Open bq-arpit-20 opened 1 year ago
Based on my current understanding of flagger, it does not currently support multi-cluster deployment functionality. This is because resource creation across multiple clusters requires providing the KUBECONFIG
of each cluster, but flagger does not have an entry point to retrieve the KUBECONFIG
.
We employ a blue/green deployment strategy utilizing two distinct clusters, denoted as c1 and c2, both seamlessly integrated with an Application Load Balancer (ALB).
In our deployment approach, c1 functions as the primary cluster, catering to live production traffic. When updates or changes are required, we systematically implement them in the c2 cluster. Following this, a meticulous testing phase is conducted to ensure the environment's stability and compatibility.
Upon successful validation, we systematically transition the traffic routing mechanism from c1 to c2, initiating the "green" environment. Subsequently, c1 undergoes its update process and comprehensive testing regimen.
Upon the completion of these updates and tests, we seamlessly reroute the traffic back to c1, marking the transition from the "green" to the "blue" environment. This method ensures minimal disruptions, robust redundancy, and high availability while maintaining the highest quality standards throughout the deployment process.
How to achieve this from flagger?