RBD mirroring for disaster recovery

einsibjani

Member
Feb 5, 2020
16
4
23
45
Hi,

I've posted before about this (https://forum.proxmox.com/threads/ceph-mirroring-between-datacenters.66890/#post-300712).
We're still working out our DR plans. I'm a little spooked about reports of greatly reduced IO on ceph when using journal based
mirroring.

We have two identical 3-node clusters in two datacenters, connected by dark fiber. The cluster is hyper-converged, with each node
both hosting VM's and a ceph node. Each node has 5 x 2 TB NVM.e SSD's.

Does anyone have experience with rbd mirroring in a similar scenario and can comment on performance? Also, what can we expect
in terms of latency, from write on cluster A before data is written to cluster B?

.einar