The Workflow That Retried a Dead Provider 200 Times: Backoff, Jitter, and Circuit Breakers in n8n

Description and use

In n8n, a provider outage caused 200 retries, wasting quota and delaying jobs. The fix involved exponential backoff capped at five minutes, honoring Retry-After, adding jitter, and a circuit breaker that opens after five consecutive failures, deferring jobs until the breaker closes.

Problem

A provider outage caused a workflow to retry the same failed call 200 times, burning quota, filling logs, and delaying queued jobs.

Outcome

The workflow now limits retries to five attempts, spreads retries with jitter, respects provider hints, and stops retrying after a circuit breaker opens, preventing 200 redundant calls and preserving resources.

Workflow

• Enable retry-on-failure in n8n
• Configure exponential backoff: 1s, 2s, 4s, 8s, up to 5min cap
• Add logic to honor Retry-After header if present
• Implement full jitter: random wait between 0 and backoff cap for each attempt
• Persist backoff state per job so it does not reset on new execution
• Add circuit breaker: after 5 consecutive retryable failures, open breaker for a fixed window
• When breaker is open, route jobs to deferred queue with reason "provider circuit open"
• After window, allow one probe call; if success, close breaker, else reopen
• Ensure breaker state lives in Postgres table shared across executions
• Add alerts for breaker opening
• Classify failures: retryable (429, 5xx, timeouts) vs non-retryable (4xx, 401/403)
• Test with mock provider that fails for 10 minutes then recovers, asserting limits and behavior

Evidence

By the time the provider came back, the workflow had retried the same failed call two hundred times.

Source

The first fix is backoff: each attempt waits longer than the last. Mine doubles the wait — 1 second, 2, 4, 8 — up to a cap of five minutes.

Source

The fix is jitter: add a random spread to each wait, so retries land across a window instead of on a tick.

Source

After a set number of consecutive failures — mine is five for provider calls — the breaker opens, and the workflow stops calling that provider entirely for a fixed window.

Source

None of this applies to every error. A retry policy that treats all failures the same will retry the ones that can never succeed.

Source