx 150 (w2k) clients accessing 60+ gdb's. The client apps
are written in Delphi7+IBO 4.3a & have the windows build of
fbclient.dll (1.5.1.4481) installed.

All is running well :-) ... except at approx 17:30 most days -
presumeably as everyone is logging out - an error is reported in
firebird.log and rapidly & continously repeated filling the 5GB
partition in about 45 minutes.

The processor useage also climbs. Killing the fbserver & letting the
guardian restart it is the only solution I currently have.

#firebird.log snip ...

mydomain.myserver.co.uk (Server) Wed Aug 25 17:35:17 2004
Super Server/main: Bad client socket, send() resulted in
SIGPIPE, caught by server
client exited improperly or crashed ????

mydomain.myserver.co.uk (Server) Wed Aug 25 17:35:17 2004
INET/inet_error: send errno = 32

<repeat -forever!>
#

Error 32 is a broken pipe. Is it a crashed client? Even so, why does
Firebird not recover?

Any help much appreciated.