Shorticle 997 – Release management with Google Cloud Deploy and Flagger
Deploying to cloud services should be carefully designed to handle controlled management of cloud resources so that any failure due to incorrect configuration or high load can be easily handled by reverting or resolving the problem quickly. Also, we need to use multiple stages of environments like development, testing, QA/UAT and production so that release can be managed through a staged approach and a workflow to be created to promote deployment once it passes its previous environment/stage of deployment.
This kind of deployment process is called canary deployment and integrated cloud native services helps to achieve canary deployment through its DevOps pipeline. For example, in Google cloud, we can develop canary deployment architecture by using Google Cloud deploy, Flagger and Kubernetes Gateway API to build and deploy Kubernetes applications in Google Kubernetes Engine (GKE).
The advantage of canary deployment is you can fail fast at early stage as failure at higher stage proves to be costly for business disruption and hence you work on a controlled deployment using this deployment model. Cloud deploy is Google native service for build and deployment of cloud native applications including Kubernates application.
Flagger is a popular deployment tool popularly used as cloud agnostic deployment particularly in ServiceMesh (eg: Istio or Anthos) and Kubernetes applications. It can also monitor the healthcheck of applications and prepares for self-healing services for clustering additional services as needed depending on service failure and it integrates with Prometheus and Google cloud monitoring for health monitoring of cloud services.
领英推荐
Kubernetes Gateway API helps to manage traffic for Kubernetes clusters and if any additional instances are spun up with Flagger, it gets attached to load balancer through Gateway API and workload traffic are distributed to all the clustered nodes easily.