shibd leaves connections in CLOSE_WAIT state

Roland Tanner roland at
Mon Nov 14 19:22:48 GMT 2011


I have shibd (version 2.3.1) running on an Apache2 server for 
authentication of UK academic institutions. (Installed via apt-get on a 
Debian Lenny server.)

I'm finding that over time the server quickly accumulates an increasing 
number of connections in CLOSE_WAIT state which never completely close. 
Here's and example output from netstat.

Proto    Recv-Q    Send-Q Local Address                      Foreign Address                  State

tcp       38      0       localserver:41904                   CLOSE_WAIT
tcp       38      0       localserver:39484                   CLOSE_WAIT

tcp       38      0       localserver:39484              CLOSE_WAIT

The only thing that clears these connections is a forced close of shibd 
(eg by killing the process or a force-restart). Restarting networking 
with, eg, '/etc/init.d/networking restart' fails to clear them, and in 
fact networking won't close down while the connections remain open.

Any ideas about a solution? So far, it does not seem to happen so often 
as to exhaust the available connections, but the possibility of 
crashing/blocking the server seems to be there if they accumulate like 
this. From what I understand, CLOSE_WAIT status gets stuck if the local 
application does not close the connection, after receiving a request to 
close from the remote server.

Many thanks in advance,

More information about the users mailing list