Error Handling
Robust pipelines must anticipate and gracefully handle failures.
Strategies
- API Failures: Retry with exponential backoff. Alert if persistent.
- Validation Failures: Log bad records, do not fail the entire batch unless error rate > threshold (e.g., 5%).
- Database Locks: DuckDB is single-writer. Ensure pipeline orchestrator prevents concurrent write runs.
- Notifications: Send alerts to a webhook (Slack/Discord) if a pipeline step fails.