Proxmox 3 Node Cluster with CEPH - proper network configuration for VM failover?

T

theprez1980

New Member
Feb 25, 2024
16
0
1
Hey All -

Trying to understand how this should be configured - it's currently in a lab environment, but thats easy as everything on the same 10.0.1.0/24 subnet in the lab. My main question is, how does Proxmox's network configuration be configured so that if the node goes down, the VM on the other nodes can still browse and remain functional?

More specifically, here's the network layout

Site 1 - Local - 192.168.0.X/24. Proxmox is running on 192.168.0.1, VMs are running from 192.168.0.10 to .20
Site 2 - Offsite, colocation - 192.168.1.X/24. Proxmox is running on 192.168.1.1, VMs are running from 192.168.1.10 to .20
Site 3- Offsite, colocation - 192.168.2.X/24. Proxmox is running on 192.168.2.1, VMs are running from 192.168.2.10 to .20

All 3 sites are interconnected via dedicated OpenVPN servers and can see each other and ping/access fine. Sites are connected with 1Gig/1Gig links.

My question is, a VM running on site 2, with an IP of 192.168.1.14, if that node goes down, and it fails over to another node - how does connectivity remain functional? I assume all 3 nodes need some kind of bridge or some other configuration to maintain connectivity.

Thanks
 
gurubert said:
You cannot failover in a WAN setup with VPN. The latencies are too high to setup a HA cluster.
Click to expand...
I'm getting 8-10 MS between the 3 nodes, is that too high? The 3 nodes are located in the same city, 2 of which have the same transport provider if that matters any
 
theprez1980 said:
I'm getting 8-10 MS between the 3 nodes, is that too high? The 3 nodes are located in the same city, 2 of which have the same transport provider if that matters any
Click to expand...
One could try reading the manual: https://pve.proxmox.com/pve-docs/pve-admin-guide.html#pvecm_cluster_network_requirements

5.7.1. Network Requirements​

The Proxmox VE cluster stack requires a reliable network with latencies under 5milliseconds (LAN performance) between all nodes to operate stably.
 
  • Like
Reactions: Johannes S and gurubert
You must log in or register to reply here.
Top Bottom