ScyllaDB on Kubernetes (ScyllaDB Operator)

ScyllaDB Operator is the official Kubernetes Operator for deploying and managing ScyllaDB clusters. It introduces custom resources, most importantly ScyllaCluster, that let Kubernetes manage ScyllaDB the way it manages any other workload, while still respecting ScyllaDB's own requirements around CPU pinning, local NVMe storage, and rack/zone topology.

This skill covers the official Operator model (the ScyllaCluster, NodeConfig, and ScyllaOperatorConfig custom resources, installed via Helm or GitOps manifests). It does not teach any one team's custom deployment scripts, several exist in the community, including a detailed sed-templated bash pipeline in the tluck/Scylla-K8s-Example repository, and those are useful references for a working end-to-end example, but they are one practitioner's tooling choices, not ScyllaDB's own documented interface. When a user's question is about the underlying Operator resources and requirements, ground the answer in this skill; when it's specifically about someone else's deployment scripts, say so explicitly rather than presenting their conventions as official.

When to Use This Skill

• The user is deploying ScyllaDB on Kubernetes for the first time (EKS, GKE, OpenShift, or generic)
• The user is writing or reviewing a ScyllaCluster manifest
• The user is configuring node-level prerequisites: CPU manager policy, local storage, node labels
• The user is troubleshooting a ScyllaDB Operator installation, upgrade, or a ScyllaCluster that won't reach Available
• The user references ScyllaDB Manager running inside Kubernetes (it ships as a sidecar container in each ScyllaDB pod, not a separate install)
• The user's application can't reach the cluster, or they're planning external/cross-datacenter client access (exposeOptions, load balancers, broadcast addresses)
• The user is securing a cluster: authentication, the default superuser, TLS/mTLS encryption, or extracting client certificates
• The user is scaling, adding/removing racks, or building a multi-datacenter cluster with externalSeeds

For schema design once the cluster is running, defer to scylladb-data-modeling. For DynamoDB-API access, defer to scylladb-alternator, Alternator can be enabled on a Kubernetes-deployed cluster the same as anywhere else.