[Help/Question] Multi machines cluster using k3d

[dblas]

Hello to all,
I wondered if it would be possible to instantiate nodes on different machines using k3d instead of having all the nodes on the same machine.
For example : 3 machines with, for each one, 1 master node and a worker node all belonging to the same cluster.

What would be the tricks and or pbs to get in there?
Thank you.
db

[iwilltry42]

Hey @dblas , thanks for opening this issue!
With the current version of k3 this is not officially "supported" and I haven't tried it personally (was possible with v1.x though, so I know that it works).
For best functionality though you may want to use --network host for not having to bother with ports.
If this is going to be a production-like setup, I'd recommend you to go with plain k3s instead to avoid all the pitfalls and limitations introduced by the additional docker layer.
Some starting points:

1. create initializing server node: k3d node create --role server myfirstserver --k3s-server-arg '--cluster-init' --tls-san <IP-of-Host> --token somesecrettoken
2. create the other nodes (servers and agents) like this: k3d node create --role server|agent myothernode --token somesecrettoken --env K3S_URL=https://<IP-of-myfirstserver-host>:6443

Disclaimer: haven't tested this, but I guess that's how I'd try it.

[shahamit]

Would have been good to have this enhancement working since its very easy with k3d to create and delete clusters. Without k3d I found this link that demonstrates how to do with k3s.

Looking forward to set up a multi machine cluster with k3d.

[iwilltry42]

Closing this in favor of the older issue over here: #280
Just keeping the backlog clean and collecting all information in a single place 👍