Explain SRE architecture and troubleshooting scenarios

Quick Overview

This question evaluates a candidate's competency in SRE architecture and reliability engineering, covering Kubernetes fundamentals, pod failure troubleshooting, SLO/SLA/SLI reasoning, multi-layer latency diagnosis, and SQL versus NoSQL trade-offs within the System Design domain.

Explain SRE architecture and troubleshooting scenarios

Company: TikTok

Role: Software Engineer

Category: System Design

Difficulty: hard

Interview Round: Technical Screen

Explain the difference between a Pod and a Deployment in Kubernetes and describe the role of a Service. How would you debug a Pod that fails to start—list concrete steps and typical kubectl commands you would use. Define and contrast SLO, SLA, and SLI; if the actual error rate exceeds the SLO, what actions would you take and how would you prioritize them. Given a distributed system with elevated request latency, outline a structured troubleshooting approach across the network, load balancer, database, cache, and the service itself, including the specific metrics to check and experiments to run. Compare SQL and NoSQL data stores and describe scenarios where NoSQL is the better fit.

Quick Answer: This question evaluates a candidate's competency in SRE architecture and reliability engineering, covering Kubernetes fundamentals, pod failure troubleshooting, SLO/SLA/SLI reasoning, multi-layer latency diagnosis, and SQL versus NoSQL trade-offs within the System Design domain.

|Home/System Design/TikTok
TikTok logo
TikTok
Sep 6, 2025, 12:00 AM
hardSoftware EngineerTechnical ScreenSystem Design
8
0

Kubernetes, Reliability, and Data Store Concepts (Technical Screen)

Context: Assume you operate a high-scale, latency-sensitive microservices platform running on Kubernetes. Answer the following practical questions.

1) Kubernetes Building Blocks

  • Explain the difference between a Pod and a Deployment in Kubernetes.
  • Describe the role of a Service.

2) Debugging a Pod That Fails to Start

List concrete steps and typical kubectl commands you would use to troubleshoot a Pod that is Pending, ImagePullBackOff, CrashLoopBackOff, OOMKilled, or failing health probes.

3) SLO, SLA, and SLI

  • Define and contrast SLO, SLA, and SLI.
  • If the actual error rate exceeds the SLO, what actions would you take and how would you prioritize them?

4) Elevated Request Latency in a Distributed System

Outline a structured troubleshooting approach across the network, load balancer, database, cache, and the service itself. Include:

  • Specific metrics to check at each layer.
  • Concrete experiments to run to isolate the bottleneck.

5) SQL vs NoSQL

Compare SQL and NoSQL data stores and describe scenarios where NoSQL is the better fit.

Submit Your Answer to Earn 20XP

Sign in to leave a comment

Loading comments...