HA & DR Interview Q&A
Q: RTO vs RPO?
RTO (Recovery Time Objective): how long can we be down? Target for restoring service after failure. RPO (Recovery Point Objective): how much data can we lose? Maximum acceptable data loss measured in time.
Q: Active-Active vs Active-Passive?
Active-Active: both sites serve traffic simultaneously. Zero downtime, complex data sync, expensive. Active-Passive: primary serves all traffic, secondary on standby. Simpler, cheaper, but failover has downtime (minutes typically).
Q: How do you design a multi-region K8s deployment?
Primary cluster in region A, replica in region B. Global load balancer (AWS Route53/Azure Traffic Manager) with health checks. Data layer: geo-replicated database (CockroachDB/Cosmos DB/Aurora Global). Deploy with identical Helm charts via ArgoCD ApplicationSet across both clusters.
Quick Reference — Ha Dr
Key Points for Revision
Related Topics
Explore these connected topics to build complete understanding:
Practice Approach
Further Learning
Connect this topic to the broader devops ecosystem.
Each tool in this academy is designed to work with others —
understanding the integration points is what separates intermediate from senior practitioners.
Q: What is the difference between HA and DR?
HA (High Availability): keeps the system running through component failures — redundancy within a region, no planned downtime. DR (Disaster Recovery): recovers from catastrophic failure (data center loss, ransomware) — restores from backup or failover to another region. HA prevents outages; DR recovers from them.
Q: How do you test DR without causing an actual outage?
Chaos engineering: controlled, planned failure injection (Chaos Monkey, Gremlin, AWS Fault Injection Simulator). Game days: scheduled DR tests where team simulates failover. The key is testing before you need it — discover gaps in the runbook while there is no pressure.
Q: What is an RPO of zero and how do you achieve it?
RPO=0 means zero data loss — any transaction committed before failure must be preserved. Achieved with: synchronous database replication (writes are confirmed only after secondary acknowledges), multi-region active-active writes, distributed consensus databases (CockroachDB, Spanner). Expensive in latency — every write waits for remote acknowledgment.
Q: Explain the DR strategy pyramid from cheapest to most expensive.
This section is part of the SynfraCore learning platform. Use the sidebar to navigate to Overview, Fundamentals, Advanced, and Cheatsheet sections for comprehensive coverage of this topic.