I have a Rancher single node install that’s managing clusters that are starting to take on production work - I am taking regular backups of the rancher data, but I’m thinking I should probably upgrade to a high availability rancher deployment to better insure rancher availability. Is this possible, and if so, how best to go about it?
Anyone running production clusters w/ a single node install, or are folks mostly running production clusters with the Rancher high availability deployment?
There’s a spectrum, and people use both ends…
If it’s just for you or a small company then a single container backed by persistent storage (e.g. an EBS volume) can be fine for your “production”. If the node running it dies and you still have the data, you can quickly spin up a replacement. In the meantime the clusters managed by Rancher continue running fine even while it’s gone. So HA may not be necessary or worth the effort or infrastructure cost.
If on the other hand there’s 100 people in your company relying on being in Rancher all day and they can’t do their jobs if the control panel is down, you definitely need HA.
If you are on the fence, you can also do the “HA” install with
--set replicas=1 to get the full helm install on standalone cluster, etc, but running only one of everything. Then that can be scaled up later. It’s just quite a bit more involved than the single-container
docker run ... rancher/rancher.
There is not a super official easy path to migrating out of single container, but it is possible.
That’s helpful, thanks vincent.