How do you retry a failing step without overloading the service it depends on?
- Module
- 02/05
- To read
- ~7 min
A payment processor came back after a four-minute outage, and a dozen services each had tens of thousands of retries queued for it. What do you think happened next?
In the module you make your guess first, then read what happened and why.
Exponential backoff formula: min(cap, base × 2^attempt)
Full jitter prevents thundering herd
Retry budgets bound total retry cost
Quarantine separates unrecoverable items
A retry plan for one outside service
A spec for your own app, as a text file you download. The module also gives you the prompt for Claude Code, Codex or Antigravity, a checklist to check what it built, and recommended skills.
- Queues and message brokers
- Rate limiting
- Idempotency