Nest Engineering Docs
Datalake

Operations

Deployments, SLOs, and monitoring for Datalake

Deployments

  • Build from services/datalake/Dockerfile.
  • The container runs granian against services.datalake.main:app.
  • Default listen port is 8080 and honors the PORT environment variable.

Startup behavior

  • Loads DATALAKE_SENTRY_DSN from Secret Manager.
  • Initializes BigQuery client and Pub/Sub publisher/subscriber.
  • Starts a Pub/Sub subscription callback for change stream messages.

Monitoring and alerts

  • Sentry is required; startup fails if the DSN is missing.
  • Health endpoint exposes start_time and messages_processed.

SLOs and SLIs

  • TBD. Define throughput and error targets for change stream ingestion.

Scaling and performance

  • Stateless service with Pub/Sub flow control.
  • Ack deadlines are extended while processing large payloads.
  • BigQuery inserts retry up to three times per message.

Last updated on