MidBehavioural

Tell me about your worst on-call week. What actually happened, and what did you change afterwards?

What they are really testing: They want to know whether a bad week produced a fix or just a story. The subtext is, do you treat operational pain as data or as fate.

A real interview question

Tell me about your worst on-call week. What actually happened, and what did you change afterwards?

What most people say

drag me

It was brutal, I barely slept, but I got through it. That is just part of being on call, you have to be able to take it.

It frames suffering as a badge instead of a signal, and it gives the interviewer no evidence that anything improved. A team that hires this mindset keeps the same broken rotation forever, and pages stay noisy.

The follow-ups they ask next

  • How did you convince the team to actually delete alerts, given the fear of missing something?

    Use the data, an alert nobody acts on is already being missed. Offer a safety net, downgrade to a digest for a month before deleting.

  • What do you do when you are too tired to be safe mid incident?

    Say it out loud and hand over. Name the concrete protocol, call the secondary, write your current state in the incident channel before you drop.

  • Whose job is it to fix the rotation, yours or your manager's?

    Show shared ownership. You bring the evidence and the proposed cuts, the manager funds the time. Neither works alone.

What the interviewer is listening for

  • Counted the pages and classified them instead of describing feelings
  • Recognised fatigue itself as an operational risk and escalated for help
  • Turned the bad week into a durable alerting rule the team kept

What sinks the answer

  • Wears sleeplessness as a badge of honour with no follow up work
  • Blames a specific colleague for the noisy rotation
  • Cannot name a single alert that was changed or removed afterwards

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.

I have not carried a pager yet, but here is how I would reason about it. I would log every page with a note on whether a human action was actually needed, and bring that count to the retro, because an alert nobody acts on should either be fixed or stop paging.

Keep going with behavioural

All 336 cloud 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