two cluster design possible?

[piramidata]

usually i build the following in Windows Server AD and Hyper-V members, but would like to know if this architecture is possible with Proxmox VE.

Two clusters named pve-ams and pve-fra. Between them is dark fiber running MPLS IP/VPN with multiple VRF's (VLAN's).

Each cluster has N members on which the VM's are created. Failover can be between local members, but also cluster to cluster.

Each cluster owns a single - redundant - repository with ISO files and each VM runs on an iSCSI storage space. The LAN & WAN are no bottlenecks.

Can i design it like this?

 

[t.lamprecht]

Each cluster has N members on which the VM's are created.

Can you please clarify this part? Does "N members" menas that you'll use nested PVEs or are those simply the VMs on the two nodes?

In general, yes, you can created two node clusters itself just fine. But, without further modification they cannot do (HA) failover if one Node fails. For that there are some solutions to mitigate it. The easiest one would be to use a QDevice, which can be on any linux box (does not needs to be a PVE) and has not the same latency needs as a real cluster member, see https://pve.proxmox.com/pve-docs/chapter-pvecm.html#_corosync_external_vote_support
There are some other tricks but those are all not a full solution, so I won't elaborate them here.

 

[piramidata]

The cluster servers have a management function ONLY. No storage or VM's run on them. The VM's run on hosts which are following Moore 's law (newest/latest trend). These host are processors, memory and small M.2 for OS/Proxmox servers startup. Actual VM's are on iSCSI.

These hosts can have Intel i, xeon or ARM cpu's. Also various AMD walk the parade. If Elon decides to put MUSK cpu's (CISC, RISC, MUSK) on the market, i will use them too to launch VM's into LEO (Low Earth Orbit).

In the case of a Cisco ASA, you replace CPU blades, power modules and interface blades. Not the backplane or chassis. That happens only during MOH.