Sentinel: the cluster monitor that files tickets

Metrics dashboards answer "what is happening?". The question that actually costs you weekends is "what should I do about it?". Sentinel is our in-cluster monitor built around that second question.

Checks, not charts

Sentinel runs a battery of opinionated checks — pod and node health, PVC usage, certificate expiry, GitOps application state, backup recency, storage health, stale image tags, ageing sealed secrets — and reduces them to a short list of actionable maintenance items. The output reads like a to-do list, because that is what it is.

From log noise to a filed issue

The pod-log scanner tails logs across the namespace, fingerprints unique error signatures, and drafts a GitLab issue against the failing service's repository — summarised, deduplicated, and filed where the owning team will see it. A stability window keeps transient deploy restarts from generating noise.

Upgrades as merge requests

When a Helm chart has a new release, Sentinel opens a one-line MR against the infrastructure repo with a summarised release-notes brief. Staying current becomes review-and-merge. Critical events still page you in Slack — but far less often, because the routine work stopped queueing up.