Hello,

It's not the first time, our firebird server crash.
yesterday for exemple, The server accept lot of new connection but not answer to them (making all the client "freezing" undefinitivelly as they are no timeout in the API). also one of the connection was activelly doing something (100% cpu usage)... but don't know what exactly

let write all the circonstance when the server crash :

1/ WinDows 2008 Server & FB classic server 2.1.2
2/ one fb_inet was always to 100% CPU usage
3/ All the other FB_inet was "freeze", even firebird accept new connection, but immediatly "freezing" the new connection... like this using all (even more) than memory available (more than 500 simultaneous client yesterday, when in normal time, only 80)
4/ killing the client does not shunt down the fb_inet... i have no other choice than restart the server to kill them
5/ in the log i see lot of
AKSERVER Tue Nov 24 02:10:50 2009
INET/inet_error: read errno = 10054
and some
AKSERVER Tue Nov 24 01:19:44 2009
SERVER/process_packet: broken port, server exiting
but i don't know if these error happen "before" the server start to crash or after ...

any idea will be welcome !

what do you thing of theses supposition :

1/ network probleme (router shuntdown/ddos attack/etc..) make all the fb_inet in "freezing" state. one fb_inet do a tcp loop because of this and using 100% CPU... if it's a probable scenario how to prevent it ?

2/ database corruption? i try gfix -v -f and no errors reported.. what else i can do to be sure that the database is not corrupted ?

3/ Firebird bug ?

thanks by advance for your help !!
stephane