Engineers keep patching live Deployments with kubectl, and the cluster no longer matches what the repository says. Which GitOps practice addresses this?
Show answer and explanation
Answer: C
In GitOps, the Git repository holds the declarative desired state of the cluster, and an agent continuously compares the live cluster with that state and corrects differences. Changes arrive through reviewed commits, which gives an audit trail and makes drift visible and self-correcting. Manual patches made directly on the cluster are overwritten by the reconcile loop, which is the intended outcome.
Why the other options are wrong
- A. Exporting manual changes after the fact keeps the cluster as the source of truth. Git then records what happened instead of driving the desired state, which defeats the GitOps model.
- B. Broad admin rights make drift more likely and remove the review step. Speed is not improved in a way that justifies losing the audit trail.
- D. A shared drive has no review or reconciliation. Copying files on request is manual and cannot guarantee that the cluster matches a single source of truth.