I've just landed some changes on juju tip that enable "highly available" juju environments. In a HA environment, there are several state servers - if one dies, one of the others should take over.
To make a HA environment, run the "ensure-availability" command. For example to make a HA environment with three state servers, run: $ juju ensure-availability -n 3 This will add new servers as necessary. It will also maintain the current state server set if some servers have gone down. A new field in the status output, state-server-member-status, present only for state server machines, indicates the current status of each state server. When the status of a majority of state server machines is "has-vote", the environment should be resilient to any one of them going down. If a state server machine goes down and you run ensure-availability again, it should detect that the machine has gone down and create a new machine to replace it. When the new machine is available, the original (down) machine's status will become "no-vote". If you run ensure-availability again, it will see that the machine has no vote and is down, and will remove its state server status. At that point the machine can actually be removed (you'll need the --force flag). Anyone that's interested, please try it out. It would be great if people could see whether it works for them - there are bound to be some teething problems! There is one major caveat: upgrading an existing environment to HA is not supported currently (Andrew Wilkins has a branch in review that should fix this). Please note that the transition times when mongo is failing over can be quite long (on the order of minutes), so you'll need to be patient. Many thanks to everyone that's pitched in to help get this done. cheers, rog.
-- Juju-dev mailing list [email protected] Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/juju-dev
