Canary releases

Overview

Canary deployment in Facets is an Argo Rollouts-based deployment strategy that incrementally rolls out changes to a subset of your infrastructure before promoting to the full environment. It reduces risk by validating changes in a limited scope before full production exposure.

Supported Resource Type: service

Prerequisites

• A service resource configured in your Facets blueprint.
• A Kubernetes cluster with Argo Rollouts (included with Facets).

Required permissions:

• Plan Release / Apply Release Plan: to trigger releases.
• Promote: graduate a canary to stable. Abort: halt a canary rollout.

Step 1: Configure Canary Strategy on the Service Resource

In your service resource configuration JSON, set the deployment strategy to Canary and define rollout behavior:

{
  "spec": {
    "release": {
      "strategy": {
        "type": "Canary",
        "canary_strategy": {
          "enable_auto_abort": true,
          "auto_abort_timeout": 120,
          "max_surge": "1",
          "max_unavailable": 0,
          "min_ready_seconds": 30,
          "progress_deadline_seconds": 600,
          "scale_down_delay_seconds": 30
        }
      }
    }
  }
}

Key Configuration Options

Every key below sits under spec.release.strategy.canary_strategy on the service resource. Options left unset fall back to the default shown.

Step 2: Trigger a Release

• Navigate to your Project → Environment in the Facets UI. Go to Releases and click Trigger Release.
• Select the service resource(s) with canary strategy configured.
• Click Plan to preview the release changes (dry-run).
• Review the plan output and click Apply to begin the canary rollout.
• Facets will initiate the Argo Rollouts controller, which creates the canary pod set alongside the stable set.

Step 3: Monitor the Canary Rollout

During an active canary rollout, monitor progress via:

• Live Insights & Operations dashboard: real-time pod-level metrics (CPU, memory, restart count).
• Kubernetes Dashboard: pod status, events, and deployment progression.
• Kubernetes Events Dashboard: filter events by namespace to track rollout activity.

Watch for:

• Pod restart frequency or CrashLoops in canary pods.
• Elevated error rates or latency from canary traffic.
• progress_deadline_seconds breaches indicating a stuck rollout.

Step 4: Promote or Abort

Promote (Graduate Canary to Stable)

If the canary is healthy and validated:

• In the Facets UI, navigate to the resource's Live Tab. Click the Promote action.
• This advances the canary version to stable and scales down the old stable pods.

Manual vs. automatic promotion: A canary advances through its configured steps on its own. Set enable_manual_rollout: true to start the rollout paused instead, so it only moves forward when you click Promote. The canary strategy has no timed auto-promotion setting; auto_promotion_seconds belongs to the blue-green strategy.

Abort (Rollback Canary)

If issues are detected:

• Click the Abort action in the UI.
• Argo Rollouts immediately stops the canary rollout and scales the canary pods down, restoring full traffic to the stable version.

Automatic abort: If enable_auto_abort: true and the rollout exceeds auto_abort_timeout seconds, Facets aborts automatically, preventing prolonged exposure to a faulty update.

Canary Metadata (Optional)

You can apply distinct labels and annotations to stable vs. canary pod sets for observability and traffic filtering:

{
  "stable_metadata": {
    "labels": { "track": "stable" },
    "annotations": {}
  },
  "canary_metadata": {
    "labels": { "track": "canary" },
    "annotations": {}
  }
}

This is useful for routing canary traffic through a separate ingress path or for filtering metrics by deployment track in your monitoring stack.