Kubernetes Deployment Validation in Two Minutes
A green deployment result does not guarantee a healthy Kubernetes application. Pods can enter crash loops, fail to become ready or leave a workload unstable even after the rollout reports success. Checking those conditions manually means moving between clusters, dashboards and logs. Across an environment with many clusters, that repetitive work can consume time engineers would otherwise spend investigating failures or improving reliability.
Sai Joshitha Kathari, senior site reliability engineer at Visa, joins Alan Shimel to explain how she moved those checks into the deployment pipeline. Building on her Cloud Native Now article on automating deployment validation, Kathari describes reducing validation time from 45 minutes to two. Her approach starts with the checks engineers already perform, rather than treating a completed deployment as sufficient evidence that the application is running correctly. It makes validation part of the release workflow instead of a separate manual exercise afterward.
The Jenkins pipeline captures cluster state before deployment, then checks pod readiness, failure rates and workload stability after the new version rolls out. Crash loops, failed pods and workloads that do not come up appear in the pipeline logs. That changes where troubleshooting begins. Instead of visiting every cluster to discover whether something went wrong, an engineer can identify the affected cluster and use the available logs to investigate. Successful deployment and successful validation remain distinct results, both of which matter before declaring the application healthy.
Kathari presents the work as a starting point other SRE and DevOps teams can adapt, and she invites improvements from engineers solving similar problems. The benefit is not simply completing a checklist faster. Bringing validation into CI/CD gives engineers a more useful account of what happened during a rollout and directs their attention toward the systems that need it. Time recovered from repetitive inspection can go into troubleshooting, further automation and application reliability work.
