Data Platform
Keep the controls.
Skip the cluster ops.
and automate cluster changes while Tinybird operates the underlying platform.
A Tinybird cluster runs two replicas. Load rises until both are saturated. The customer adds a third replica from the cluster management interface or with a POST to the Organizations API, and Tinybird provisions it while the first two keep serving traffic. Once the new replica is caught up, read and write weights are rebalanced across three replicas and utilization returns to a healthy range.
Observe. Decide. Resize. Verify.
# Check cluster health from average CPU utilization
Choose your operating model
The change happens
when you decide it does
while your team stays in control of when and how the cluster changes.
Self-managed
Maximum control because your team owns the entire ClickHouse stack.
- Host and operate the entire stack
- Build and maintain cluster runbooks
- Execute every capacity change yourself
- Diagnose and recover failures
Tinybird managed
Direct cluster control on Dedicated infrastructure without taking on ClickHouse operations.
- Inspect workload and capacity signals
- Resize through the Organization UI or API
- Decide the timing while we operate the platform
Conventional managed
The provider operates the stack and controls the change process.
- Request changes through support
- Wait for the provider to act
- Follow the provider's schedule
- Work within exposed controls
Customer proof
Production scale, without
operating ClickHouse
serves over 1 billion requests per month for millions of developers.
per month
per month
latency
“Tinybird is the most exciting data company since Snowflake. It's revolutionized the way we think about real‑time data analytics at Vercel.”
Guillermo Rauch
CEO at Vercel
Built for control
Cluster control without
ClickHouse ops
and resize through a supported workflow.
Observe before changing
- Query CPU, memory, errors, and replica queues
- Review recent service activity
- Apply your own thresholds and timing
Plan known peaks
- Add capacity before launches or seasonal traffic
- Use guided UI changes or customer-authored automation
- Keep capacity decisions explicit
Distribute workloads
- Set read weights by replica
- Set write weights by replica
- Route copy-job work independently
Resize with a safe flow
- Add, reroute, verify, then remove
- Keep old replicas active while validating new capacity
- Plan for temporary dual capacity and cost
Keep storage durable
- Object storage remains the source of truth
- Local storage acts as a performance cache
- New replicas warm their cache while serving workloads
Leave operations to Tinybird
- Tinybird monitors infrastructure and responds to operational alerts
- Tinybird manages storage and recovery
- Your team owns capacity and workload decisions
FAQs
Do I need to self-host ClickHouse for cluster control?
No. You control supported Dedicated capacity and workload distribution while Tinybird operates ClickHouse and the surrounding platform. Compare infrastructure models
Can I configure cluster alerts?
Tinybird monitors the underlying infrastructure and responds to operational alerts to keep the service available. Cluster Management does not currently provide customer-configurable alert rules. To monitor workload and capacity signals, query organization.metrics_logs or scrape the Prometheus metrics endpoint into your observability platform, where you can define your own thresholds and notifications. Review cluster metrics
Does Tinybird automatically scale my cluster?
No. You decide when and how to resize your Dedicated cluster. Tinybird monitors the underlying infrastructure and responds to operational alerts to maintain availability. Our team can help review sustained capacity pressure and recommend changes, but capacity changes remain visible and under your control. Use the Organization UI or the Organizations API to apply them. Read the cluster management guide
What happens to service and data while resizing?
The documented resize workflow is designed to keep the cluster serving traffic while capacity changes. Object storage remains the durable source of truth while new replicas warm their local cache. Add the new replicas, reroute workloads, verify the new configuration, and only then remove the old replicas. Review the resize workflow
Can I manage clusters through both the UI and API?
Yes. Organizations running on Dedicated infrastructure can manage replicas and workload distribution through Organization settings or the Organizations API. View the Organizations API
Can I automate capacity changes?
Yes. You can use the Organizations API to build workflows around the signals and thresholds that matter to your team. Tinybird does not automatically resize the cluster for you: your automation decides when to initiate each change, and the result remains visible through the Organization UI and API. View the Organizations API
How do capacity changes affect cost?
Dedicated billing reflects active infrastructure and related resources. A safe resize temporarily runs old and new replicas together, so expect temporary dual capacity and cost until the old replicas are removed. Review the calculator and your agreement for pricing details. Review Enterprise pricing
Plan and operate your Dedicated cluster
- Choose between Shared and Dedicated infrastructure.
- Inspect and resize clusters with the cluster management guide.
- Automate cluster operations with the Organizations API.
- Monitor resource usage through Service Data Sources.
