• Start

High availability

How fault tolerance and availability differ between SurrealDB Cloud Start and Scale plans.

High availability (HA) means your database stays reachable and consistent when individual components fail or during planned maintenance. What you get on SurrealDB Cloud depends on your plan and instance topology.

Start instances run SurrealDB on a single compute node with dedicated storage. This model is simple and cost-effective for development, staging, and workloads that tolerate brief downtime.

  • There is no multi-node failover within the instance — if the node, disk, or underlying host fails, connections drop until the platform recovers it.

  • Pausing an instance stops compute deliberately; resuming brings it back on the same configuration.

  • Protect production data with automated backups and tested restore procedures rather than assuming in-instance redundancy.

Start on Cloud adds managed patching, snapshots, and operations without you running the server. For workloads that must keep serving through hardware or zone failures, use Scale or architect retries and failover at the application layer.

Scale clusters run multiple SurrealDB query nodes against SurrealDS distributed storage. HA is built into the architecture:

  • Multi-node clusters — Scale starts at three nodes and you can add more horizontally as throughput grows.

  • Survive node loss — SurrealDS coordinates replication and quorum consensus so the cluster keeps serving when one node is unavailable, instead of treating three separate single-node databases.

Scale’s three-node minimum leaves headroom during failures, upgrades, and maintenance while one node is out of service.

Scale is the managed option when you need fault-tolerant production without operating your own Kubernetes cluster, storage replication, and upgrade runbooks. SurrealDB Cloud provisions the cluster, applies patches, manages backups, and coordinates rolling version upgrades.

See Cloud architecture and Scaling.

RequirementStartScale
Accept brief maintenance or node recovery windowsSuitableSuitable
Survive node failure without manual failoverLimitedBuilt-in (multi-node + SurrealDS)
Horizontal query scaleResize vertically onlyResize vertically or add nodes

Was this page helpful?