Federating reliability ownership to product teams
WHAT IT TESTS: Whether you can scale reliability by enabling teams, not gatekeeping. OUTLINE: Build a self-service reliability platform (golden paths, paved roads), train teams and embed SLO/on-call practices, and govern with standards plus error budget…
Capacity planning for a distributed cache
WHAT IT TESTS: Whether you can pick the right cache metrics and forecast scaling. OUTLINE: Track hit ratio, memory and eviction rate, throughput/latency, and connections; correlate with growth to forecast when to add capacity before the hit ratio or evictions…
Capacity planning for distributed stateful systems
WHAT IT TESTS: Whether you plan capacity beyond raw resources. OUTLINE: Account for replication and cross-region network, IOPS and disk throughput, connection limits, partition/shard balance, and N+1 region failover headroom; validate with load and chaos…
Designing a feature flagging service
WHAT IT TESTS: System design for low-latency config delivery and HA. OUTLINE: Control plane (UI, store, targeting), SDKs that cache flags locally for zero-latency evaluation, streaming/CDN delivery for near-real-time updates, and stale-flag lifecycle tooling.
Managing a risky release with feature flags
WHAT IT TESTS: Whether you can decouple deploy from release and roll out safely. OUTLINE: Deploy code dark behind an off flag, enable for internal then small percentage, monitor metrics, ramp gradually, then remove the flag.
Automating actions on error budget burn
WHAT IT TESTS: Whether you can operationalize error budgets with automation. OUTLINE: Use multi-window burn-rate alerts; on fast burn, auto-trigger actions like halting deploys, rolling back, or scaling, with escalating tiers.
Safeguards for automated remediation runbooks
WHAT IT TESTS: Whether you can build safe auto-remediation. OUTLINE: Add rate limits and circuit breakers on restarts, escalate to humans after N attempts, log all actions, and check for cascading failure before acting.
Breaking tunnel vision during an incident
WHAT IT TESTS: Whether you can counter confirmation bias under pressure. OUTLINE: Call out the assumption, ask for disconfirming evidence, list parallel hypotheses, split responders to investigate them, and anchor on what changed and the data.
Key roles in incident response
WHAT IT TESTS: Whether you understand structured incident command. OUTLINE: Incident Commander coordinates and decides but does not fix; Communications Lead handles stakeholders; Operations/SMEs do hands-on diagnosis and remediation.
OpenTelemetry agent and gateway architecture
WHAT IT TESTS: Whether you can architect a scalable telemetry pipeline. OUTLINE: Agents run per-node for local collection and host enrichment; gateways are central, horizontally scaled pools for batching, tail sampling, and routing.
First-SRE 90-day plan at a startup
WHAT IT TESTS: Whether you can introduce SRE incrementally and show value. OUTLINE: Listen and measure first, pick one high-impact service, define SLIs/SLOs and basic alerting, then show reduced toil or incidents to leadership.
Designing shallow vs deep health checks
WHAT IT TESTS: Whether you understand health-check semantics and failure amplification. OUTLINE: Shallow checks confirm the process is alive; deep checks verify dependencies; use shallow for liveness/load-balancer routing and deep sparingly to avoid…
Rolling update vs blue-green deployment
WHAT IT TESTS: Knowledge of deployment strategies and their costs. OUTLINE: Rolling replaces instances gradually with minimal extra capacity but mixes versions; blue-green runs two full environments for instant switch and rollback at double the cost.
Writing high-quality postmortem action items
WHAT IT TESTS: Whether postmortems lead to real change. OUTLINE: Good action items are specific, assigned to an owner, prioritized, tracked to completion, and ideally prevent recurrence rather than just detect faster.
Calculating downtime for a 99.9% SLO
WHAT IT TESTS: Whether you can compute and apply an error budget. OUTLINE: 0.1% of 30 days is roughly 43 minutes of allowed downtime; healthy budget enables faster shipping while depletion slows or freezes deploys.
Designing an error budget policy
WHAT IT TESTS: Whether you can make SLOs enforceable, not decorative. OUTLINE: Define SLO and budget, tiered consequences as burn worsens, a feature freeze on exhaustion, and concrete earn-back criteria.
Keeping a postmortem blameless after an admission
WHAT IT TESTS: Whether you can protect blameless culture under pressure. OUTLINE: Acknowledge the courage, redirect from who to why the system allowed it, ask what guardrails were missing.
Conducting a Production Readiness Review
WHAT IT TESTS: Whether you can systematically gate a launch on reliability. OUTLINE: Assess monitoring and alerting, capacity and load testing, failure modes and dependencies, on-call and runbooks, and rollback or release safety.
Embedded vs consulting SRE engagement models
WHAT IT TESTS: Understanding SRE team topologies and scaling. OUTLINE: Embedded SREs sit inside one team for deep impact but limited reach; consulting SREs advise many teams broadly but shallowly. RED FLAG: Treating either as universally best.
A team keeps blowing its error budget. First steps?
WHAT IT TESTS: Data-driven, collaborative incident reduction. OUTLINE: Analyze where the budget is burning via SLIs and postmortems, validate the SLO and SLIs are sound, then partner blamelessly on the top fixes.