--- name: azure-waf-reliability metadata: category: WellArchitectedFramework description: >- Generates guidance for reliability, resilience, availability, redundancy, failure recovery and disaster recovery for Azure workloads, based on the Reliability pillar of the Azure Well-Architected Framework. Use when evaluating, designing or improving how an Azure workload withstands and recovers from failure. --- # Azure Well-Architected Framework skill for the Reliability pillar ## Overview The Reliability pillar helps a workload withstand faults and recover to a working state within agreed targets. Decisions start from business requirements, expressed as availability and recoverability targets, and they are weighed flow by flow, because not every part of a workload needs the same protection. ## Core principles The recommendations below follow the published framework. Written for this agent from Microsoft's published design review checklist. Codes are Microsoft's. - **Keep the design simple** (RE:01): Add only the complexity that a business goal justifies, and prefer platform capabilities to custom code. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/simplify - **Rate user and system flows** (RE:02): Identify the flows through the workload and rate each for criticality, so effort goes to what matters most. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/identify-flows - **Analyse failure modes** (RE:03): List dependencies and failure points, with a mitigation for each. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/failure-mode-analysis - **Set reliability and recovery targets** (RE:04): Define SLOs, RTO and RPO, and build the health model on them. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/metrics - **Add redundancy where it counts** (RE:05): Replicate compute, data and network across availability zones, or regions, in proportion to each flow's criticality. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/redundancy - **Scale in time** (RE:06): Scale the application, data and infrastructure from actual or predicted use, with little manual work. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/scaling - **Self-preserve and self-heal** (RE:07): Use retries, circuit breakers, queues and health-based routing so the workload degrades gracefully and recovers by itself. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/self-preservation - **Test resiliency** (RE:08): Inject faults with chaos engineering to prove the workload withstands them and recovers within target. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/reliability-test - **Plan disaster recovery** (RE:09): Keep structured, documented and rehearsed plans that cover every component and the system as a whole. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/disaster-recovery - **Measure health continuously** (RE:10): Track uptime and reliability indicators across components and critical flows, and keep the data for detection and post-incident analysis. Grounding document: https://learn.microsoft.com/azure/well-architected/reliability/monitoring ## Relevant products The following are _examples_ of products relevant to this pillar: - **Compute**: Availability zones, Virtual Machine Scale Sets, AKS, App Service, Container Apps - **Networking**: Azure Front Door, Traffic Manager, Application Gateway, Azure Load Balancer - **Data**: Zone-redundant storage, geo-redundant storage, Azure SQL auto-failover groups, Cosmos DB multi-region - **Recovery**: Azure Backup, Azure Site Recovery - **Testing and health**: Azure Chaos Studio, Azure Monitor, Azure Service Health, Azure Resource Health ## Workload assessment questions Ask appropriate questions to understand the requirements and constraints of the workload and the organization. Choose from this list: - Which user and system flows matter most to the business, and what happens if each one stops? - What are the SLO, RTO and RPO for each critical flow? - Which dependencies, inside and outside the workload, can fail, and how is each failure handled? - How is the workload deployed across availability zones and regions? - How does it scale, and what are its quota and capacity limits? - How are backups taken and restores tested? - How often is failover rehearsed, and with what result? - How is workload health measured and reported? ## Validation checklist Use this checklist to evaluate the architecture against this pillar: - Critical flows are named and rated, and each has a target. - Failure mode analysis exists for the critical flows. - No critical flow depends on a single instance, zone or unmanaged dependency. - Zonal or regional redundancy matches the stated SLO, RTO and RPO. - Autoscaling is configured, and quotas are checked against peak demand. - Transient faults are handled with retries, timeouts and circuit breakers. - Backups exist for all stateful components, and restores are tested. - The disaster recovery plan is documented and rehearsed. - Health signals exist for each critical flow, with alerts.