Piye cara milih ambang weker supaya bisa ngindhari kapetengan weker lan positif palsu?

Question

Accepted Answer

Aturan inti: **weker ing simtom, dudu saben**, lan **nang-kene mung yaiku sing bisa ditindakake lan darurat**. Weker sing ribut sing murub saben bengi bakal ditilem utawa dikunci — dadi risiko sejati dudu weker sing ilang, nanging on-call sing desensitized sing turu ing sawah.

## Napa penting

Weker ing **SLO sing ngadhepi pengguna** (error rate, latency, availability), ora ing sebab internal kaya "CPU > 80%". CPU dhuwur bisa uga ora apa-apa; sing penting yaiku apa pangguna kena tepak.

```text
BAD  (cause)     CPU > 80% for 5m            → fires constantly, often no impact
GOOD (symptom)   error-rate SLO burn is fast → fires only when users hurt
```

## Weker rate-burn multi-window

Satunggal ambang statik yaiku kakehan kuwat utawa alon. Ananging, weker on bab cepet kowe murub **budget salah**, nggunakake rong windows dadi combustion cepet nang-kene langsung lan slow burn nang-kene mung yen berkelanjutan.

```yaml
# Page: 2% of monthly budget burned in 1h AND still burning over 5m
- alert: HighErrorBudgetBurn
  expr: |
    (slo:error_ratio_1h > 14.4 * 0.001)   # fast window: catch big outages
    and
    (slo:error_ratio_5m > 14.4 * 0.001)   # short window: confirm it's still happening
  for: 2m
  labels: { severity: page }
```

Windows cendhak nyegah nang-kene ing masalah sing wis ngrampungake dhewe; windows dawa nyegah nang-kene ing campuran singkat.

## Nang-kene vs tiket vs dashboard

```text
PAGE       actionable + urgent  → wake a human now (SLO at risk, checkout down)
TICKET     actionable, not urgent → fix in business hours (disk 70%, cert in 20 days)
DASHBOARD  informational         → no alert, just visible (per-endpoint traffic)
```

Yen weker ora bisa ditindakake, ora kudu nang-kene — ganti dadi tiket utawa panel dashboard. Banjur **tune over time**: deleng saben nang-kene, busak sing ora ana sing tumindak.

Kapetengan weker yaiku risiko keandalan, ora mung rasa: jalma kanthi cermat saring saluran ribut, dadi luwih akeh positif palsu sing kowe kirim, luwih bisa disimpulake pabrik sejati kepuntren. Weker ing simtom karo windows burn-rate, lan nganggep nang-kene kanggo urgensi sing bisa ditindakake, njaga saben nang-kene bermakna — apa sing njaga on-call responsif.