Hello,
My answers hereunder. I respond quick because it is a non paid service.
I invite you to contact an official partner to build the infrastructure
that suits your need (as I am).
BR
Gautier Husson.
Le 18/06/2026 à 11:38, Iván Méndez Cairós a écrit :
Hello,
We would like to know whether the following Disaster Recovery scenario can be
implemented with Proxmox VE and what would be the recommended architecture.
Our environment consists of:
* A primary datacenter with more than three Proxmox VE hosts connected to
a SAN storage array.
* A secondary datacenter with equivalent infrastructure (Proxmox VE hosts
and a storage array).
* Storage replication from the primary array to the secondary array.
* Both datacenters are connected through a LAN network with very low
latency and sufficient bandwidth, so network performance is not a concern.
Our objective is to ensure business continuity in the event of a complete
failure of the primary datacenter.
In such a scenario, we would like the virtual machines to be started in the
secondary datacenter with as much automation as possible and with minimal
manual intervention.
We would appreciate your guidance regarding the following points:
1. Does Proxmox VE officially support this type of architecture?
yes
2. Would you recommend a stretched cluster across both datacenters, or two
independent Proxmox clusters?
stretched (QDevice on a 3rd site + primary and secondary datacenter have
their own path to connect the QDevice) ==> HA will be automatic
3. What are the requirements regarding quorum, fencing, HA, networking, and
storage?
- same number of nodes both sides
- a QDevice in a 3rd location (can be > 5ms away)
- HA enabled on VMs + Node affinity for the 1st site
- ping < 5ms between ALL Proxmox nodes (QDevice does not matters)
- storage : SYNCHRONOUS replication
4. Is it possible for Proxmox HA to automatically start the virtual
machines in the secondary datacenter if the primary datacenter becomes
completely unavailable?
yes
5. What are the main limitations or considerations for this design?
None if all is respected. If a node is out of order for long time, you
should rebalance vote numbers between DC 1 and 2.
6. What would be the recommended Proxmox architecture and best practices to
achieve an automated Disaster Recovery solution in this scenario?
Any recommendations, documentation, or real-world deployment examples would be
greatly appreciated.
Official Proxmox Training of course :-)
