Standard ConfigurationΒΆ

A configuration designed for scalable and highly available setups.

../../_images/starlingx-deployment-options-controller-storage.png

Note

With Rook, the storage function can be distributed over controllers and/or workers in this configuration.

See Deployment Configuration Terminology for a description of common deployment configuration terminology, including definitions of different node types and network types.

  • Architecture:

    Includes 2 controller nodes, up to 8 storage nodes, and up to 100 worker nodes.

  • Typical Deployment:

    • Ideal for large-scale, high-performance environments.

    • Suitable for environments requiring flexible resource allocation and modular scaling.

  • Storage Options:

    • A local, internal Rook-Ceph cluster deployed across controller or worker nodes using either the Controller, Dedicated, or Open Rook deployment model.

      • Controller Model (default):

        • For a small-scale storage solution.

        • OSDs are placed only on controller nodes.

        • Replication factor is configurable up to size 3 for Standard controller deployments as data Replication is between controller hosts.

      • Dedicated Model:

        • OSDs are placed only on a subset of worker nodes that are fully dedicated to storage, not to hosting containerized applications.

        • Replication factor configurable up to size 3.

      • Open Model:

        • No restrictions on OSD placement.

          Typically, all or most worker nodes host containerized applications and participate in the local storage function, with one or more disks used as Ceph OSDs.

        • The replication factor has no limitations.

    • External storage backends can be configured as an alternative or in addition to the local Rook-managed Ceph cluster. Supported external backends include, for example:

      • External NetApp Trident backend

      • External Dell Storage Array backend

  • Redundancy:

    • HA controller services run across the two controller nodes in Active/Standby mode.

    • The Rook-managed Ceph Cluster provides HA through replication in its deployment models. However, for AIO-SX deployments, an HA Ceph monitor configuration is not possible because there is only a single host. HA for Ceph data is only possible if multiple OSDs are configured on that host.

    • In the event of an overall controller node failure, all controller HA services become active on the remaining healthy controller node, and the above-mentioned Rook-managed Ceph replication protects the Kubernetes PVCs.

    • In the event of a worker node failure, hosted application containers on the failed worker node are recovered on one or more remaining healthy worker nodes.

    • Network link redundancy can be improved by deploying redundant Top-of-Rack (ToR) switches. Each link in a link aggregate is connected to a different switch, ensuring connectivity if one switch fails. For more information, see Redundant Top-of-Rack Switch Deployment Considerations .