Every uncaught exception, structured and searchable.
Self-hosted Sentry-lite running on the 11 most critical Lambdas. Every unhandled exception writes a structured JSON record to S3 with traceback, request ID, log stream, and event shape. The platform tells on itself.
Error distribution
Where errors are concentrated. A healthy platform has empty rows everywhere — concentration on one Lambda points to a specific bug or upstream data issue.
Last 200 errors
Reverse chronological. Click a row to expand traceback excerpt. Full traceback + log stream link is in s3://justhodl-dashboard-live/errors/<lambda>/<timestamp>.json.
Errors recorded by the @track_errors decorator at the top of each instrumented lambda_handler. Detail records are write-only (one per error). The recent.json rolling tail keeps the last 200 summary records (no traceback to keep the file small). On record failure, the original exception is still re-raised so AWS records the Lambda failure metric — error tracking never blocks failure attribution.