2 server cluster with HA? Layman tutorial?

[proxnewb!e]

Hello, I seek your help/suggestion.

I am willing to double my web hosting (LEMP stack + some custom Linux software) cost by running two Linux virtual servers (in case each of the server needs to have same full CPU/RAM/disk specs). Otherwise in case the resources sharing/load balancing would be possible, i may get even 3-4 servers with like 30% of required computing resources each.

I am seeking to increase hosting server availability and ease of restoration:
- increasing availability (one of the two servers irrecoverably die and automatically is switched to one remaining server or i can do it manually using couple of clicks or commands)
- ideally also sharing resources so one of the two cluster members is not just an idle hot-swap
- easier to bring the website back online after one of the servers gets damaged
- avoiding any complex re-installation/restoration process with many steps while the site is offline

My idea is to have 2 KVM VPS in two different location/providers (i know it is not ideal due to high latency) and run Proxmox LXC on it so I can snapshot the server for ease of restoration. I have also read about ZFS filesystem replication, though i worry about filesystem curruption on a regular/budget KVM VPS.

I would need a tutorial or the commands that i can put inside a bash script so i can restore the hosting server very quickly with a couple of clicks or couple of bookmarked commands.

I am NOT looking for:
- cloud servers from famous brands which would solve this increased availability issue (have numerous reasons for that)

What is a good solution for this and can you please link to a layman friendly turnkey tutorial?

I have already seen:
https://pve.proxmox.com/pve-docs/chapter-ha-manager.html
https://pve.proxmox.com/pve-docs/ha-manager.1.html
https://pve.proxmox.com/pve-docs/chapter-pvecm.html
https://knowledgebase.45drives.com/kb/kb450268-clustering-proxmox-nodes/ - basic cluster and HA setup, yet fails to followup with some details explanation how that works and how can be configured.

Thank you

 

[esi_y]

Hello, I seek your help/suggestion.

I am willing to double my web hosting (LEMP stack + some custom Linux software) cost by running two Linux virtual servers (or 4-5, where 3rd and 4th would be cheap ~256MB VPS) in order to increase hosting server availability and ease of restoration:

You would not be able to provide the latency requirements on that cluster "network":
https://pve.proxmox.com/wiki/Cluster_Manager#pvecm_cluster_network_requirements

As a result, you will end up with...

- increasing availability (one of the two servers irrecoverably die and automatically is switched to one remaining server or i can do it manually using couple of clicks or commands)

Lower availability.

- ideally also sharing resources so one of the two cluster members is not just an idle hot-swap
- easier to bring the website back online after one of the servers gets damaged

More complex recovery.

- avoiding any complex re-installation/restoration process with many steps while the site is offline

See above.

My idea is to have 2 KVM VPS in two different location/providers (i know it is not ideal due to high latency) and run Proxmox LXC on it so I can snapshot the server for ease of restoration.

It will not work, there's enough threads on the forum with people having 2 nodes in 2 different buildings and experiencing issues. Also, 2-anything does not play well with HA.

I have also read about ZFS filesystem replication, though i worry about filesystem curruption on a regular/budget KVM VPS.

This would actually not be a problem.

I would need a tutorial or the commands that i can put inside a bash script so i can restore the hosting server very quickly with a couple of clicks or couple of bookmarked commands.

I am NOT looking for:
- cloud servers from famous brands which would solve this increased availability issue (have numerous reasons for that)

You need a load balancer instead. If you can, change your app architecture so that it can run on something like Kubernetes.

What is a good solution for this and can you please link to a layman friendly turnkey tutorial?

I have already seen:
https://pve.proxmox.com/pve-docs/chapter-ha-manager.html
https://pve.proxmox.com/pve-docs/ha-manager.1.html

Thank you

Sorry for being the party pooper.

 

[UdoB]

You need a load balancer instead.

This!

OP: High Availability in PVE solves a lot of issues, but it does not come for free and it does not fit onto every hardware - there are some requirements that are a must.

If your application can HA-migrate in PVE from one node to another (with a two minute stop = with loss of the current sessions) it should also be possible to construct HA on application level. The easiest approach would be working with a virtual IP address handled by something like HAproxy. This is probably much easier to build than a whole cluster and at the end it works quicker.

Just my 2€¢...

 

[esi_y]

- easier to bring the website back online after one of the servers gets damaged
- avoiding any complex re-installation/restoration process with many steps while the site is offline

Also, if this is what you care about a lot and want to be provider agnostic, consider looking at cloud-init and/or ansible. And database replication would be best handled on the application level.