Purpose
This is the top-level gate: the review that happens once, before a cluster stops being a project and starts being infrastructure other teams depend on. The narrower checklists in this course each cover one surface in depth. This one covers the whole cluster shallowly and exists to catch the thing nobody owned.
Order of work
The list runs in three passes. The daemon pass is mon-health, mgr-active
and osd-state: the control plane and the data plane, in that order, because
a cluster with a marginal quorum has no business being assessed on anything
else. The layout pass is crush and pool-policy, which together are the
cluster’s durability story - the map has to match the room, and every pool has
to be bound to a rule that fits what it holds. The operations pass is
monitoring, backup, dr and runbooks: the four items that are not
about the cluster at all but about whether anyone can run it at three in the
morning.
Before you start
Have the rack diagram, the pool inventory with owners, the on-call rotation the alerts will reach, and the date of the last successful restore.
Verification
Production readiness is a signed decision, not a passing score. Every critical item is checked, every warn item has an owner and a date, and the storage lead signs the result. Mark an item N/A rather than leaving it unchecked when it does not apply to this cluster.
Escalation
Any critical item that cannot be satisfied blocks the handover. Take it to the storage lead and the service owner together.