Large single-replica services

Understand availability and scaling for large single-replica services in ClickHouse Cloud

Large single-replica services let you run a ClickHouse Cloud service with one compute replica while using the configurable compute resources available on the Scale and Enterprise tiers. Unlike fixed-size Basic tier services, their scaling behavior depends on the tier and hardware profile.

Availability

Large single-replica services are available on the Scale and Enterprise tiers. Basic tier services also use a single replica, but are fixed at 8 GiB or 12 GiB of total memory and don’t support vertical or horizontal scaling.

On the Scale and Enterprise tiers, you can create a primary or secondary service with one replica without contacting support. Services in a warehouse scale independently.

When to use a single-replica service

A large single-replica service can reduce compute costs for workloads that need more CPU or memory than a Basic tier service but don’t require compute redundancy. It can also be useful for development, testing, batch processing, and other non-critical workloads.

Single-replica services have reduced availability and aren’t recommended for production workloads that require high availability. Use at least two replicas for high availability.

Scaling

Vertical scaling behavior depends on the tier and hardware profile:

• Scale tier services and Enterprise tier services with standard 1:4 profiles support vertical autoscaling. You can configure minimum and maximum CPU and memory resources in the Cloud Console.
• Enterprise tier services with custom profiles don’t support vertical autoscaling or manual vertical scaling in the Cloud Console. Contact support for vertical resizing.

Scale and Enterprise tier services support horizontal scaling, so you can scale out to multiple replicas and scale back in to one replica.

Service limits

Single-replica services are limited to 100 databases and 500 tables on every tier, including Scale and Enterprise. Other service guardrails (views, dictionaries, columns, partitions, parts, and so on) follow the standard limits. For details, see Usage limits.