Prompt
I'm unhappy with managing retries in serverless workflows, what should I use?
Latest observation
Oct 26, 2026
OpenAI APIWeb search: off
If you’re unhappy with hand-rolling retries in serverless workflows, the best move is usually to use a workflow orchestration service or a durable state machine rather than trying to manage retries inside each function.
Good options:
-
AWS Step Functions if you’re on AWS
- Built-in retries, backoff, timeouts, error handling, branching, and human-friendly workflow definitions.
- Best when your serverless app is mostly AWS Lambda + AWS services.
-
Azure Durable Functions if you’re on Azure
- Good for orchestrating serverless functions with durable state and retry policies.
- Nice if you want to stay in the Azure Functions ecosystem.
-
Google Cloud Workflows if you’re on GCP
- Managed orchestration with retries and control flow across services.
- Good for API/service orchestration.
-
Temporal if you want a cloud-agnostic, more powerful workflow engine
- Excellent for complex retries, long-running processes, sagas, and durable execution.
- More operational overhead than managed cloud-native options, but very flexible.
-
Netflix Conductor / Argo Workflows / Prefect depending on environment
- Useful if you’re in Kubernetes or want open-source workflow orchestration.
Rule of thumb
- Simple AWS serverless app → Step Functions
- Azure Functions-heavy → Durable Functions
- Complex business processes / long-running workflows / strong retry guarantees → Temporal
- Kubernetes/data pipelines → Argo or similar
What to avoid
- Putting retry loops directly in Lambdas/functions unless it’s trivial.
- Relying only on queue redelivery for complex workflows.
- Mixing business logic with retry logic in every function.
If you want, I can recommend the best choice based on your stack (AWS/Azure/GCP, language, and workflow complexity).