Resilience & operational
Keeping flows healthy under failure and load.
These patterns protect flows from transient faults, failing endpoints, and traffic spikes, and give bad messages a safe place to go instead of blocking the queue behind them.
Retry. Recover from transient failures by retrying on a send port until success or a limit.
Circuit breaker. Stop hammering an endpoint that is clearly down, then test once after a cooldown.
Throttling / rate limiting. Cap the delivery rate so a fragile or rate-limited downstream is not overwhelmed.
Store-and-forward. Hold messages durably while a destination is unavailable and forward on its return.
Poison-message handling. Remove a message that fails every time so it stops blocking the queue behind it.
Flexible validation. Validate in a pipeline component so a harmless deviation from the agreed shape is accepted and reported instead of rejected, while an unsafe one still fails.