Technical leadership

1.You lead SRE for a checkout service. It has spent its whole monthly error budget in ten days. The signed error budget policy says feature launches pause. The product lead says a holiday promotion must ship Friday. What do you do first?

2.A payments team keeps blowing its error budget. Each time, its director grants an exception to keep shipping. This has happened four months in a row. As the staff SRE, what is the most useful move?

3.You are drafting the first error budget policy for a ride-booking platform. Which feature matters most for it to hold up later?

4.A team's error budget is exhausted. Most of the burn came from one bad dependency upgrade that has since been rolled back. The policy says to freeze features. The team asks for a waiver. What is a sound response?

5.Your error budget dashboard shows a search service burned 70 percent of its budget in a week. The team has not noticed. Nobody has paged. What is the right first step?

6.A product manager says error budgets are just SRE's way of saying no. How do you reframe it so they see it as a tool for them?

7.A team is freezing features after spending its error budget. Two weeks in, engineers are idle because nobody planned reliability work. What should you have pushed for when the policy was written?

8.A leader asks you to raise a team's availability objective from 99.9 to 99.99 percent, with no change in staffing or architecture. What is the most useful reply?

9.Product wants a 100 percent availability objective for a new booking service. What is the strongest argument to change their mind?

10.You are negotiating a service level objective (SLO) with a product lead for a video upload feature. Which input should carry the most weight?

11.A product lead and a backend lead disagree on a latency objective. Product wants 200 ms at the 99th percentile. Backend says 800 ms is all they can do. You are asked to settle it. What do you do?

12.A service has had a 99.95 percent objective for a year. It has never come close to missing it, and users have never complained. What does that suggest?

13.Sales has promised a customer a 99.99 percent contractual uptime that engineering never agreed to. The internal objective is 99.9 percent. What do you do?

14.You want to introduce service level objectives across forty services. Teams are skeptical. What is a good way to start?

15.A product lead asks why their dashboard shows 99.9 percent availability while users report outages. You find the indicator measures only server responses, not client failures. What do you do?

16.Product asks for an objective on a feature that is still an experiment with a few hundred users. What is a reasonable stance?

17.Your company writes good postmortems, but an audit finds 60 percent of action items are still open after six months. As the reliability lead, what change helps most?

18.A postmortem lists twelve action items, ranging from a one-line config fix to rewriting the scheduler. What should you push for?

19.In a postmortem review, a director asks, 'Who pushed the bad config?' The room goes quiet. You are facilitating. What do you say?

20.A team keeps closing action items by filing new tickets, such as 'investigate retry logic'. The tickets then sit for months. What standard should you set?

21.A key action item belongs to a team that says it has no capacity this quarter. The same failure could cause a large outage. What is the right step?

22.You run the weekly incident review. Attendance has dropped and the reviews feel like a formality. What change is most likely to make them useful again?

23.After a large outage, leadership wants the postmortem in 24 hours. The investigation is not finished. What do you do?

24.Your rotation of six engineers averages 40 pages a week. Two people have asked to leave on-call. What should you do first?

25.You want to set a toil cap for your SRE team: no more than half of each engineer's time on manual, repetitive work. How do you make it real?

26.An alert fires nightly for disk usage on a log server. The on-call engineer always runs the same cleanup command. It has gone on for a year. What is the right fix?

27.A senior engineer on your rotation takes every hard page, even when off shift. Others have stopped learning the systems. What should you do?

28.Leadership asks why toil reduction should get headcount over new features. Which framing is most persuasive?

29.A team wants to automate a weekly manual failover test that takes 20 minutes. Building the automation would take three weeks. How do you decide?

30.After a hard on-call week with five overnight incidents, the primary engineer is due in a planning meeting at 9 a.m. What do you do as the lead?

31.Your rotation spans two time zones, but nearly all pages land overnight for the European half. What is a good structural fix?

32.Half the pages in your rotation are later marked 'no action needed'. What should you do about these alerts?

33.A product team wants SRE to take over on-call for their new service, saying 'reliability is your job'. What is the right response?

34.You want service teams to own their reliability. Which change makes that ownership most real?

