# What Is ScyllaDB Operator? ScyllaDB Operator is a [Kubernetes Operator](https://kubernetes.io/docs/concepts/extend-kubernetes/operator/) that automates the deployment and lifecycle management of [ScyllaDB](https://www.scylladb.com) clusters on Kubernetes. ## What problems does it solve? Running ScyllaDB on Kubernetes without an operator requires manually handling many tasks that the Operator automates: | Concern | Without the Operator | With ScyllaDB Operator | |-------------------------|----------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------| | **Provisioning** | Manually create StatefulSets, Services, ConfigMaps, PVCs, and configure ScyllaDB seed lists, snitch settings, and rack/DC placement. | Declare a `ScyllaCluster` resource. The Operator creates and configures all Kubernetes objects. | | **Scaling** | Manually adjust StatefulSet replicas, update seed lists, and run `nodetool cleanup` on existing nodes after the new node finishes streaming. | Change the `members` count in the spec. The Operator handles the StatefulSet update and runs cleanup automatically when token ownership changes. | | **Upgrades** | Manually update the container image, coordinate a rolling restart, and verify each node rejoins before proceeding to the next. | Update the image field in the spec. The Operator performs a rolling upgrade one node at a time, verifying health at each step. | | **Node replacement** | Manually drain the node, remove it from the cluster, recreate the pod, and pass the `replace_address_first_boot` flag. | Label the node’s Service for replacement. The Operator handles draining, decommissioning, and bootstrapping the replacement. | | **Performance tuning** | SSH into each Kubernetes node to run `perftune`, configure IRQ affinity, set sysctls, and prepare RAID/filesystem layouts. | Create a `NodeConfig` resource. The Operator deploys privileged DaemonSets and Jobs that tune each node and prepare disks automatically. | | **Repairs and backups** | Install and configure ScyllaDB Manager separately, register clusters manually, and schedule tasks. | Define repair and backup tasks in the cluster spec or as `ScyllaDBManagerTask` resources. The Operator deploys Manager and registers clusters. | | **Monitoring** | Deploy and configure Prometheus, Grafana, ServiceMonitors, and dashboards manually. | Create a `ScyllaDBMonitoring` resource. The Operator provisions the monitoring stack with ScyllaDB-specific dashboards. | | **TLS** | Generate and distribute certificates, configure ScyllaDB to use them, and handle rotation. | Enable TLS in the spec. The Operator uses cert-manager to provision and rotate certificates automatically. | ## How it works ScyllaDB Operator extends the Kubernetes API with [CustomResourceDefinitions (CRDs)](https://kubernetes.io/docs/concepts/extend-kubernetes/api-extension/custom-resources/) and runs controllers that continuously reconcile the desired state you declare with the actual state of the cluster. The Operator ships as a single binary that includes: - **Controllers** that watch CRDs and create or update the underlying Kubernetes objects (StatefulSets, Services, ConfigMaps, Jobs, PodDisruptionBudgets, and more) that orchestrate the ScyllaDB clusters you define. - **Admission webhooks** that validate configuration correctness. - **A sidecar** that runs inside each ScyllaDB pod to configure and manage the ScyllaDB process. - **Node setup agents** that run on Kubernetes nodes to prepare disks and tune the host. ## Single-DC and multi-DC deployment models `ScyllaCluster` is the stable (`scylla.scylladb.com/v1`) API for deploying ScyllaDB on Kubernetes. Each `ScyllaCluster` resource represents a single datacenter. You define racks within the datacenter, and the Operator creates a StatefulSet per rack. The Operator automates all single-datacenter operations: provisioning, scaling, upgrades, node replacement, tuning, repairs, backups, and monitoring. ### Multi-datacenter clusters To deploy a ScyllaDB cluster that spans multiple Kubernetes clusters or geographic regions, create one `ScyllaCluster` resource in each Kubernetes cluster and connect them using the `externalSeeds` field. Each `ScyllaCluster` must use the same `.metadata.name` (which becomes the ScyllaDB cluster name) and specify seed node addresses from the other datacenters in `externalSeeds`. The Operator manages each datacenter independently. The following must be managed manually when using multiple `ScyllaCluster` resources for multi-datacenter deployments: - **Seed configuration** — the `externalSeeds` field of each `ScyllaCluster` must be populated with the node addresses of the other datacenters. These addresses must be kept up to date if nodes are replaced or IPs change. - **ScyllaDB Manager auth token synchronization** — when ScyllaDB Manager is deployed in one datacenter, its auth token Secret must be manually copied to each other datacenter and the Manager Agent pods restarted. See [ScyllaDB Manager](https://operator.docs.scylladb.com/v1.22/understand/manager.md). - **Replication factor configuration** — keyspace replication factors must be updated via CQL after adding a new datacenter, using `ALTER KEYSPACE`. - **Network peering and firewall rules** — inter-cluster Pod-to-Pod connectivity must be configured at the infrastructure level. ## Key user-facing resources The Operator provides the following custom resources: | Resource | Scope | Purpose | |------------------------|------------|------------------------------------------------------------------------------------------------------------------------------| | `ScyllaCluster` | Namespaced | A ScyllaDB datacenter (one resource per DC; for multi-DC, create one per Kubernetes cluster and connect via `externalSeeds`) | | `NodeConfig` | Cluster | Node-level disk setup, RAID, filesystem, and performance tuning | | `ScyllaOperatorConfig` | Cluster | Global Operator configuration (images, cluster domain) | | `ScyllaDBMonitoring` | Namespaced | Monitoring stack (Prometheus + Grafana) for ScyllaDB | | `ScyllaDBManagerTask` | Namespaced | Backup or repair task managed by ScyllaDB Manager | The Operator also uses several internal CRD types under the hood to orchestrate operations such as datacenter management, Manager cluster registration, and bootstrap synchronisation. These are not intended for direct user interaction. ## Supported platforms ScyllaDB Operator is tested and supported on: - **Google Kubernetes Engine (GKE)** with Ubuntu node images - **Amazon Elastic Kubernetes Service (EKS)** with Amazon Linux node images - **Oracle Kubernetes Engine (OKE)** - **Architectures**: amd64 and arm64 For the full support matrix including supported Kubernetes versions, ScyllaDB versions, and known platform incompatibilities, see [Releases](https://operator.docs.scylladb.com/v1.22/reference/releases.md). ## Next steps - [Install](https://operator.docs.scylladb.com/v1.22/install-operator/index.md) ScyllaDB Operator in your Kubernetes cluster. - [Deploy your first cluster](https://operator.docs.scylladb.com/v1.22/deploy-scylladb/deploy-your-first-cluster.md) once the Operator is installed. - See a complete end-to-end walkthrough for [GKE](https://operator.docs.scylladb.com/v1.22/deploy-scylladb/reference-deployments/reference-deployment-gke.md), [EKS](https://operator.docs.scylladb.com/v1.22/deploy-scylladb/reference-deployments/reference-deployment-eks.md), or [OpenShift](https://operator.docs.scylladb.com/v1.22/deploy-scylladb/reference-deployments/reference-deployment-openshift.md). - Learn about the Operator’s [architecture](https://operator.docs.scylladb.com/v1.22/understand/index.md).