Open Stack

> On 29 Jun 2016, at 18:33, Kosnik, Lubosz <lubosz.kosnik at intel.com> wrote:
> 
> May you specify what exact use-case you have to upload incompatible images?
> In my opinion we should prepare a flow which is like you said building new instance, configuring everything, adding that amphora into load balancer and removing old one. In that way we will be able to minimize retry to specific Amphorae not all load balancers.
> Everything depends of that are we able to do something with Amphora image that it will not work any more in cluster with older versions.

You picked just the third minor concern I had, and the one that is indeed quite theoretical. I just feel safer when my cluster runs the same version of the image, but yeah, generally it won’t be an issue.

I think we should primarily stick to two other concerns I raised: 1. HA assumption 2. cluster degradation for a non negligible time.

Ihar