35.SRE currently runs on-call for twenty services. You plan to hand them back to the service teams. How should you sequence it?

36.A service team has a high page rate and asks SRE to fix it. You see the cause is retry logic in their own code. What do you do?

37.A launch review checklist, run by SRE, has become a gate teams resent. Launches wait weeks for SRE sign-off. What is a better model?

38.A VP says, 'We have an SRE team, so why do product engineers need to care about reliability?' How do you answer?

39.You are creating a shared on-call standard for all service teams. Which part is most important to get right?

40.For the fifth time this year, a certificate expired and caused an outage. Each postmortem blamed a different team's missed renewal. What do you do?

41.You notice that one in four incidents last quarter came from config changes that skipped staged rollout. No single postmortem flagged it. What is the best next step?

42.A recurring database failover issue has caused three outages. Everyone agrees it should be fixed, but no team owns the shared database layer. What do you do?

43.You want to make recurring incidents visible to leadership. Which approach works best?

44.A noisy neighbor problem on shared hosts causes small outages every few weeks. The fix is a large platform change. Leaders keep deferring it. What strengthens your case?

45.A recurring incident has a known, cheap mitigation, but the root fix will take a quarter. Product wants only the mitigation. What do you recommend?

46.You are starting a disaster recovery program. Teams claim their services can fail over, but nobody has tried. What is the first meaningful step?

47.A failover test of the billing system reveals the recovery time would be eight hours. The stated recovery time objective is one hour. What do you do?

48.Teams want disaster recovery tests only on weekends at 3 a.m., with notice weeks ahead. What is your view as the program lead?

49.Your company's disaster recovery plan lists one engineer who knows how to restore the main database from backup. What is the most urgent fix?

50.A director asks for a single report showing disaster recovery readiness for 30 services. What should the report show?

51.A failover game day causes a 10-minute customer-facing outage. An executive wants to cancel future game days. How do you respond?

52.Your disaster recovery program has strong momentum, but you are moving to another team in three months. What do you focus on?

53.You want to move all services from host-based alerts to alerts based on service level objectives. Ten teams will be affected. What do you write first?

54.Your design doc for a new deploy pipeline has been open for comment for three weeks. It has 80 comments and no decision. What do you do?

55.A design doc you wrote proposes replacing the company's alerting tool. Which section most helps reviewers judge it fairly?

56.You are setting a two-year reliability direction for the platform group. What makes it most likely to guide real decisions?

57.A peer asks you to review their design doc for a new caching layer. You think the whole approach is wrong. How do you give that feedback?

58.Your team has many tribal rules about how to run releases safely, but none are written. New hires keep breaking them. What is a good first step?

59.In a design review for a multi-region rollout, a senior engineer objects strongly, but the rest of the group agrees. What do you do?

60.You run the review board for changes to shared infrastructure. Reviews take a month and teams work around them. What should you change?

61.After a review, you decided to use a single shared queue for alerts. A team that disagreed keeps reopening the debate in every meeting. What do you do?

62.In a design review, the most junior engineer has not spoken, but you know they have worked on the system under review. What do you do?

63.You must choose between two load balancers for a new edge tier. Benchmarks are close, and more testing would take a month. The launch is in six weeks. What do you do?

64.A year ago you chose a metrics database. It now struggles with scale. A teammate asks why you picked it. What helps most at this point?

65.You decided to roll out a new deployment system gradually. Three weeks in, rollback rates are higher than with the old system. What do you do?

66.Leadership asks for a firm date to finish moving all services to a new region. Half the services have never been tested there. What do you give them?

67.You must decide whether to delay a database upgrade that has a known bug affecting 1 percent of queries. The vendor fix is weeks away. What helps you decide?

68.You lead a six-month project to add automated canary analysis for all deploys. At month two, a reorg cuts your team in half. What do you do?

69.Three teams ask your SRE group to build them custom dashboards this quarter. You already have a full roadmap. What do you do?

70.You are planning a multi-quarter observability overhaul. How should you sequence the work?

