Hi.
Since some time now I run a clustered setup of ejabberd.... and it basically breaks all the time and doesn't work at all.
There are two hosts in different networks, their connection is secured via IPsec (which alone works flawlessly).
The nodes run Debian sid with ejabberd 2.1.10 and erlang 15.b.
If you need more data please ask.
My general idea was to setup clustering for high availability, not for load balancing.
If one server is unavailable, the other should take over.
Some (of many) issues I've noticed):
- if the masterserver (a) is (or becomes) unavailable , b doesn't work correctly
- sometimes it looks as if was working, but connections to other servers just don't work (e.g. gmail).
- sometimes it's starting, and the process runs, but it doesn't bind to the xmpp ports
- sometimes when I then restart b, a process comes up, but not even any output to the logs.
The above happens in both cases, either when both are running fine and a goes away,... or if I freshly start b when a is already away.
- sometimes, when I try to stop the slave server (b), just nothing happens,... it claims it would have been stopped, but the process remains there forever
- when one of the two (master or slave) goes away and reappers the other typically doesn't notice this and usually the reappearing one doesn't work then.
Am I doing something wrong?
Find my config as well as the output of mnesia info (when both servers are running correctly) at:
Thanks,
Chris.
Oh and my servers can't contact jabber.org since months now,... absolutely no idea why.
