> > I was getting identical Socket.pm errors. See > > http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5388 for my > diagnosis. > > Let me know if you'd like me to explain further. > > Aha! > > We have a similar setup here with the BigIP having two addresses on the > machine (a machine ip and a service ip). I just looked on the two > machines that weren't having any problems and I noticed that I had > apparently added the service address in the /etc/hosts file. On the > hosts that are having the problem, there is no such entry. > > I vaguely remember tracking down the code to a peername call, but I > hadn't figured out what peername was tripping it. I'll try adding an > entry in /etc/hosts and see how things progress.
I had tried that as well, to no avail. I suspect that in our setup, the tcpcheck from LVS/keepalived was closing the connection so quickly after discovering that the port was open, the socket was closed even before ->peername was called. -Matthew
