powered by Jive Software

Question about openfire clustering with hazelcast and Haproxy

Hi Guys,

We will use hazelcast and haproxy to do cluster for openfire, and I checked many posts brefore, seems if one node is down, the user on it should logout and relogin, is that correct? the cluster plugin couldn’t handle it but we need to do this in our client?

Looking forward to your reply, thanks.

In our environment, the user will typically get kicked out or timeout when a node goes down. Most clients will reconnect after 30s or some other timeout.

Obviously if the client doesn’t have the user’s password saved it’s not going to work, but most people save it.

Thanks David, the information is quite helpful. Seems we need to do more things in our client code.

Hi Reed -

The internal routing mechanism for Openfire currently assumes that a dropped client will establish a new session (including reauthenication). I believe this can be improved in the cluster scenario, but since the most common XMPP clients already provide this reconnect behavior, the development effort has had a bit lower priority. I have created a JIRA ticket ( OF-689 ) to track this request.

Regards,

Tom

Glad to hear that Tom, that’s will be great!