Hey,
I am planing to create a 4 Node Streched Cluster with Ceph. Given that I have 4 nodes means 2 on each side I am in the need of a Quorum for Proxmox and a Tiebreaker Monitor for Ceph.
As I read the Ceph Tiebreaker can even be in the Cloud or another location because no OSD is speaking with them and he will only elect an Winner in a net-split situation. That means I am not in need of a very low latency and much throughput. Is this Correct?
To create a Quorum for Proxmox in a different Location I need a Qdevice or one more Proxmox Server. What I have read so far is that a Qdevice can work with higher latency than a Cluster Member. Is that even Correct? Would be 20ms to much for such an Qdevice?
My Idea was to create a fifth Proxmox Server which is not particapting in the Ceph Storage Cluster and only have the Monitor as Tiebreaker installed. Also I could restrict the Host from hosting VMs. In this way I would be able to have everything under Support and a convienend update and management way. The only question now is that I would not have qdevice which can have higher Latencys. It would be a normal Cluster node. Would that mean that my Quorum cant have anymore than 5ms of Latency?
What would be your suggestions to archive this kind of Cluster with an easy to Manage solution?
I am planing to create a 4 Node Streched Cluster with Ceph. Given that I have 4 nodes means 2 on each side I am in the need of a Quorum for Proxmox and a Tiebreaker Monitor for Ceph.
As I read the Ceph Tiebreaker can even be in the Cloud or another location because no OSD is speaking with them and he will only elect an Winner in a net-split situation. That means I am not in need of a very low latency and much throughput. Is this Correct?
To create a Quorum for Proxmox in a different Location I need a Qdevice or one more Proxmox Server. What I have read so far is that a Qdevice can work with higher latency than a Cluster Member. Is that even Correct? Would be 20ms to much for such an Qdevice?
My Idea was to create a fifth Proxmox Server which is not particapting in the Ceph Storage Cluster and only have the Monitor as Tiebreaker installed. Also I could restrict the Host from hosting VMs. In this way I would be able to have everything under Support and a convienend update and management way. The only question now is that I would not have qdevice which can have higher Latencys. It would be a normal Cluster node. Would that mean that my Quorum cant have anymore than 5ms of Latency?
What would be your suggestions to archive this kind of Cluster with an easy to Manage solution?