Devoxx Greece 2024 - Kubernetes Resiliency by Chris Ayers

Creating a Baseline for Kubernetes

• Set a baseline for resource requests and limits in Kubernetes applications
• Requests should be based on average usage, not minimum

Resiliency and Availability

• Availability zones are critical for resiliency in Kubernetes
• Use availability zones to spread workloads across multiple regions
• Haikus can be used to monitor availability

Monitoring and Observability

• Use metrics like CPU usage, memory usage, and queue lengths to monitor applications
• Leverage tools like Open Telemetry and distributed tracing for observability

Node and Resource Management

• Use node pools and resource requests to manage compute resources
• Limitations are crucial for resource management
• Use feature flags to manage rollout of new features and versions

Scaling and Autoscaling

• Use horizontal pod autoscalers to scale applications based on demand
• Leverage pod disruption budgets to handle scaling and autoscaling

Failure Domains and Rollback

• Identify failure domains in Kubernetes applications
• Use probes like liveness, readiness, and startup probes to detect failures
• Roll back deployments when failures occur

Backup and Disaster Recovery

• Plan for backup and disaster recovery in Kubernetes applications
• Use tools like Chaos Mesh for testing and validation

Monitoring and Testing

• Monitor applications and nodes in Kubernetes
• Load test applications to ensure they can handle demand
• Validate test results using metrics and logging