Enquiry on cluster setup for different DC

[uglyxiaodi]

Dear All,

Im new to proxmox, not sure if it is possible to link up different DCs in USA in one cluster ( for example OVH, Digital Ocean, hostwinds, tzulo etc..)

If yes, any guidance for me to refer to? I read before OVH vRack is able to link up all nodes from OVH in one region ( USA) not sure if this is true

How about other DCs? extra NIC to create vlans , allow the firewall port for proxmox?

Thank you
Xiaodi

 

[t.lamprecht]

Hi,

PVE native clusters are only available on LAN performance network, so roughly less than 6-8 ms latency - we heard of people getting it to work with 10-12ms too, but there you need quite stable net not to many nodes and a bit of luck, I guess.

We plan to add a datacenter manager in the future, providing a central management interface for managing multiple clusters and/or single nodes.

What you could do now is creating multiple cluster on each side and use Ceph mirror (or pve-zsync for single nodes or non-ceph setups) to mirror the storage of the VM/CTs to the other cluster in the other DCs, for redundancy.

 

[uglyxiaodi]

Hi,

PVE native clusters are only available on LAN performance network, so roughly less than 6-8 ms latency - we heard of people getting it to work with 10-12ms too, but there you need quite stable net not to many nodes and a bit of luck, I guess.

We plan to add a datacenter manager in the future, providing a central management interface for managing multiple clusters and/or single nodes.

What you could do now is creating multiple cluster on each side and use Ceph mirror (or pve-zsync for single nodes or non-ceph setups) to mirror the storage of the VM/CTs to the other cluster in the other DCs, for redundancy.

"What you could do now is creating multiple cluster on each side and use Ceph mirror (or pve-zsync for single nodes or non-ceph setups) to mirror the storage of the VM/CTs to the other cluster in the other DCs, for redundancy."

does it mean H.A still possible? to move all VMS from one node to another? but I believe there is a clause that it cant migrate from one cluster to another cluster

 

[t.lamprecht]

does it mean H.A still possible? to move all VMS from one node to another? but I believe there is a clause that it cant migrate from one cluster to another cluster

The Proxmox VE HA manager work only inside a cluster, so automatic HA failover would not be possible with that one.
But one could prepare the fail over VMs on the other side and then, after closely evaluating that it's safe to do so and switching ceph from secondary to primary.
https://pve.proxmox.com/wiki/Ceph_RBD_Mirroring
https://docs.ceph.com/en/latest/rbd/rbd-mirroring/

but I believe there is a clause that it cant migrate from one cluster to another cluster

There's actually some ongoing work to implement cross-cluster migrations, that feature may be quite a bit closer than the DC manager (as its a preparatory step for that one).