Events Extension

The events extension surfaces Kubernetes events inside ArgoCD. Developers can see warnings, scheduling failures, image pull errors, and other cluster events for their resources without leaving ArgoCD.

Registration

The extension registers as:

  • Resource tabs on Pod, Deployment, StatefulSet, ReplicaSet, DaemonSet, Job, and CronJob resources
  • App view on any ArgoCD Application (Events)

What it shows

Resource tabs

Per-resource event views:

  • Kubernetes events filtered to the specific resource (by kind and name)
  • Event type color coding (Warning in red, Normal in green)
  • Time range picker (1h, 6h, 24h, 7d)
  • Expandable rows with full message, source component, host, and timestamps
  • Event count and relative timestamps (e.g. “5m ago”)
  • Auto-refresh every 30 seconds

Application view (Events)

Namespace-wide event aggregation for the entire application:

  • All events from the application’s destination namespace
  • Summary cards: total events, warnings, normal
  • Type filter buttons (All, Warning, Normal)
  • Free-text search across reason, message, object kind, and object name
  • Time range picker
  • Expandable detail rows with source and timestamp information

No external dependencies

Unlike other extensions, the events backend queries the Kubernetes API directly. No Prometheus, Loki, or other observability stack is required. If your cluster runs Kubernetes, events work.

Backend

The events backend queries the Kubernetes Events API:

EndpointDescription
GET /api/v1/eventsList events with filters

Configuration

Environment VariableRequiredDefaultDescription
PORTNo8085HTTP server port
LOG_LEVELNoinfoLog level
CLUSTER_NAMENoArgoCD name of the cluster this backend reads from. When set, apps whose spec.destination.name differs get a 403. See “Multi-cluster ArgoCD” on the ArgoCD Configuration page.
CLUSTER_SERVERNoAPI server URL of that cluster, matched against spec.destination.server
MULTI_CLUSTERNofalseRead events from the cluster each Application deploys to, instead of only this one. Mutually exclusive with CLUSTER_NAME/CLUSTER_SERVER. Requires rbac.namespacedRules.
ARGOCD_NAMESPACENopod’s own namespaceWhere the ArgoCD cluster Secrets live. Set by the Helm chart; only needed if ArgoCD runs elsewhere.

Reading other clusters

With MULTI_CLUSTER enabled, one backend beside ArgoCD serves every registered cluster, the same way the vulnerabilities backend does. It resolves each Application’s spec.destination against ArgoCD’s cluster Secrets and builds a client for that cluster, caching it for five minutes.

Enabling it takes two Helm values together:

extensions:
  events:
    env:
      MULTI_CLUSTER: "true"
    rbac:
      namespacedRules:
        - apiGroups: [""]
          resources: ["secrets"]
          verbs: ["list"]
Warning
MULTI_CLUSTER with the Secret rule lets this backend read argocd-secret and every registered cluster's credentials. Anyone who compromises the backend can then mint ArgoCD admin tokens. Read "Multi-cluster ArgoCD" on the ArgoCD Configuration page before enabling it.

Both default to off, so a single-cluster install is unaffected. The spoke’s credentials need read access to events there. Supported auth, the exposure the Secret rule carries, and the GKE Workload Identity requirement are covered under “Multi-cluster ArgoCD” on the ArgoCD Configuration page.

RBAC

The backend needs read access to Kubernetes events:

rules:
  - apiGroups: [""]
    resources: ["events"]
    verbs: ["list", "watch"]
  - apiGroups: ["events.k8s.io"]
    resources: ["events"]
    verbs: ["list", "watch"]

ArgoCD RBAC

p, role:developer, extensions, invoke, events, allow