Tinybird Cluster Management

Product · SaaS / cloud software · Operations & supply chain by Tinybird

Keep the controls. Skip the cluster ops.

Tinybird Cluster Management gives enterprise teams direct control over dedicated ClickHouse infrastructure. It enables on-demand resizing and workload rebalancing through a UI or API while Tinybird handles the underlying operations. Choose it for production-scale analytics that require high availability and custom capacity planning.

Category: Data Warehouse

Deployment
Cloud (AWS, GCP) and Dedicated Infrastructure
Platforms
Web, API
APIs and SDKs
Organizations API
Default Replicas
2 replicas
Scaling Type
Vertical and Horizontal
Infrastructure Type
Dedicated Infrastructure
  • Add or remove dedicated replicas on demand
  • Rebalance read and write weights across replicas
  • Zero-downtime resizing with active old replicas
  • Query live CPU, memory, and error signals
  • Automate changes via the Organizations API

Overview

Tinybird Cluster Management is a dedicated infrastructure control service for enterprise teams using the Tinybird data platform. It provides direct control over ClickHouse clusters without the operational burden of manual database management. Users can add or remove replicas, rebalance workloads, and automate cluster changes through a guided UI or the Organizations API. The service is designed for large-scale production workloads, allowing teams to plan for traffic peaks and distribute query loads independently. Tinybird operates the underlying platform, including infrastructure monitoring and recovery, while the customer owns capacity and workload decisions. This model ensures high availability and performance for mission-critical analytical features.

Cluster Operations and Control

  • On-demand Resizing: Users can add or remove dedicated replicas through the Organization UI or with a POST request to the Organizations API.
  • Safe Resize Workflow: Tinybird provisions new replicas while existing ones continue to serve traffic, syncing metadata and warming caches without data copying.
  • Workload Distribution: The platform allows teams to distribute query and copy-job work independently by setting specific read and write weights per replica.
  • Cache Warming: New replicas warm their local performance cache from durable object storage while serving active workloads.

Observability and Monitoring

  • Workload Observability: Teams can query CPU, memory, errors, and replica queues to make informed capacity decisions based on live signals.
  • Custom Thresholds: Users can apply their own thresholds and timing for capacity changes based on observed workload signals.
  • Capacity Planning: The service allows for planning known peaks by adding capacity before launches or seasonal traffic events.
  • Metrics Integration: Workload and capacity signals can be monitored by querying organization.metrics_logs or scraping the Prometheus metrics endpoint.

Managed Platform Services

  • Automated Recovery Management: The platform manages storage and recovery processes while the customer retains ownership of capacity and workload decisions.
  • Durable Storage Management: Object storage remains the durable source of truth, ensuring data durability during cluster resizing.

Use cases

  • Enterprise teams add dedicated replicas to handle planned traffic peaks during product launches.
  • Data engineers rebalance read and write weights across replicas to optimize query throughput.
  • Operations teams automate cluster resizing using the Organizations API based on custom CPU utilization thresholds.
  • Software teams validate new capacity by keeping old replicas active during the resizing workflow.

Specifications

Deployment
Cloud (AWS, GCP) and Dedicated Infrastructure
Platforms
Web, API
APIs and SDKs
Organizations API
Limits
2 replicas included by default on Enterprise plans; unlimited vCPUs on dedicated clusters
Default Replicas
2 replicas
Scaling Type
Vertical and Horizontal
Infrastructure Type
Dedicated Infrastructure
Control Interfaces
Organization UI, Organizations API
Compute Allocation
Unlimited vCPUs (Enterprise)
Storage Type
SSD + Cloud Storage (Enterprise)
Storage Capacity
Bottomless storage (Enterprise)

How it works

  1. Observe Workload Signals: Monitor cluster health through average CPU utilization and other live signals in the Organization UI or via API.
  2. Provision New Capacity: Decide on capacity needs and add a replica using a POST request to the Organizations API or through the management interface.
  3. Automatic Provisioning and Sync: Tinybird provisions the new replica and syncs metadata while existing replicas continue serving traffic.
  4. Rebalance and Verify: Once the new replica is caught up, rebalance read and write weights across all replicas to return utilization to healthy ranges.

FAQ

Does Tinybird automatically scale my cluster?

No, Tinybird does not automatically scale clusters. Users decide when and how to resize their dedicated cluster using the Organization UI or the Organizations API.

Added by AntSpace from public sources (tinybird.co).

Related

More products by Tinybird