Proxmox / Ceph HA Platform

Proxmox works well when the surrounding design is solid: storage, quorum, networking, backups, upgrades, and failure handling.

This mission covers new clusters, migrations from vSphere or older platforms, and reviews of existing Proxmox/Ceph setups before they become production-critical.

When this is NOT a fit

• You need a fully managed hypervisor service and do not want to operate storage, quorum, backups, and hardware lifecycle internally.
• The available network cannot provide predictable low-latency paths for Corosync, migration traffic, and Ceph replication.
• The hardware mix is too uneven to make capacity planning, failure domains, or performance baselines reliable.
• There is no maintenance window or test environment for validating failover, restore, and node replacement procedures.

Common failure modes we’ve seen

• Corosync shares congested production networks, leading to quorum instability and split-brain risk during ordinary switch or host issues.
• Ceph is sized from raw disk capacity instead of usable capacity, recovery bandwidth, OSD memory, and failure-domain math.
• HA is enabled before fencing, backups, and restore tests are proven, so an automated restart can make a storage or network incident worse.
• Cluster upgrades are treated like package updates instead of coordinated platform changes with migration, rollback, and workload placement rules.