Speaker · Technical talk
AWS Community Day Perú 2026
When an external API fails: designing resilient integrations on AWS with ECS Fargate
Original title: Cuando una API externa falla: diseñando integraciones resilientes en AWS con ECS Fargate
I shared how to design integrations that recover from partial failures: record the state, distinguish temporary errors from business errors, and resume a process without repeating completed operations.
Leer en español
- Date
- Venue
- UTEC · Barranco, Lima
Inside the talk
Design for the moment something fails.
Resilience starts with knowing what happened, what is pending, and how to continue.
Recover from a partial failure
What happens when an external operation completes but the flow fails before finishing? Resuming from the pending step requires knowing what already happened.
Persist useful state
Recording the current step, attempt count, and external identifier makes it possible to reconstruct the process and determine the next action with traceability.
Retry with intent
Distinguish transient errors from business errors and consider idempotency to avoid repeating effects that have already occurred.
Separate application and infrastructure
ECS, Fargate, and CloudWatch support execution and observability. The application remains responsible for process state and recovery.
From concepts to code
A lab for exploring failures.
The companion repository includes a local integration demo and an architecture prepared for ECS Fargate. It supports exploration of operation state, errors, and recovery.
The presentation uses ephemeral PostgreSQL as part of the lab. It is a teaching resource for exploring design decisions.
Resources
Explore the talk at your own pace.
The slides, the lab, and the event publications, together in one place.
A look inside the slides
Five moments from the presentation. Open each image for a closer look.
From the community
A place to share and learn.
AWS Community Day Perú brings the community together to exchange experiences and knowledge about AWS.
