Re: [syslog-ng]problem on HPUX with s_tcp snd d_tcp
I've just upgraded all version to 1.6.4 and the problem is the same. After a kill -HUP on the hpux-server the clients are going into CLOSE_WAIT state where the server is in the FIN_WAIT_2 state client: tcp 1 0 <client>:55682 <server>:10514 CLOSE_WAIT server tcp 0 0 <server>.10514 <client>.55682 FIN_WAIT_2 After a 'service syslog-ng restart' on all clients they open a new connection to the server and the connection will be established as I expect. I hope that helps.. Stephan
2004-07-01, cs keltezéssel 17:06-kor Stephan Hendl ezt írta:
I've just upgraded all version to 1.6.4 and the problem is the same.
After a kill -HUP on the hpux-server the clients are going into CLOSE_WAIT state where the server is in the FIN_WAIT_2 state
client: tcp 1 0 <client>:55682 <server>:10514 CLOSE_WAIT
server tcp 0 0 <server>.10514 <client>.55682 FIN_WAIT_2
After a 'service syslog-ng restart' on all clients they open a new connection to the server and the connection will be established as I expect.
I hope that helps..
hmmm syslog-ng relies that it gets an error when it tries to send to an already closed TCP connection. (basically it expects an EPIPE), can you check whether the clients get this kind of error? (using strace, truss or ktrace) a quick workaround seems to be to enable keep-alive() on your tcp sources, this way the TCP connection will not be closed, however this would not solve the caswe when you stop and start syslog-ng. -- Bazsi PGP info: KeyID 9AF8D0A9 Fingerprint CD27 CFB0 802C 0944 9CFD 804E C82C 8EB1
participants (2)
-
Balazs Scheidler
-
Stephan Hendl