High Availability Deployment

The Highly Available (HA) Deployment of MSR 4 is distributed across three or more worker nodes, ensuring resilience and reliability through multiple service instances. For installation guidance, refer to the Install MSR with High Availability.

A key aspect of this deployment is that Job Service and Registry utilize a shared volume, which should be backed by a non-local, shared file system or external storage cluster, such as Ceph (CephFS). Additionally, Redis and PostgreSQL run in a replicated mode within this example, co-hosted on the same worker nodes as MSR 4’s core services. However, it is also possible to integrate existing corporate Redis and PostgreSQL instances outside of these nodes, leveraging an enterprise-grade key-value store and database infrastructure.

The following diagram illustrates the service placement in an HA deployment. Dashed boxes indicate potential additional replicas for certain services. As a reference, we recommend deploying at least two instances of Portal, Core, Job Service, Registry, and Trivy—though this number can be adjusted based on specific requirements, workload, and use cases. These services are not quorum-based.

While the number of replicas for these services can scale as needed, Redis and PostgreSQL must always have a minimum of three replicas to ensure proper replication and fault tolerance. This requirement should be carefully considered when planning a production deployment. Redis and PostgreSQL are quorum-based services, so the number of replicas should always be odd, specifically 1, 3, 5, and so on.

The reference HA deployment of an MSR 4 is presented in the following diagram.