Valkey provider moves to valkey-operator v0.9.0 and validates instances at admission time
provider-valkey v0.1.11 moves the provider onto valkey-operator v0.9.0, and adds validation that catches misconfigurations before anything is scheduled.
valkey-operator v0.9.0 (breaking): the provider now runs on valkey-operator v0.9.0, so OpenEverest-managed Valkey picks up the operator’s newest behavior: ACL users move to spec.acl.users, the Valkey default user is off unless a cluster declares it, replicasPerShard replaces patching replicas, a cluster can expose a Service per node and a separate client CA for mTLS, and a crashed primary’s data is kept when its shard has a replica. The bump is marked breaking for a reason: the operator’s upgrade notes require applying both v0.9.0 CRDs before upgrading, because the older CRDs prune the fields the new operator reads — including the ACL revision each node has loaded, and networking.perNodeServices — and a replicas to replicasPerShard patch against the old CRD leaves a cluster with zero replicas.
Admission-time validation: instance specs are now validated through an admission webhook that returns pure validation results with exact field paths, so a misconfigured instance is rejected at apply time, with the error pointing straight at the offending field.
The release also brings a first contribution from @chilagrow — exploring cluster-scoped instance presets (#27, reverted in #37) — and the provider’s CI now tracks new valkey-operator releases through Dependabot (#40), so future operator updates reach the provider sooner.
To learn more, visit the v0.1.11 release notes.
