All topicsTopic 18 / 24
Intermediate 1 minute
Fault Tolerance
Fault tolerance means a system continues providing an acceptable service when part of it breaks. Designs use redundancy, health checks, timeouts, retries with limits, graceful degradation, and recovery procedures. It does not mean failures disappear. It means failure is expected, contained, visible, and recoverable without losing more data or overwhelming healthy components.
Key idea
Assume components fail; isolate impact and preserve the most important path.
See it in one picture
Follow the arrowsRequest
Healthy path
Failed node
Fallback
Keep the important path working
Real-world example
If a recommendation service times out, a shopping site can still show products and checkout with a default recommendation list. The degraded response is better than making every page unavailable.
Quick check