On 05/13/2011 09:56 AM, stoyan wrote:
after some struggle i managed to run a 0.10 broker in cluster on my local machine and started digging why it wouldn't start on the other servers. after some strace-ing i noticed that on the local server qpid attempted to connect(10, {sa_family=AF_FILE, path=@"corosync.ipc"}, 110) = 0 while on the other machines it failed when trying to connect(10, {sa_family=AF_FILE, path=@"libais.socket"...}, 110) = -1 ECONNREFUSED (Connection refused)now apparently it would fail because i started corosync service, not openais. on the server i have openais 0.80.6 (yum installed); i manually built and installed corosync 1.3.0. qpid's configure found both and i don't know what decisions it made, but eventually i only managed to start qpid with cluster on a machine without ais and i subsequently i gave up starting openais service after a bunch of these (with valid confs in proper locations, i believe): aisexec: [MAIN ] AIS Executive Service RELEASE 'subrev 1887 version 0.80.6' aisexec: [MAIN ] Copyright (C) 2002-2006 MontaVista Software, Inc and contributors. aisexec: [MAIN ] Copyright (C) 2006 Red Hat, Inc. aisexec: [MAIN ] AIS Executive Service: started and ready to provide service. aisexec: [MAIN ] parse error in config: No interfaces defined aisexec: [MAIN ] AIS Executive exiting (reason: could not read the main configuration file).
That may be a problem in your AIS configuration. There's a checklist at https://cwiki.apache.org/qpid/starting-a-cluster.html
--------------------------------------------------------------------- Apache Qpid - AMQP Messaging Implementation Project: http://qpid.apache.org Use/Interact: mailto:[email protected]
