IntegrationIntermediatemulesoftretryreliability
Design retry policies with exponential backoff for MuleSoft Salesforce connectors
Real World Scenario
Transient Salesforce 503 errors during maintenance cause MuleSoft flows to fail permanently; ops manually replays 200 failed orders nightly.
Expected Answer
• Until-successful retry with exponential backoff: 1s 2s 4s 8s max 5 attempts
• Retry only on transient HTTP codes 429 503 504 not 400 401
• Idempotency key on upsert preventing duplicate orders on retry
• Dead letter queue after max retries with ops alert
• Jitter on backoff preventing thundering herd on recovery
• Separate retry policy for read vs write operations
• Dashboard retry rate trending indicating systemic issues
Follow-Up Questions & Answers
Click to expand — each follow-up includes a direct, interview-ready answer
Direct answer: Until-successful retry with exponential backoff: 1s 2s 4s 8s max 5 attempts Also consider: Retry only on transient HTTP codes 429 503 504 not 400 401 In practice: Idempotency key on upsert preventing duplicate orders on retry Balance speed of delivery with maintainability.
Architect Perspective
Retry without idempotency creates duplicates—pair backoff with external ID upsert always.