Introduction

Brief overview of Azure Availability Zones and their importance in ensuring high availability. Importance of setting up Azure resources to withstand zone failures. General Guidelines for Azure Setup

Explanation of Azure Availability Zones and Regions. Best practices for designing resilient architectures in Azure. Overview of Azure SLAs related to availability and redundancy. Specific Considerations for AKS

Introduction to Azure Kubernetes Service (AKS) and its benefits. Understanding AKS availability zones support and limitations.

Designing AKS Clusters for Zone Resilience

Choosing the right Azure region and availability zones for AKS deployment. Configuring AKS with multiple availability zones. Best practices for distributing node pools across availability zones. Network and Security Considerations

Ensuring network redundancy and isolation in AKS clusters. Implementing network security policies to protect against zone failures. Using Azure Load Balancer and Application Gateway for zone-resilient traffic management. Data Management and Persistence

Strategies for deploying stateful applications in AKS across availability zones. Using Azure Disks and Azure Files for persistent storage across zones. Implementing data replication and backup strategies to safeguard against zone-level data loss. Monitoring and Recovery

Setting up Azure Monitor for AKS to monitor cluster health across zones. Implementing automated recovery and failover mechanisms. Using Azure Kubernetes Event-Driven Autoscaling (KEDA) for efficient resource utilization and resilience.

Conclusion

Summary of key takeaways for designing zone-resilient Azure setups, particularly focused on AKS. Future trends and advancements in Azure availability and resilience.

References

List of resources, Azure documentation links, and further reading materials.

Zone-failure simulation in Azure Kubernetes Service (AKS)

The resources in this repository aims to help in the simulation of Zone Failure of an Azure Kubernetes Service (AKS) cluster.

This repository includes a FastAPI-based Python API (server.py) which features 3 endpoints. Depending on the endpoints, the API fetches data either from the MySQL or SQL Server instance. For our testing we have used Azure Db for MySQL and Azure SQL MI respectively. The repo also includes the necessary Dockerfile and Kubernetes manifest files for Deployment and Service creation along with the Python-based insert scripts of the sample data for MySQL (mysqlinsertscript.py) and SQL Server (sqlserverinsertscript).

The repository also includes a bash-based test harness (apitestharness.sh) which has the capability to continuosly hit the desired endpoint exposed by the API and to print the HTTP success status.

Another important code file is the bash-based zone fail simulation (akszonefailsimulation.sh) which has the necessary steps to simulate the Zone Failure of the desired zone in Azure for the AKS cluster as well as Azure Db for MySQL and Azure SQL MI.

Source files

Database insert scripts

mysqlinsertscript.py: Python-based insert scripts of the sample data for MySQL sqlserverinsertscript: Python-based insert scripts of the sample data for SQL Server

Application files

server.py: FastAPI-based Python API requirements.txt: Python package requirements file

Kubernetes manifest and Docker files

Dockerfile: Dockerfile required for containerization of the API k8sdeployment.yaml: Kubernetes manifest file for deployment k8sservice.yaml: Kubernetes manifest file for service

Files needed for Zone Fail simulation

apitestharness.sh: A bash-based test harness which has the capability to continuosly hit the desired endpoint exposed by the API and to print the HTTP success status akszonefailsimulation.sh: A bash-based zone fail simulation which has the necessary steps to simulate the Zone Failure of the desired zone in Azure for the AKS cluster as well as Azure Db for MySQL and Azure SQL MI

Execution steps

1. Provision the Azure Db for MySQL - Flexible and Azure SQL MI database servers. Kindly note that these database servers need to be zone-redundant.
2. Provision the Zone-redundant AKS cluster. The system node pool of the AKS cluster should be spanning across all the 3 Availability Zones. 3 different user node pools should also be created - with each user node pool "pinned" to just one unique Availability Zone.
3. Set the database server details in the server.py file and containerize the application and push it to the desired Azure Container Registry (ACR).
4. Update the k8sdeployment.yaml file with the container image details from the ACR and create the K8S deployment in the AKS cluster.
5. Create the K8S service in the AKS cluster using k8sservice.yaml.
6. Update the apitestharness.sh file with the IP address of the K8S service and other endpoint details.
7. Update the akszonefailsimulation.sh with the necessary details as mentioned through the placeholders.
8. Execute (step-by-step to begin with) the akszonefailsimulation.sh script.