Question: How far can a cluster be from the Rancher instance?

Sorry If this is a stupid question. I tried to google this, but I couldn’t find an appropriate answer.
Here is a Little bit a background info. Currently We have a HA Rancher 2.5 cluster running on our hardware here in the USA. It manages a separate cluster in the same data center. Its for our deployments. We use Jenkins to auto deploy beta versions of our softwares to it.

Here comes the problem. We have an India Office with developers, and we are now adding servers over there for CI/CD. We want to have rancher over there to help them out. We have a connection that hooks their network with ours. The only problem is that it frequently gets to be like 200-300 ping at times.

This is where it gets interesting. Do we (1) Set up a separate Rancher ha cluster over there in India. Set up a separate Jenkins and then have to worry about two separate instances and have double the work. (Not a fan of this one.) (2) Use the Current HA here in the USA then have it link over to the India server to manage the cluster instead of creating two HA Rancher instances. Note that there is a high ping and I don’t know how rancher will work with that. This will allow one Jenkins, and one rancher HA cluster I have to worry about, so almost no extra work load for me.

My main concern is the ping between the servers.
If any body has any idea about this, it would help a lot!

If you need more info about the problem just let me know, I’ll try to explain bettter.

There’s no particular limit, but the UI and the default way your kubeconfig/kubectl will talk to the cluster is through the management server.

So that’s a long round trip if people in India right next to it are using it (browser in India → rancher server in US → cluster in India → back to server → back to browser).

Thanks @vincent for responding. Is there a better way that you know of? Or is it that there is no easy way around this?

  • Live with it; it may be acceptably slow; the Ui caches a lot of stuff up front on pageload. You can setup Authorized Cluster Endpoint so kubectl access goes straight to the cluster without going through Rancher.
  • Separate servers
  • Split the difference with one server in a location that’s more evenly inconvenient from everyone (Europe?)

Okay Thanks, I think I got a better understanding!