powered by Jive Software

High availability/load balancing discussion

This is in reference to JM-191.

All the threads I found regarding load balancing or high availability were archived, so I’'d like to bring the discussion back up.

Currently for high availability, I am using Microsoft cluster services for an active/passive failover situation. This means that if the Jive service (or any other resource on one cluster node) fails and cannot restart, the cluter IP and all resources will fail over to the second node. The obvious disadvantage of this is that there is a 5-15 second downtime of the service and all clients will be disconnected.

An active-active setup where the two Jive Server nodes can talk with eachother would be ideal. A peer-to-peer setup (where there is no need for a router component) seems to be the most logical. Of course a load balancer of some sort would be needed to balance the packets between the servers.

Do the server-to-server connections allow for anything like this?

Message was edited by:


See the Project Pampero forum for discussion on clustering/load balancing/high availability: http://www.jivesoftware.org/community/forum.jspa?forumID=48

Pampero addresses scaling but not clustering. We’‘ve talked about clustering approaches several times in the past. The best approach will likely be to use Tangosol’‘s Coherence or Jboss cache. Both options provide clustering capabilities, although we feel much more comfortable with Coherence. We’'re planning to address clustering after Pampero and scalability, however.