RetryPilot

RetryPilot is a web app for teams running Kafka, RabbitMQ, SQS, or Pub/Sub who keep getting burned by retries, dead-letter queues, and “poison pill” messages. It connects read-only to your brokers and consumer logs, then builds a unified view of failure hotspots: which message types fail, which consumers loop retries, and where backoff policies are misconfigured. The product’s core is a rules-and-simulation workspace: define retry policies (max attempts, jitter, exponential backoff, DLQ routing, quarantine topics), simulate impact on throughput and lag, then export policy templates and alerting thresholds. It also provides a lightweight sidecar/SDK to standardize error metadata (failure reason, idempotency key, payload hash) so debugging isn’t guesswork. This is not a full APM replacement; it’s narrowly focused on message failure handling and operational hygiene, where teams lose real time and money.