1 min lesson
Limit retries across all traffic
Set retry eligibility, backoff and a traffic-wide cap that prevents retry storms.
Step 1 of 2
Limit retries across all traffic
A fixed retry count applies to every failed request. During a broad outage, that multiplies traffic when the provider has the least capacity. A global retry budget limits retries to a chosen share of all requests.
- 1Retry only failures the shared error policy marks as temporary, such as selected 503 responses or timeouts.
- 2Use exponential backoff with jitter so retries do not start at the same time.
- 3Choose a global retry cap from load tests and capacity data. When the cap is spent, shed load or use a healthy fallback.
- 4Send a retry to a healthy provider when the original provider is unavailable. Repeating the same failed route wastes the budget.