1 · Cluster Profile
2 · Control Plane (masters + etcd) size-driven, workload-agnostic
Control plane sizing scales with cluster size (node/object count), not workload type. Red Hat guidance: 4/16 ≤25 workers, 8/32 ≤100, 16/64 ≤250, 32/128 beyond. etcd needs fsync p99 < 10 ms — NVMe/SSD only.
3 · Infrastructure Nodes ingress / monitoring / logging / registry
Move Ingress, Monitoring (Prometheus is heaviest), Logging and the internal Registry onto 3 dedicated infra nodes — isolates platform services and can remove them from subscription core counts. Hybrid clusters monitor more series, so size infra higher.
4 · Worker Pools
5 · Storage Backend
Mandatory design constraints — ground-up external Red Hat Ceph cluster:
Expert Service
Turn this plan into a validated build
Indicative cluster design from Red Hat published practices. NCINGA delivers reference architectures, sizing validation, installation and Day-2 operations for production OpenShift platforms.
VM Workloads — detailed sizing feeds the Virtualization pool
Node capacity (cores, RAM, node count, N+1) is inherited from your Virtualization worker pool(s) on the Cluster Design tab. Set overcommit, storage and templates here; totals flow back into the Used-vs-Available view.
Overcommit & Storage
VM Templates & Quantities
| Template | vCPU | RAM GiB | Disk GiB | Dedicated CPU | Storage | Qty | Mem OH | Eff RAM | Threads | Eff Disk |
|---|
Per-VM overhead (Red Hat): 0.002×RAM + 218 MiB + 8 MiB×vCPU + 16 MiB. Dedicated-CPU VMs cost vCPU+1 threads. Block RWX = raw SAN (no overhead, live-migratable); File RWX = live-migratable +FS overhead; File RWO = node-bound, blocks live migration.
VM Capacity Result
Object Storage (Backup & Import)
Object (S3) cannot host running VM disks — it sizes the backup/DR target (OADP/Velero) and MTV/CDI import staging. Set capacity to 0 to exclude.
Cluster Node & Capacity Roll-up
| Role / Pool | Nodes | vCPU/node | RAM/node | Total vCPU | Total RAM |
|---|
Worker Capacity — Used vs Available after reservations & N+1
Available = (cores×SMT − reserved) × (nodes − N+1), per pool. Reserved per node covers system/kube + DaemonSets, plus OpenShift Virtualization management on virtualization pools. Used = the workload demand entered on each pool. Enter demand on the Cluster Design tab to populate utilisation.
Red Hat Recommended Host Practices
● Must ● Recommended ● Optional / context
References
- Red Hat KCS 7132959 — OpenShift Control Plane Sizing Guidelines.
- Red Hat docs — Control plane node sizing & Recommended etcd practices (fsync p99 < 10 ms, SSD/NVMe).
- Red Hat docs — Recommended host practices; default 250 pods/node (configurable).
- Red Hat docs — Creating infrastructure machine sets; MachineConfigPools per node role.
- Red Hat OpenShift Virtualization — cluster resource requirements & live migration.