71.A director asks your team to also take on cost optimization for cloud spend, on top of reliability work. You think it would stretch the team too thin. What do you do?

72.Halfway through a project to automate database failover, you find a simpler approach that covers 80 percent of cases in a third of the time. What do you do?

73.Your monitoring stack has years of debt: duplicate agents, dead dashboards, and alerts nobody owns. How do you get time to fix it?

74.A team wants to rewrite its deploy tooling from scratch. The current tooling is ugly but reliable. What question should you ask first?

75.A finance leader asks why your team spends 20 percent of its time on 'maintenance'. What answer works best?

76.You are leading a migration of 60 services from an old container platform to a new one. Ten teams have not started, and the old platform loses support in six months. What do you do first?

77.In a platform migration, a team says its service is too special to move. Their reasons are a custom kernel setting and a fixed IP address. What do you do?

78.You are running a migration across 20 teams. What is the best way to keep everyone informed of progress?

79.A migration to a new logging pipeline is 90 percent done. The last 10 percent are hard edge cases. Leadership wants to call it done. What do you push for?

80.Several outages came from deploys with no rollback plan. How do you raise the bar across teams?

81.Code reviews for alerting rules are often rubber-stamped. Several bad alerts got through. What is the best fix?

82.Your team has a hero culture. Late-night saves get praised in all-hands, but steady preventive work goes unnoticed. What do you change?

83.You notice teams copy-paste Terraform modules for databases, each slightly different, leading to inconsistent backups. What do you do?

84.A mid-level SRE on your team wants to grow toward senior. What is the most useful opportunity you can give them?

85.You delegated a capacity planning project to a senior engineer. You disagree with one of their choices, but it is not dangerous. What do you do?

86.A new SRE is nervous about leading their first incident. What is a good way to prepare them?

87.You are the only person who approves changes to the production network config. Changes queue for days when you are busy. What do you do?

88.A team member delivered a postmortem that was thin and missed the main contributing cause. How do you give feedback?

89.You notice you are pulled into every incident across five teams, even minor ones. What is the best long-term fix?

90.You lead a project to cut deploy time from 40 to 10 minutes. It is due in four weeks, and you now think it will take ten. When do you tell your stakeholders?

91.A dependent team promised an API you need for a reliability project. They are now three weeks late, and your launch date depends on it. What do you do first?

92.Your project is slipping because two engineers were pulled into incident work for a month. Your manager asks how it is going. What do you say?

93.You learn that a project you inherited has been reported as on track for months, but it is far behind. What do you do?

94.During a major outage, you are the senior SRE on the call. The incident commander is a junior engineer who is doing a decent job. What is your role?

95.After a painful outage, an engineer who caused the trigger is visibly upset. A manager suggests a formal warning. What do you do?

96.You are leading an incident that has run for five hours. Engineers are tired and making small errors. What do you do?

97.After an incident, a customer-facing executive asks for 'the name of whoever messed up'. How do you respond?

98.A major incident is over, and customers are back. The incident channel is quiet. What is the most important thing to do in the next day?

99.A security team wants to roll out a new agent to every host this month. Your data shows it uses enough CPU to risk latency on the busiest hosts. What do you do?

100.A finance team wants to cut infrastructure spend by 25 percent. Their plan removes the spare capacity that keeps a service up during traffic spikes. What do you do?

101.Your manager wants to cut the on-call rotation from eight people to four to free up project time. You think it will burn people out. What do you do?

102.A principal engineer proposes a custom service mesh. You think an existing open-source one would cost far less to run. They are more senior. What do you do?

103.Your director makes a decision you argued against: adopting a new incident tool mid-year. It is now final. What do you do?

104.Your manager asks you to sign off on a launch readiness review for a service you believe is not ready. What do you do?

105.You want all teams to adopt structured logging, but you have no authority over them. What is the most effective approach?

106.A team lead keeps rejecting your suggestions to add health checks. You suspect they see SRE as outsiders. What do you try?

107.Every major architecture decision in your org waits for your review. Teams say they are blocked. What do you change?

108.You are the only one who understands the custom autoscaler. You want to move to a new project. What do you do?