Kubernetes rollouts and rollbacks, recorded automatically

Not every change to a cluster comes from a pipeline. Someone scales a deployment by hand, an HPA reacts to load, a rollout gets reverted at two in the morning. The kube-watcher observes the cluster directly, so those changes reach the timeline whether or not anything told CI about them.

What lands on the timeline

Deploy: A Deployment, StatefulSet, or DaemonSet completed a rollout, with image tag, replica count, and duration.
Rollback: A workload's image reverted to a previously seen tag, recorded as a rollback rather than a generic deploy.
Scale: A replica count changed, manually or via HPA, with direction and old and new counts.
Teardown: A workload was removed. Requires reportDeletes.
Alert: Pod failures: CrashLoopBackOff, image pull errors, OOMKilled, and unschedulable pods.

Setup

1Get a CLI token
On the dashboard, go to Settings, then CLI Token.

2Install the watcher

It runs as a Deployment using a ServiceAccount with read-only access to Deployments, StatefulSets, DaemonSets, and Pods. Install one release per cluster.

helm install kollaber-watcher oci://ghcr.io/urbangeeks/charts/kube-watcher \
  --set kollaber.env=prod \
  --set kollaber.api=https://kollaber.io \
  --set kollaber.token=<cli-token>

3Scale something
Change a replica count and the scale event appears on the timeline within seconds.

Good to know

For several clusters, install one release each and point them at matching environments with kollaber.env.
If you manage secrets externally with Vault, Sealed Secrets, or External Secrets Operator, set kollaber.existingSecret to a secret with a token key instead of letting the chart create one.
Set watchNamespace to limit the watcher to a single namespace. Empty means all namespaces.
The binary also runs outside the cluster against your kubeconfig, which is useful for trying it before deploying anything.

Questions

What access does the watcher need?

Read-only access to Deployments, StatefulSets, DaemonSets, and Pods through a ServiceAccount. It never writes to the cluster.

Does this collect metrics or logs?

No. It watches workload and pod state changes and sends events. Kollaber collects no metrics, logs, or traces.

How does it tell a rollback from a deploy?

It tracks image tags it has already seen for a workload. When a tag reverts to a previous one, the event is recorded as a rollback instead of a deploy.

Other integrations

GitHub Actions Argo CD HCP Terraform Atlantis Prometheus Alertmanager

Start recording Kubernetes changes

The free plan covers two environments and five members. The full reference for this integration lives in the docs.

Get started free All integrations