Hello guys,
I would like to setup a Proxmox HA cluster and as for now, I see multiple approaches for making the VM's accessible when moved to a different node.
For example, having a 3 node Proxmox cluster. Each PVE node has its own public static IP address. The machines will be running inside a vLAN. Having shared nfs storage.
A mailserver VM is running on node1. One day, node1 should be updated and rebooted, therefore either automatically or manually the mailserver is moved to node2. The typical problem, mailserver is now under a different public IP address.
Current DNS settings:
So this will not work that way.
Now I see a few possible approaches. Also if there are some other, better approaches, please let me know.
- Approach1: Hetzner for example, has a vSwitch and the possibility to order a subnet, which has a pretty high setup fee + a monthly fee and somehow an outgoing Limit of 1TB. So it doesn't sound that interesting.
- Approach2: Hetzner is providing a simple round robin (cloud) load balancer, which can be setup to load balance between the dedicated servers. Pricing seems better than approach1, a traffic limit of 20TB for the smallest load balancer. But wouldn't work because only
- Approach3: A cloud flare load balancer. Smallest package cheaper than Hetzners solution, has no public static IP address. How I understand it, you have to update your CNAME records to point to cloudflares load balancer. And here I don't have a lot of experience with DNS Settings, to evaluate how or even if this could work. I guess I would have to remove the above mentioned
And the overall question is, can the load balancer approach even be used, instead of approach1 (subnet)? I would say approach3, the cloud flare load balancer seems to be the best approach especially as it seems to provide the functionality to detect an unresponsive server and not redirecting to it. Means, you would set it up like the mailserver is running on all of the three nodes. The load balancer on the other hand would always see only one available instance and always redirect to that one.
Has anyone experience with cloud flare or done something like that?
I would like to setup a Proxmox HA cluster and as for now, I see multiple approaches for making the VM's accessible when moved to a different node.
For example, having a 3 node Proxmox cluster. Each PVE node has its own public static IP address. The machines will be running inside a vLAN. Having shared nfs storage.
A mailserver VM is running on node1. One day, node1 should be updated and rebooted, therefore either automatically or manually the mailserver is moved to node2. The typical problem, mailserver is now under a different public IP address.
Current DNS settings:
A
records pointing to node1 (*, mail, www, etc.), CNAME
records (autoconfig and autodiscover) + some other records ( don't think they are somehow relevant for this example).So this will not work that way.
Now I see a few possible approaches. Also if there are some other, better approaches, please let me know.
- Approach1: Hetzner for example, has a vSwitch and the possibility to order a subnet, which has a pretty high setup fee + a monthly fee and somehow an outgoing Limit of 1TB. So it doesn't sound that interesting.
- Approach2: Hetzner is providing a simple round robin (cloud) load balancer, which can be setup to load balance between the dedicated servers. Pricing seems better than approach1, a traffic limit of 20TB for the smallest load balancer. But wouldn't work because only
round robin
and least connections
available. So if the mailserver moved to node2, the load balancer has to be aware of that and only send the requests to the available VM.- Approach3: A cloud flare load balancer. Smallest package cheaper than Hetzners solution, has no public static IP address. How I understand it, you have to update your CNAME records to point to cloudflares load balancer. And here I don't have a lot of experience with DNS Settings, to evaluate how or even if this could work. I guess I would have to remove the above mentioned
A
and CNAME
records, set a CNAME
record to redirect everything to the load balancer and inside the load balancer, setup the records again like mentioned above. At least I hope this is possible that way.And the overall question is, can the load balancer approach even be used, instead of approach1 (subnet)? I would say approach3, the cloud flare load balancer seems to be the best approach especially as it seems to provide the functionality to detect an unresponsive server and not redirecting to it. Means, you would set it up like the mailserver is running on all of the three nodes. The load balancer on the other hand would always see only one available instance and always redirect to that one.
Has anyone experience with cloud flare or done something like that?
Last edited: