Planning cluster with virtualization - where should I look? HA shared storage vs replicate all data Ceph. New to Poxmox

[EngChiVi]

Greetings,

I currently run a home environment containing domain controller (win 2016 server over lenovo S510 hardware) with two VMs using Hyper-V ( VM A for Pi-hole, VM B for Unifi controller, both on Debian linux as OS, no GUI) . I also have a NAS (QNAP 332X) with SFP+ connected to it from switch that has SFP+ ports. NAS holds all of my data and backs itself to older synology unit.

Issues i want to solve :
small issue - whenever I reboot the machine (windows patches ,etc), I lose internet, DNS , and wifi. this is disruptive
big issue - while I backup a server using Veeam to NAS, what happens if MB dies? I have usb key + full backup of the server but where do I restore it to? how long would I be out ? with kids learning from home and myself/DW working from home, this is a concern. I do not have any other copies of Windows Server license so I can not do it in Windows without buying additional license.


Considerations
a) I am willing to buy few nodes to build redundancy for the primary machine I have. budget up to $1000
b) I would like to introduce resiliency and have more than one node so things like bios update, etc on the hardware itself does not bring _everything_ down.

Questions:
1) Is there a concise description of HA options possible with Proxmox VE? in my professional life I am a DBA working on Microsoft stack so I am familiar with old HCI microsoft clustering (two nodes with shared storage between them, heartbeat connecting the nodes over separate channel) . This way there is only one copy of data and nodes just determine (using quorum voting) who owns it.
I am also familiar with more recent majority node clustering where instead of single copy of data, each node has its own copy and clustering is a service that just moves the pointer for the 'guest application' to decide which one is active.

What does Proxmox VE support? HCI ? shared everything cluster? both?

2) #nodes. I use to recall that HCI I can do in Windows with two nodes and shared folder/service/witness somewhere for quorum. Is this the same for Proxmox
3) in regards to Proxmox and Ceph (which to me looks like block level replication keeping underlying data in sync between noes), is the minimal number of nodes required 3?

what questions should I be asking? where should I be starting?

My hope is that I can put proxmox node(s) up, move pi-hole out of hyper-v VM into VM(?)/container(?) if supported, find the way to host Unifi controller , and create/move/convert my Windows Server into VM that mini-cluster can support. any single physical component failure (CPU, MB ,etc) at least gives me time to get a replacement without fully taking out my network and ISP access..

thank you in advance

 

[EngChiVi]

Am I asking wrong questions ? Any other hints?

 

[Alwin]

Am I asking wrong questions ? Any other hints?

Sometimes it takes its time. Especially with bigger formulated questions.

what questions should I be asking? where should I be starting?

I'd say, please first start off with our documentation. It explains the working of HA, corosync and what types of storages are possible.
https://pve.proxmox.com/pve-docs/pve-admin-guide.html

1) majority vote (see corosync in the docs)
2) yes 3x nodes (maybe look into storage replication)

 

[EngChiVi]

Sometimes it takes its time. Especially with bigger formulated questions.
I'd say, please first start off with our documentation. It explains the working of HA, corosync and what types of storages are possible.
https://pve.proxmox.com/pve-docs/pve-admin-guide.html

1) majority vote (see corosync in the docs)
2) yes 3x nodes (maybe look into storage replication)

Thank you sir, appreciate the response. I am reading the documentation now and starting to get cold feet regarding the whole redundancy project.

a. it looks like the desired older hardware I was hoping to use (recycled business desktops with quad core i5-6500 intel CPU ) would not support cluster + Ceph as Ceph documentation requires at CPU thread for each of the three main Ceph processes leaving only one for OS itself + any VM guests.

Also , this link gives me page does not exists - may be update the primary documentation for the pve-admin-guide ? https://docs.ceph.com/docs/nautilus/start/hardware-recommendations/

i think the current documentation link changed to https://docs.ceph.com/en/latest/

b. ignoring Ceph for now, would the cluster still work with multiple nodes and iscsi storage exposed from NAS over 10Gbe link? do I still need 3 nodes for this if all I want is hardware redundancy for primary VM host? will continue reading the documentation.


Thank you for your earlier help and response.

 

[Alwin]

a. it looks like the desired older hardware I was hoping to use (recycled business desktops with quad core i5-6500 intel CPU ) would not support cluster + Ceph as Ceph documentation requires at CPU thread for each of the three main Ceph processes leaving only one for OS itself + any VM guests.

Yeah, Ceph needs some resources. That's why I added the storage replication, it may more suit what you want to do. It also works with HA.

Also , this link gives me page does not exists - may be update the primary documentation for the pve-admin-guide ? https://docs.ceph.com/docs/nautilus/start/hardware-recommendations/

Yes, docs recently changed. Updated docs follow soon.

b. ignoring Ceph for now, would the cluster still work with multiple nodes and iscsi storage exposed from NAS over 10Gbe link? do I still need 3 nodes for this if all I want is hardware redundancy for primary VM host? will continue reading the documentation.

In general yes.