Tried to run a cluster-of-one-broker on Fedora 10. Client sessions cannot be established with the broker. Here are the installed RPMS of interest: openais-0.91-2.fc10.i386 openais-devel-0.91-2.fc10.i386 cman-2.99.12-2.fc10.i386 cmanlib-2.99.12-2.fc10.i386 qpidd-0.4.728142-1.fc10.i386 qpidd-cluster-0.4.728142-1.fc10.i386 The broker trace that occurs at connection setup: 2009-jan-05 16:01:17 debug RECV [127.0.0.1:56069] INIT(0-10) 2009-jan-05 16:01:17 warning SASL: No Authentication Performed 2009-jan-05 16:01:17 debug 192.168.122.1:28204(INIT) new connection: 192.168.122.1:28204-0x9fedfc0(local) 2009-jan-05 16:01:17 trace 192.168.122.1:28204-0x9fedfc0(local) write idle. 2009-jan-05 16:01:17 trace SENT 127.0.0.1:56069 INIT(0-10) 2009-jan-05 16:01:17 trace SENT [127.0.0.1:56069]: Frame[BEbe; channel=0; {ConnectionStartBody: server-properties={qpid.federation_tag:V2:36:str16(b9650cbd-92c4-438b-bf2a-4c3cdc1ff5a0)}; mechanisms=str16{V2:9:str16(ANONYMOUS)}; locales=str16{V2:5:str16(en_US)}; }] 2009-jan-05 16:01:17 trace 192.168.122.1:28204-0x9fedfc0(local) write idle. 2009-jan-05 16:01:17 trace 192.168.122.1:28204-0x9fedfc0(local) write idle. 2009-jan-05 16:01:17 trace RECVB [127.0.0.1:56069]: 52 bytes 2009-jan-05 16:01:17 trace MCAST held: [event 192.168.122.1:28204-0x9fedfc0 data 52 bytes] This happened with both the Python and C++ clients, with auth enabled and disabled on the broker.
I'm developing on f10 now and things appear to function correctly. Can you try this with the firewall off: system service iptables off There is a bug in the f10 corosync package (https://bugzilla.redhat.com/show_bug.cgi?id=478401 ) that is causing it to use the wrong port for multicast - it's using the endian swap of the configured port. So if your firewall only allows the configured port corosync won't work.
Alan, Thanks. That solves the problem. *** This bug has been marked as a duplicate of bug 478401 ***