Ingestor

The ingestor is the backend service that receives error events from the tracker and stores them in a columnar database. Each project gets its own isolated data store.

Storage

Error data is organized in two tiers for optimal performance and storage efficiency:

• Hot storage (recent events): Fast in-memory storage for the latest error data. Handles incoming writes at high throughput.
• Archived segments (older events): Compressed archives of older data, optimized for long-term storage. Events automatically rotate to archive as hot storage approaches size limits.

The rotation is size-based (not time-based), ensuring predictable and safe storage growth. Old data remains queryable through a unified view that transparently reads from both hot storage and archives.

Each project's data is fully isolated and never mixed with other projects.