I85PREP
⌘ K
Module 15 study materials
Module 15 study resources

Module 15: Operations, Monitoring, and Maintenance

Module 15 · Work through your exam, one topic at a time.

All guide sections

Operations, Monitoring, and Maintenance

This domain tests whether you can keep a live service healthy and prove it. Expect SLIs and SLOs, alert design, on-call practice, patching, backup restore, and maintenance planning.

Core rules

  • An SLI is a measured indicator such as latency, error rate, or saturation. An SLO is a target over a window. The error budget is 1 minus the SLO.
  • Monitor symptoms users feel, not only causes. Golden signals are latency, traffic, errors, and saturation.
  • Alerts must be actionable, owned, and linked to a runbook. Page only for urgent user impact; create tickets for trends.
  • Observability needs logs, metrics, and traces with correlation IDs. Retention must match investigation and compliance needs.
  • On-call practice includes an escalation path, handover notes, and blameless review. Track toil and reduce it deliberately.
  • Maintenance includes severity-based patch cadence, maintenance windows, restore tests, not just backup jobs, and capacity trend review.
  • Asset inventory and the configuration management database must match reality. Stale inventory breaks impact analysis.
  • A change calendar and freeze periods reduce collision risk during peak events.

Worked example

Availability SLI equals successful requests divided by total requests over 30 days. SLO 99.9% gives an error budget of 0.1%, about 43 minutes per month. Alert policy: page when the burn rate is 14.4x over one hour, consuming 2% of budget; ticket at 6x over six hours. Reasoning: burn-rate alerts catch fast and slow burn, and thresholds tie to the budget rather than an arbitrary CPU threshold. CPU at 80% may be normal; user-facing errors are not.

Traps

  • Paging on resource metrics such as CPU alone, producing alert fatigue.
  • Setting a 100% SLO, leaving no budget for legitimate change.
  • Running backups without ever testing a restore.
  • Watching averages and missing p95 or p99 latency that users feel.
  • Putting an on-call engineer on rotation without runbooks or escalation.

Key points

  • SLIs measure behavior; SLOs set targets; error budgets quantify allowable risk.
  • Alert on user-facing symptoms and tie pages to runbooks.
  • Correlate logs, metrics, and traces during investigation.
  • Test restores, not just backups, and review capacity trends.
  • Keep inventory accurate and respect change freezes.