Pira lan cara circuit breaker lan retry kanthi backoff dalam sistem terdistribusi?

Question

Accepted Answer

Ing sistem terdistribusi, panggilan gagal. **Retry kanthi backoff** mulih saka kesalahan *sementara*; **circuit breaker** ngelingake sampeyan ora manyun-manyun dependensi sing *mudhun*. Padha komplementer: coba maneh nangis, mateni ing outage.

## Retry kanthi exponential backoff + jitter

Kanggoan sementara (timeout, keparekan jaringan singkat, 503), coba maneh — nanging **backoff eksponensial lan tambahin jitter** supaya retry nyebar tinimbang nyinkronisasi menyang bala tentara gundhul.

```text
attempt 1 → wait ~1s   (+ random jitter)
attempt 2 → wait ~2s   (+ random jitter)
attempt 3 → wait ~4s   (+ random jitter)
→ cap at maxRetries (e.g. 3) and a max delay → don't retry forever
```

Peringatan utama:

- **Cap retry** — retry sing ora ana wates nggawe **retry storm** sing bisa mbebani layanan sing lagi mulih.
- **Idempotency** — mung coba operasi sing aman diulang (pembayaran sing dicoba maneh kudu ora ngisi loro).
- **Jitter** — tanpa, kabeh klien coba maneh ing wektu sing padha lan munculake beban maneh.

## Circuit breaker

Circuit breaker nglacak kegagalan kanggo dependensi lan, sawisé threshold, **mateni** — gagal cepet (utawa mulih fallback) tinimbang nelpon layanan mati. Sawisé cooldown iku **setengah-open** kanggo raket pemulihan.

```text
CLOSED   → calls pass through; count failures
  too many failures → trip → OPEN
OPEN     → fail fast immediately (no call); start cooldown timer
  cooldown elapsed → HALF-OPEN
HALF-OPEN → allow a few probe calls
  success → CLOSED (recovered) ; failure → back to OPEN
```

## Nalika saben ditrapake

```text
Retry           → transient, likely-to-succeed-soon errors (1 slow call)
Circuit breaker → repeated/sustained failures (the dependency is down)
→ use together: breaker prevents retries from piling onto a dead service
```

## Ambeg penting

Tanpa iku, dependensi sing gagal setya nyuwun pangilane mudhun: panjaluk numpuk ing timeout, retry nambah beban, lan kegagalan nyebar ing layanan. Exponential backoff kanthi jitter mulih saka keparekan tanpa retry storm; capping lan idempotency njaga retry aman; lan circuit breaker mateni boros sumber daya ing dependensi mati lan menehi papan kanggo mulih. Bebarengan, maka kegagalan dependensi dadi acara sing dikungkung lan self-healing.