Google DevOps / SRE Interview Questions

1.Tell me about a time when you had to implement a monitoring solution without clear requirements or success metrics. How did you figure out what to build?

2.Tell me about a time you convinced your team to adopt a reliability practice they were initially resistant to. How did you get them on board?

3.Walk me through how you would troubleshoot a service that's experiencing 50% error rates across multiple Google Cloud regions. You have 10 minutes to present your approach.

4.Describe the deepest technical challenge you've solved in the reliability space. What made it particularly complex and how did your expertise make the difference?

5.Describe a situation where you had to make a critical reliability decision with incomplete information and tight time constraints. What was at stake and how did you decide?

6.You're tasked with improving the reliability of a system you've never worked on before and the original team has left the company. Walk me through your approach for the first 30 days.

7.You notice that writes to Spanner are taking 300ms when they used to take 50ms, but CPU and memory look normal. How would you debug this?

8.You're running a canary deployment through our traffic splitting system, and you notice the canary has 2x higher latency but the same error rate. The code changes only added logging. How do you investigate?

9.A service in our Borg cluster is getting OOMKilled every few hours, but heap dumps show memory usage is well below the limit. What could be happening?

10.Write a function that takes a stream of search query logs and returns the top 10 queries by volume in the last sliding window. The stream processes billions of queries per day.

11.Design a monitoring system that can detect when YouTube's video serving pipeline is degrading before users notice. What signals would you track and how would you aggregate them?

12.You need to migrate 50TB of data from our legacy Bigtable cluster to a new one with zero downtime. The data has a heavy read pattern during peak hours. Walk me through your approach.

13.Design a rollback system for Chrome that can safely revert a bad update pushed to 3 billion devices within 15 minutes. How do you prevent making the problem worse?

14.YouTube's recommendation engine needs to serve 2 billion users with personalized video suggestions. Design a system that can handle traffic spikes during major events while ensuring recommendations stay relevant even when the ML pipeline fails.

15.Gmail's spam detection service needs to process 300 billion emails per day with sub-100ms latency. Design the infrastructure to handle this load while ensuring the ML models can be updated hourly.

16.You need to build a configuration management system that every microservice at Google can use to fetch feature flags and settings. It must work even when its own backend is completely down.

17.Design the infrastructure for Google Maps to serve routing requests when an entire region goes offline. You have 30 seconds to reroute traffic and maintain sub-second response times globally.

18.Build a deployment system for Android system updates that can safely roll out to 3 billion devices while detecting issues like boot loops or battery drain before they affect millions of users.

19.Tell me about a time you had to lead a technical change across multiple engineering teams who each thought their current approach was best. How did you get alignment?

20.You discover that three different teams are each building their own monitoring solutions for similar problems, creating fragmentation and maintenance burden. How do you drive consolidation without stepping on technical egos?

21.Describe a situation where you discovered your team was spending 60% of their time on operational toil instead of reliability engineering work. How did you turn this around?

22.You're leading an incident response where a critical Google service is down, affecting millions of users. The incident has been going for 2 hours and your team is getting frustrated with each other. How do you handle this?

23.Tell me about a time you had to influence a senior engineering manager to change their team's practices because they were impacting reliability across the broader system. What approach did you take?

24.You're asked to lead a cross-functional effort to improve Google's infrastructure reliability, but you have no formal authority over the teams involved and they all have competing priorities. How do you make this successful?

25.Gmail serves 1.8 billion users and we're considering adding a new feature that requires an additional 10ms to every email send. How would you estimate the total infrastructure cost impact per year?

26.Search latency has increased by 15ms globally over the past week, but our monitoring shows all individual services are healthy. Walk me through how you'd investigate this.

27.YouTube's CDN cache hit rate dropped from 85% to 75% overnight with no code changes. Estimate the additional bandwidth cost and explain how you'd prioritize investigating this.

28.Google Cloud's control plane API has a p99 latency SLO of 500ms, but you're seeing it hit 800ms during peak hours. The product team wants to know if we should add more capacity or optimize the code. How do you decide?

29.Android's OTA update system needs to distribute a critical security patch to 3 billion devices within 48 hours, but our current rollout strategy takes 2 weeks. How would you redesign the deployment approach?

30.We're planning to migrate Spanner's storage layer to a new compression algorithm that could save 30% disk space but adds 5ms to read latency. How do you evaluate whether this trade-off makes sense for Google's business?