A GDCE zone is a rack or group of racks sharing space, power, cooling, and a contiguous network fabric as one failure domain. It differs from a GCE compute zone and is not returned by GCE zone listings. section 8.3.1
Use the zone boundary to reason about placement and resilience:
- Treat each zone as a single failure domain. section 8.3.1
- Deploy applications across multiple zones in the same region to improve high availability and limit the effect of zone failures. section 8.3.1
- Associate physical machines with their designated region; machines can join clusters deployed only in that region. section 8.3.1
Rack-based cluster layout
A GDCE Kubernetes cluster uses a GCP-hosted control plane and worker nodes running on physical servers in GDCE racks. Spreading workloads across separate zones separates them across failure domains. section 8.3.1
Use the resource-model figure to relate zones to machines, clusters, and node pools. section 8.3.1
