Helm Values & Environment Variables

Reference for all environment variables and Helm values used by Wave.

Core Configuration Parameters

Essential Environment Variables

License & Registry Credentials

License and image-pull credentials are set as Helm values, not environment variables:

Data Management

Worker Intervals

Monitoring & Metrics

Forecast Intervals

Platform Features

Feature Toggles

Wave 3.4.0+. Each background task can be turned off at deploy time. Turning a task off stops the task from starting at all — Core does not collect its data or run its detector, so there is no CPU, memory, or storage footprint for that feature.

How a disabled feature appears in the console. The menu entry stays visible and the page renders a panel telling you which environment variable to set to turn the feature back on. Wave Karpenter and Wave GPU behave differently: when off, their menu entries are hidden entirely.

WA_NF_ENABLED is folded together with the license: NetFUNNEL shows as available only when the license grants it and the environment variable is true. Setting it to false turns NetFUNNEL off even on a licensed cluster.

Advanced Features

Component Configuration

Autopilot Configuration

OpenStack Configuration

Karpenter Configuration

Karpenter Settings

Webhook TLS Settings

Example Configuration

# Karpenter Integration (EKS only)
karpenter:
  enabled: true
 
# Webhook TLS (required for Spot Workload Placement)
webhooks:
  tls:
    enabled: true
    port: 9443
    certManager:
      enabled: true
      duration: "8760h"
      renewBefore: "720h"
    secretName: "wave-autoscale-webhook-tls"

See the Karpenter Getting Started guide for detailed setup instructions.

Storage Configuration

Persistent Volume Configuration

spec:
  storageClassName: "gp3"  # AWS EKS example
  existingPvcName: ""       # Use existing PVC or leave empty for auto-creation

Storage Class Examples by Platform

• AWS EKS: gp3, gp2, io1
• GKE: standard, standard-rwo, premium-rwo
• Azure AKS: managed-premium, azurefile
• On-premises: local-path, nfs, ceph-rbd

Platform-Specific Configuration

Red Hat OpenShift

openshift:
  waveAutoscale:
    scc:
      create: true  # Create custom SCC for core components
  waveAutoscaleAgent:
    scc:
      create: true  # Create custom SCC for agent DaemonSet

AWS EKS with Fargate

spec:
  core:
    nodeSelector:
      eks.amazonaws.com/compute-type: fargate

Cluster Permissions

Standard Configuration

clusterRole:
  readOnly: false    # Full automation capabilities
  istioAdmin: false  # Set true for Istio service mesh integration

Read-Only Mode

clusterRole:
  readOnly: true     # Limited to monitoring only
  istioAdmin: false

Resource Sizing Guidelines

Wave resource requirements scale with cluster size (~1% of total cluster capacity):

Note: WA Metrics Agent requires 0.1 vCPU / 50MB per node (deployed as DaemonSet).

Complete Configuration Example

Most installs only need a license, image-pull credentials, and a StorageClass; the published chart already sets image repositories, tags, and default resources. Override per-component values only when you need to.

# Image-pull secret for ghcr.io/stclab-inc/* (provided by the Wave team)
ghcr:
  enabled: true
  dockerconfigjson: "eyJhdXRocyI6eyJnaGNyLmlvIjp7...}}"
 
# License: inline, or reference a pre-created Secret via license.existingSecret
license:
  key: "YOUR-LICENSE-KEY-HERE"
 
# Install the 14 wavek8s.com CRDs with the chart (default true)
crds:
  install: true
 
# Cluster permissions
clusterRole:
  readOnly: false
  istioAdmin: true
 
# Per-component overrides (optional; the chart ships sensible defaults)
spec:
  core:
    image:
      repository: ghcr.io/stclab-inc/wave-autoscale-core
      tag: <version>
    env:
      - name: WA_LOG_LEVEL
        value: "info"
      - name: WA_DATA_SAVE_DAYS
        value: "30"
    resources:
      requests:
        cpu: "1500m"
        memory: "3Gi"
      limits:
        cpu: "1500m"
        memory: "3Gi"
 
  webConsole:
    image:
      repository: ghcr.io/stclab-inc/wave-autoscale-web-console
      tag: <version>
    resources:
      requests:
        cpu: "100m"
        memory: "100Mi"
 
  intelligence:
    image:
      repository: ghcr.io/stclab-inc/wave-autoscale-intelligence
      tag: <version>
    resources:
      requests:
        cpu: "500m"
        memory: "1Gi"
 
  agent:
    image:
      repository: ghcr.io/stclab-inc/wave-autoscale-agent
      tag: <version>
    resources:
      requests:
        cpu: "100m"
        memory: "50Mi"
 
  cadvisor:
    image:
      repository: gcr.io/cadvisor/cadvisor
      tag: v0.52.1
    resources:
      requests:
        cpu: "100m"
        memory: "200Mi"
 
  # Storage configuration
  storageClassName: "gp3"
  existingPvcName: ""
 
# Platform-specific settings (uncomment as needed)
# openshift:
#   waveAutoscale:
#     scc:
#       create: true
#   waveAutoscaleAgent:
#     scc:
#       create: true

Validation and Troubleshooting

Pre-Installation Checklist

1. Verify Kubernetes version compatibility (≥ 1.26)
2. Ensure sufficient cluster resources based on sizing guidelines
3. Validate license key availability
4. Confirm storage class availability
5. Check network policies allow component communication

Common Configuration Issues

License Validation

kubectl logs -n wave-autoscale wave-autoscale-0 -c core | grep LICENSE

Metrics Collection

# Verify agent is collecting metrics
kubectl logs -n wave-autoscale -l app=wave-autoscale-agent | grep metrics

Storage Issues

# Check PVC status
kubectl get pvc -n wave-autoscale

Performance Tuning

High-Load Clusters (>1000 vCPUs)

• Increase WA_METRIC_STORE_SAVE_BATCH_WORKER_CNT to 100
• Set WA_EVENT_QUEUE_CHANNEL_CAPACITY to 128
• Adjust worker intervals based on workload volatility

Limited Resources

• Set WA_DATA_SAVE_DAYS to 7 to reduce storage requirements
• Increase worker intervals to reduce CPU usage
• Consider read-only mode for monitoring-only deployments

Security Considerations

Network Policies

Ensure proper network policies allow:

• Core component communication on ports 3024-3027
• Agent communication with kubelet (port 10250)
• External access to Web Console (port 3025) if required

RBAC Configuration

Wave requires specific RBAC permissions:

• Read access to all Kubernetes resources
• Write access to workload resources (Deployments, StatefulSets, etc.)
• Metrics collection from nodes and pods

GPU Features (preview)

Wave's GPU features are gated behind feature flags during the preview period. Enable them via these env vars:

See the Wave GPU section for feature details.