Set a short reconciliation interval on the Kustomization object to reduce the window between a drift event and its detection
Add a healthChecks block listing the Deployments, StatefulSets, or other workload resources whose Ready status Flux should verify after applying manifests
Configure Flux notification-controller with an Alert targeting a Provider such as Slack or PagerDuty, and use the event severity level to differentiate drift warnings from hard failures
Simulate drift by manually scaling a Deployment outside of Git and observe the Kustomization condition transition to a failed or degraded state within one reconciliation interval
Review the Flux event stream and notification delivery to confirm the alert reaches the configured provider and contains actionable context about the drifted resource
Known gotchas
Flux drift detection reverts changes on the next reconciliation cycle rather than alerting and halting; if you need a break-glass override without immediate revert, you must suspend the Kustomization object before making the manual change
healthChecks only validate readiness of explicitly listed resources; resources not listed can be degraded without triggering a Kustomization failure, so incomplete healthChecks give false confidence
The notification-controller Alert filters events by source kind and name; a misconfigured event source selector means no alerts are ever sent even when reconciliation fails
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp