SeniorObservability

Define an SLO for a checkout service. What would you pick and why?

What they are really testing: Whether you can pick a measurable indicator tied to user experience and set a target from business consequence rather than from a habit of writing nines.

A real interview question

Define an SLO for a checkout service. What would you pick and why?

What most people say

drag me

Five nines availability, since checkout is the most critical service.

Five nines is about 26 seconds of downtime a month, which is extraordinarily expensive and almost certainly beyond what the business needs or the architecture supports. Picking a number by importance rather than by cost and consequence is the tell.

The follow-ups they ask next

  • Why 28 days rather than a calendar month?

    It contains exactly 4 of each weekday, so weekly traffic patterns are weighted evenly, and it avoids the budget resetting on the 1st, which otherwise encourages risky behaviour at month end.

  • The service is at 99.99% and the target is 99.9%. Is that good?

    Not necessarily. Consistently overshooting suggests you are spending too much on reliability and could take more risk, ship faster, or reduce cost. An unused error budget is unrealised velocity.

What the interviewer is listening for

  • Includes latency in the success definition
  • Derives the target from business consequence
  • Measures at the client or edge, not only server side

What sinks the answer

  • Picks nines by perceived importance
  • Availability only, ignoring latency
  • No consequence attached to breach

If you genuinely do not know

Say this instead of freezing. Reasoning out loud from what you do know beats silence every single time, and a good interviewer is listening for exactly that.

The indicator is [proportion of checkout requests succeeding within 2 seconds over 28 days], since [a slow success is still a failure]. The target comes from [what failure costs versus what each nine costs], usually landing near [99.9%, about 43 minutes a month]. Frame it as [an error budget with an agreed consequence], and measure [at the edge, since server metrics miss requests that never arrived].

Keep going with observability

All 87 devops engineer questions

Knowing the answer is not the same as recalling it under pressure

Sign in to send the questions you fumble to spaced recall, so they come back right before you would forget them, and learn the concepts behind them with hands-on labs.

Start free