Skip to main content
Ingester receives telemetry from monitored clusters and writes it to ClickHouse.

​
Talks To

​
How It Fits

Ingester is the write path for telemetry. Exporters in monitored clusters send telemetry to the hub ingress path, and that traffic is routed to Ingester for validation, normalization, batching, and storage. The service writes the resulting metrics, traces, logs, profiling data, Kubernetes resource updates, and usage records into ClickHouse. It uses the authenticated request context to attach organization and environment information to incoming data. Ingestion is more CPU intensive than the UI and API request path. Ingester is split out from Apiserver so telemetry processing can scale independently without forcing the control-plane API to carry the same CPU profile.

​
Operational Notes

Ingester depends on ClickHouse for durable telemetry storage. If ClickHouse is unavailable or slow to accept writes, ingestion latency and backlog symptoms will show up at this layer. Ingester is stateless and is intended to scale horizontally as telemetry volume increases. Operators should watch request latency, dropped-record signals, memory readiness, and ClickHouse insert health when diagnosing ingestion problems.