Client
US food manufacturer
Industry
Food and dairy manufacturing
Service
24x7 managed Kubernetes
Scale
About 45 microservices
Manufacturing does not stop for software. When an application that supports production or orders goes down, the plant feels it. The client's platform had grown into dozens of microservices across on-premises and cloud Kubernetes, with releases managed from an older version-control set-up and little visibility into why services restarted.
The client wanted one team accountable for keeping the platform healthy, a safer way to ship changes, and a recovery path if a cluster failed.
The desired state of every service lives in Git, and automated pipelines build, test and deploy it. Each change is reviewed, repeatable and easy to roll back.
A standby cluster mirrors production so workloads can fail over if the primary environment has a problem. We test the failover path rather than assuming it works.
Prometheus and Grafana track cluster and service health, uptime checks watch the endpoints users depend on, and alerts go to the 24x7 team. Message queues are monitored alongside the services that use them.
The service runs against defined uptime and response targets, with tiers the client chooses by how critical each application is.
Every month we report on each microservice restart: when it happened, what caused it and what we changed. Early reports surfaced repeat causes, such as data-serialisation errors and dropped database connections. Fixing those at the source, rather than restarting and moving on, has cut restarts significantly.
Container scanning
We scan the images in the client's private container registry for known vulnerabilities and track fixes by severity.
Backup and recovery
Next step proposed: daily snapshots of cluster state, backups of message-queue data, and offsite, tamper-proof copies, so the platform can be rebuilt even after a serious incident.
Agentic AI runs on the same platforms that run the business today. A Kubernetes estate with GitOps, tested failover, root-cause discipline and clean images is the foundation that lets a manufacturer add AI agents to operations safely. It is also where site-reliability agents such as AgentSRE can take on monitoring and root-cause work next.
See how the stack fits together on the XenonStack ecosystem page and in ElixirClaw. To see what happens before you book, read how the Context Workshop works.