Three habits for building reliable backends

A placeholder post with a little more structure so you can see code blocks and lists. Rewrite it with your own lessons.

1. Make failure boring

Design so that the common failure — a timeout, a dropped connection — is handled the same dull way every time. Retries with backoff, idempotent writes, clear error types.

func withRetry(fn func() error) error {
    for i := 0; i < 3; i++ {
        if err := fn(); err == nil {
            return nil
        }
        time.Sleep(backoff(i))
    }
    return ErrExhausted
}

2. Log the decision, not just the event

"Charged user 42" is an event. "Charged user 42 because subscription was past_due and grace period expired" is a decision. The second one saves you at 3am.

3. Test the unhappy path first

The happy path tends to work by construction. Bugs live where things go wrong — so write those tests first.

← Back to all posts