Dear Members,
we're planning to integrate a Proxmox/Ceph Instance into our production-enviroment. After testing Proxmox for several weeks we think this might be a good idea for some of our services to rely on.
We've accomplished to archive the following so far:
Questiontime:
Appreciated
we're planning to integrate a Proxmox/Ceph Instance into our production-enviroment. After testing Proxmox for several weeks we think this might be a good idea for some of our services to rely on.
We've accomplished to archive the following so far:
- Setup 2 Bladecenters in two independend Datacenters for the beginning (Interconnected through 2x40 Gbit)
- Reserved 6 Nodes in each BladeCenter (each has 2x E5-2620 v3, 128GB Ram, HW-Raid 1TB Intel DC SSDs, 10 GBit LACP)
- Nodes are interconnected using 4 VLANs for Mgmt, Corosync, VM Migration and CephStorage
- Setup CEPH-Crushmap to accomplish our storage and redundancy needs
- We CANNOT use another Datacenter for now, so we have to stick with 2 locations.
- Having HA throughout our complete setup (if one Datacenter dies the other 6 Nodes still get quorum)
- If we encounter a splitbrain, both BladeCenters with 6 Nodes are getting fenced
- We can run with just one-side of our 12 Nodes (so just 6 Nodes active).
- Give each Node in DC 1 a Quorumvote +2
- Give each Node in DC 2 a Quorumvote +1
Questiontime:
- Is it possible to have more Quorumvotes than the number of nodes in our setup?
- Is this a decent tactic / practice?
- Is there anyone else who would like to share their experience with a large scale setup?
Appreciated
