Operator configuration

This page documents the process-level configuration of the Momento Valkey Operator: the valkey-operator-config ConfigMap, how its keys reach the operator, and how to change them. It is for platform teams who own the operator installation.

The operator installation (operator.yaml) creates the ConfigMap valkey-operator-config in the valkey-operator namespace. The ConfigMap is optional: if it is absent, the operator runs with compiled-in defaults.

Configuration keys

Key	Values	Default	Effect
defaultZoneSpread	bestEffort | required	bestEffort	The operator-wide default for zone spreading of Valkey pods, used by every ValkeyCluster that does not set placement.zoneSpread itself. bestEffort schedules pods even when zone spread cannot be satisfied; required leaves pods Pending until a valid zone placement exists. See Zone-aware placement.

A per-cluster placement.zoneSpread always overrides this default. The value spellings are identical in the ConfigMap and in the ValkeyCluster field: exactly bestEffort or required.

Environment variable mapping

The operator reads its settings from environment variables; the Deployment wires each ConfigMap key to its variable:

ConfigMap key	Environment variable
defaultZoneSpread	DEFAULT_ZONE_SPREAD

The Deployment marks the reference optional, which is why a missing ConfigMap or key falls back to the compiled default rather than blocking startup.

The Deployment also sets RUST_LOG=info, which controls the operator's JSON log verbosity. It is plain environment configuration, not a ConfigMap key. See Monitoring.

Changes require an operator restart

The operator reads configuration once at startup and holds it for the life of the process: there is deliberately no live reload. Changing the ConfigMap alone has no effect on a running operator. To apply a change, patch the ConfigMap and restart the operator Deployment:

kubectl patch configmap valkey-operator-config -n valkey-operator \
  --type merge -p '{"data":{"defaultZoneSpread":"required"}}'

kubectl rollout restart deployment/valkey-operator -n valkey-operator
kubectl rollout status deployment/valkey-operator -n valkey-operator

Running Valkey clusters keep serving traffic during the restart; reconciliation pauses briefly and resumes when the new operator pod is up.

注記

A changed defaultZoneSpread affects only pods created after the restart. Existing pods keep the spread behavior they were scheduled with; the operator does not roll clusters for a zone-spread change. See Zone-aware placement.

Invalid values fail startup

If DEFAULT_ZONE_SPREAD is set to anything other than bestEffort or required, the operator logs the error and exits at startup, and the Deployment goes into a crash loop. This is deliberate fail-fast behavior: a misconfigured operator does not start, rather than running silently with a wrong or guessed default.

If you encounter this after a configuration change, fix the ConfigMap value and restart again:

kubectl logs deployment/valkey-operator -n valkey-operator
kubectl patch configmap valkey-operator-config -n valkey-operator \
  --type merge -p '{"data":{"defaultZoneSpread":"bestEffort"}}'
kubectl rollout restart deployment/valkey-operator -n valkey-operator

Existing Valkey clusters continue serving while the operator is down; no reconciliation (scaling, replacement, failover recovery) happens until it is healthy again. See Failure modes.