Exact nominal delay
Useful for deterministic inspection, but synchronized clients can collide.
Retry Planning
Generate bounded exponential-backoff schedules with Retry-After handling, jitter ranges, cumulative wait time, and code templates.
RETRY PLAN
Use the defaults or enter the retry policy you want to inspect.
| Attempt | Nominal delay | Jittered delay range | Cumulative nominal wait |
|---|---|---|---|
| Generate a schedule to see retry delays. | |||
IMPLEMENTATION TEMPLATE
The template returns one delay. Your retry loop must still decide whether the failure is transient and whether replaying the operation is safe.
Generate a schedule to create an implementation template.
HOW IT WORKS
Each nominal delay multiplies the previous backoff and stops growing at the selected maximum. Jitter turns that nominal value into a safe random range so many clients are less likely to retry at the same instant.
Nominal delay is initial delay multiplied by the multiplier raised to the retry index.
The configured maximum bounds the calculated backoff before jitter is applied.
A supplied provider delay becomes the minimum for the first retry, even when it exceeds the local maximum.
Choose no jitter, full jitter, equal jitter, or a percentage range based on the failure mode.
Useful for deterministic inspection, but synchronized clients can collide.
Spreads retries across the widest range and can reduce coordinated retry spikes.
Preserves a minimum pause while still distributing retry timing.
Varies around the nominal delay. The configured maximum caps the upper bound.
Implementation template: Retry only failures that are safe and likely to be transient. Honor provider instructions, protect non-idempotent operations, cap total retry time, and add observability. A generated schedule is not a production guarantee. Learn how backoff and jitter policies work. See four retry schedules compared in a reproducible simulation.