Progressive Delivery with Argo Rollouts & Flagger

Manage progressive deployments using kubectl-mcp-server's rollout tools (11 tools).

Requires the rollouts toolset: start kubectl-mcp-server with --toolsets default,rollouts (or MCP_TOOLSETS=default,rollouts).

When to Apply

Use this skill when:

• User mentions: "canary", "blue-green", "progressive delivery", "Argo Rollouts", "Flagger"
• Operations: rolling out new versions, traffic splitting, automated rollbacks
• Keywords: "gradual rollout", "traffic shift", "analysis run", "promote", "abort"

Priority Rules

Priority	Rule	Impact	Tools
1	Detect Argo Rollouts installation first	CRITICAL	detect_ecosystem
2	Check rollout status before promoting	HIGH	get_rollout_status
3	Monitor analysis runs for failures	HIGH	get_analysis_runs
4	Abort immediately on critical failures	CRITICAL	abort_rollout

Quick Reference

Task	Tool	Example
Detect Argo Rollouts	detect_ecosystem	detect_ecosystem()
List rollouts	get_rollouts_list	get_rollouts_list(namespace)
Get rollout status	get_rollout_status	get_rollout_status(name, namespace)
Promote rollout	promote_rollout	promote_rollout(name, namespace)

Check Installation

detect_ecosystem()

Argo Rollouts

List Rollouts

get_rollouts_list(namespace="default")

# Shows:
# - Rollout name
# - Strategy (canary/blueGreen)
# - Status
# - Desired/Ready replicas

Get Rollout Details

get_rollout(name="my-rollout", namespace="default")

# Shows:
# - Spec (strategy, steps)
# - Status (phase, conditions)
# - Current step

Check Rollout Status

get_rollout_status(name="my-rollout", namespace="default")

# Returns detailed status with:
# - Current step index
# - Canary weight
# - Stable/canary replicasets

Promote Rollout

# Promote to next step
promote_rollout(name="my-rollout", namespace="default")

# Full promote (skip remaining steps)
promote_rollout(name="my-rollout", namespace="default", full=True)

Abort Rollout

abort_rollout(name="my-rollout", namespace="default")
# Reverts to stable version

Retry Rollout

retry_rollout(name="my-rollout", namespace="default")
# Retry failed rollout

Restart Rollout

restart_rollout(name="my-rollout", namespace="default")
# Triggers new rollout with same spec

Analysis Runs

# List analysis runs
get_analysis_runs(namespace="default")

# Analysis runs verify rollout health:
# - Prometheus metrics
# - Web hooks
# - Custom jobs

Create Canary Rollout

kubectl_apply(manifest="""
apiVersion: argoproj.io/v1alpha1
kind: Rollout
metadata:
  name: my-rollout
  namespace: default
spec:
  replicas: 5
  strategy:
    canary:
      steps:
      - setWeight: 20
      - pause: {duration: 1m}
      - setWeight: 40
      - pause: {duration: 1m}
      - setWeight: 60
      - pause: {duration: 1m}
      - setWeight: 80
      - pause: {duration: 1m}
  selector:
    matchLabels:
      app: my-app
  template:
    metadata:
      labels:
        app: my-app
    spec:
      containers:
      - name: app
        image: my-app:v2
        ports:
        - containerPort: 8080
""")

Create Blue-Green Rollout

kubectl_apply(manifest="""
apiVersion: argoproj.io/v1alpha1
kind: Rollout
metadata:
  name: my-rollout
  namespace: default
spec:
  replicas: 3
  strategy:
    blueGreen:
      activeService: my-app-active
      previewService: my-app-preview
      autoPromotionEnabled: false
  selector:
    matchLabels:
      app: my-app
  template:
    metadata:
      labels:
        app: my-app
    spec:
      containers:
      - name: app
        image: my-app:v2
""")

Flagger

List Canaries

get_flagger_canaries(namespace="default")

# Shows:
# - Canary name
# - Status (Initialized, Progressing, Succeeded, Failed)
# - Weight

Get Canary Details

get_flagger_canary(name="my-canary", namespace="default")

Create Flagger Canary

kubectl_apply(manifest="""
apiVersion: flagger.app/v1beta1
kind: Canary
metadata:
  name: my-canary
  namespace: default
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  service:
    port: 80
  analysis:
    interval: 30s
    threshold: 5
    maxWeight: 50
    stepWeight: 10
    metrics:
    - name: request-success-rate
      threshold: 99
      interval: 1m
    - name: request-duration
      threshold: 500
      interval: 1m
""")

Progressive Delivery Workflows

Canary Deployment

1. get_rollouts_list(namespace)
2. # Update image in rollout
3. get_rollout_status(name, namespace)  # Monitor progress
4. promote_rollout(name, namespace)  # Promote when ready
5. # Or: abort_rollout(name, namespace) if issues

Blue-Green Deployment

1. get_rollout(name, namespace)  # Check current state
2. # Update image
3. get_rollout_status(name, namespace)  # Wait for preview ready
4. # Test preview service
5. promote_rollout(name, namespace)  # Switch traffic

Troubleshooting

Rollout Stuck

1. get_rollout_status(name, namespace)  # Check current step
2. get_analysis_runs(namespace)  # Check analysis
3. kubectl_get(resource_type="events", namespace)  # Check events
4. # If analysis failing:
   abort_rollout(name, namespace)

Canary Failing Analysis

1. get_analysis_runs(namespace)
2. # Check metrics source (Prometheus, etc.)
3. # Verify threshold configuration
4. retry_rollout(name, namespace)  # Retry if transient

Related Skills

• k8s-deploy - Standard deployments
• k8s-service-mesh - Traffic management with Istio