Create a cluster with remote node

[fruchtzwerg]

I would like to create a cluster of a node inside my home-net and a remote vps (public IP).
I thought of tailscale to easily installng it.
So I created a new cluster on the remote node (with public IP). Unfortunately the node inside my homenet behind my router cannot join the cluster.

My questions:
(1) How can I connect all the nodes through tailscale's given IP's of the nodes?
(2) Do I have to create an interface for the tailscale net? Because just now proxmox only shows me the public IP as the IP.
Until now I have two interfaces for proxmox: first vmbr0 connected to the public IP and second vmbr1 which creates an intern network with an IP-range for all VM/LXC.

So what I am missing?

 

[gurubert]

What is the latency between the two nodes? The cluster manager needs low latency (small number of milliseconds).

 

[jlauro]

Is there a reason you want a cluster with a remote node? You are probably better off with it as two separate clusters (or individual nodes). What is the latency (RTT on ping) between the cluster and the remote node?

 

[basil]

This won’t work;

you need a connection from a dedicated server to a local server, but your local server is behind the router and dont have real ip

 

[jlauro]

I wouldn't say it couldn't work as they don't strictly have to be local... you can do a site to site vpn between the locations using vms on the internal networks (or on the nodes, but I would use a vm tied to a specific node). The only stated requirement I can find is the latencies are under 5 milliseconds, but are fine behind routers. You should be able to be about 50 miles apart and stay within that requirement assuming the same ISP or dedicated fiber. That said, if it's a random VPS not specifically in the same or near-by city, the latency is probably >5ms.

Still the question of why do you want them to be part of the same cluster? Are they going to have any sort of shared storage that you would want to move vms between them?

 

[fruchtzwerg]

Thanks for responding.
The main-reason to connect both nodes inside a cluster is that I want to secure the availability of the LXCs/VMs. If one node is offline the other can provide the services. Shared Storage would be great.
The node on the vps has limited ressources (2 Cores, 2 GB RAM, 60GB storage) but a 24/7 safe online connection
The node behind the router has much more ressources but limited bandwidth and highly depend on the home-setup.

I though mixing this could gain the potential and reduce the limits of both nodes separetly.
Maybe that requires much more knowledge than my skills.

 

[gurubert]

The main-reason to connect both nodes inside a cluster is that I want to secure the availability of the LXCs/VMs. If one node is offline the other can provide the services. Shared Storage would be great.

You cannot secure the availability of VMs or containers without shared storage.

 

[jlauro]

You are generally better off replicating the services within the VMs and not have them in the same cluster, unless you have either shared or synchronous storage (either of which you are going to need <5ms, maybe 10ms if your I/O is low, but asking for problems). You haven't mentioned your RTT yet to even know if that is possible. If not, then it depends on what is running on the LXCs/VMs how best to provide or take over the service.