Score OpenShift cluster health from node readiness, pod failures, etcd fsync latency, API latency and degraded operators.
The score is a weighted blend of the signals that actually predict an OpenShift outage. etcd fsync latency carries as much weight as node readiness because etcd is the single point of failure for the whole control plane — once fsync exceeds 10 ms the leader loses elections and every API call starts timing out, regardless of how healthy the workers look. Symptoms surface far from their cause in Kubernetes: slow etcd disks present as failing deployments, and a degraded operator presents as an upgrade that silently will not start. Scoring the signals together points at the layer that is actually broken.
OpenShift Cluster Health
score = node readiness × 0.25 + pod health × 0.20 + etcd latency score × 0.25 + API latency score × 0.15 + operator score × 0.10 + storage score × 0.05.
score = node readiness × 0.25 + pod health × 0.20 + etcd latency score × 0.25 + API latency score × 0.15 + operator score × 0.10 + storage score × 0.05. The score is a weighted blend of the signals that actually predict an OpenShift outage. etcd fsync latency carries as much weight as node readiness because etcd is the single point of failure for the whole control plane — once fsync exceeds 10 ms the leader loses elections and every API call starts timing out, regardless of how healthy the workers look.
Symptoms surface far from their cause in Kubernetes: slow etcd disks present as failing deployments, and a degraded operator presents as an upgrade that silently will not start. Scoring the signals together points at the layer that is actually broken.
This calculator takes 8 inputs: Nodes in cluster, Nodes not ready, Pods scheduled, Pods not running or ready, etcd 99th percentile fsync, API server 99th percentile latency, Degraded cluster operators, PersistentVolumeClaims pending. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
etcd must durably commit every write to its write-ahead log before acknowledging it, so fsync latency is the floor for every API write. Beyond 10 ms at the 99th percentile, heartbeats miss their deadlines and the cluster starts electing new leaders under normal load.