Skip to main content
This page is the implementation-level reference for Radar’s Argo CD and Flux integrations. Start with the GitOps workspace for the shared workflow and controller capability comparison, or the FluxCD integration for a screenshot-led Flux walkthrough and setup guidance.

Flux CRDs

Radar extracts failing Ready, Healthy, Released, and TestSuccess conditions, plus Stalled=True and Reconciling=True state, into issues. Kustomizations and HelmReleases support Reconcile, Sync with source, Suspend, and Resume. Flux desired manifests are not currently available to Radar for Git-to-live field diffs. HelmRelease-managed resources also do not carry the client-side last-applied annotation that backs Radar’s local diff fallback. See #601 for desired-state coverage work.

Argo CD CRDs

Application diagnosis includes operation failures, Application conditions, unhealthy or missing children, manual drift, and drift that persists after a successful automatic sync. Managed resources can include recent Kubernetes events and field-level drift. Applications support Sync with options, Refresh, Hard refresh, Terminate, Suspend or restore auto-sync, Rollback, and Selective sync.
Field-level drift depends on desired-state availability. The local fallback requires kubectl.kubernetes.io/last-applied-configuration; server-side-applied resources do not carry that annotation. When configured, Radar’s Argo CD API integration can retrieve Git-rendered desired state for canonical resource diffs.

Shared lifecycle and scope

For both controllers, Radar treats a resource with metadata.deletionTimestamp as terminating. It suppresses stale sync and health state, disables mutating operations, reports deletion age and finalizer ownership, and adds controller-health attribution when RBAC permits. Topology is cluster-local. Radar can connect a GitOps parent to managed resources only when those resources are visible in the connected cluster. This is usually true for Flux. Argo CD hub-and-spoke deployments may show the Application without workload edges when the workloads live in another cluster.