hi, it would be a great benefit to be able to upgrade a cluster node-by-node. therefore you'll hava a time where your cluster exists of nodes of different versions (e.g. 1.4 and 1.5). this should be possible to do a smooth upgrade with minimal downtime (just moving the VMs) I don't need fancy stuff like live-migration between different version, as these may be really different to implement, but at least all your cluster communication should be able to speak to nodes of the last version.
communication protocol compatibility
- Thread starter pille
- Start date
we do not have the resources here (test equipment, man power, etc.) to test and support these scenarios. in fact, 1.4 and 1.5 nodes 'can' communicate but this is a non-supported and non-tested way and could lead to issues.
but anyways, if you need maximum up time on your VM´s, you should run a parallel test system where you can test all what you need before you go to the live system.
but you are right, it would be nice to have this.