values.yaml

# Default values for groundlight-edge-endpoint.
# This is a YAML-formatted file.
# Declare variables to be passed into your templates.

# Global configuration for dependencies
global:
  # Enable OpenTelemetry dependency when needed for logging
  # Use: --set global.otelEnabled=true for local-splunk or cloud-splunk modes
  otelEnabled: false

# The Kubernetees namespace where the Edge Endpoint will be deployed
# We recommend that you don't change this value unless you're doing something unusual
# like running more that one instance of the Edge Endpoint in the same cluster
namespace: "edge"

# The tag to use for the images. The value of `imageTag` is used for both the edge-endpoint and
# inference images, unless you add a specific tag for one or the other image
imageTag: "release"
edgeEndpointTag: ""
inferenceTag: ""

# Whether to use the minimal image for the inference server. Currently, the minimal image only
# supports binary and multiclass detectors, so be sure to keep this set to false if you're using
# an object detection or counting model. 
useMinimalImage: false

# The image pull policy for the containers.
# The default value is "Always" which means that Kubernetes will always check to see if there's 
# a new version of the image with the requested tag available when starting containers.
imagePullPolicy: "Always"

# The port that the Edge Endpoint will listen on on the host. 
# Within the cluster, edge-endpoint always listens on port 30101 and can be addressed by 
# other services in the cluster on that port using
# http://edge-endpoint-service.edge.svc.cluster.local:30101/ (or substitute the appropriate
# namespace if you've overriden the default value).
edgeEndpointPort: 30101
edgeEndpointHttpsPort: 30143

# This is used as the base of the name for the PersistentVolume. The full name of the volume is 
# created by appending the namespace to this value. Generally, this should not be overridden.
# The PersistentVolume is used to store the model files and other data that the Edge Endpoint wants
# to persist between restarts. It is mapped to `/opt/groundlight/edge` on the host.
persistentVolumeNameBase: "edge-endpoint-pv"

# Edge Endpoint will default to using the GPU for inference. If you want to use the CPU instead,
# set this value to "cpu".
inferenceFlavor: "gpu"

# The user must provide the Groundlight API token as input or the deployment will fail
groundlightApiToken: ""

# For escalations and audits and calls that aren't handled by the Edge Endpoint, we forward
# to the Groundlight service in the cloud. If you're testing against another version
# of the Groundlight service (e.g., your own dev environment or the integ environment),
# you can override this value.
upstreamEndpoint: "https://api.groundlight.ai"

# Currently, all Groundlight services are deployed in the us-west-2 region
awsRegion: "us-west-2"

ecrRegistry: "767397850842.dkr.ecr.us-west-2.amazonaws.com"

# This sets the log level for all the containers, both edge endpoint and inference.
logLevel: "INFO"

# These values override the automated settings in _helpers.tpl to keep this short and sweet
# Don't override these
nameOverride: "edge-endpoint"
fullnameOverride: "edge-endpoint"

# Set some sensible limits on memory usage to avoid system crashing
k3sConfig:
  enabled: true
  evictionHardPercent: "15"
  evictionSoftPercent: "25" 
  evictionHardMinGB: "4"
  evictionSoftMinGB: "8"
  evictionGracePeriod: "10s"

serviceAccount:
  create: true
  name: ""

# Logging Configuration
# loggingMode options:
#   - "standard": Basic logging to stdout/files (no Splunk, no OTel)
#   - "local-splunk": Deploy local Splunk container
#   - "cloud-splunk": Does not deploy local Splunk container, points to cloud instance instead
loggingMode: "standard"

# Balena device configuration
balena:
  deviceUuid: "none"

# Splunk Configuration (used for local-splunk and cloud-splunk modes)
splunk:
  # Local Splunk container settings (for local-splunk mode)
  local:
    image:
      repository: splunk/splunk
      tag: "9.3.5"
    password: "admin123"
    hecToken: "abcd1234-5678-90ef-ghij-klmnopqrstuv"
    service:
      type: NodePort
      webNodePort: 30080    # Splunk Web UI
      hecNodePort: 30088    # HTTP Event Collector
      managementNodePort: 30089  # Management port
    persistence:
      dataSize: "10Gi"
      etcSize: "2Gi"
      storageClass: ""  # Use default storage class
    resources:
      limits:
        cpu: 2000m
        memory: 4Gi
      requests:
        cpu: 500m
        memory: 2Gi
  
  # Cloud Splunk settings (for cloud-splunk mode)
  cloud:
    endpoint: "https://your-splunk-instance:8088/services/collector"
    token: "your-hec-token-here"
  
  # Common Splunk settings
  index: "edge_app"
  source: "edge-endpoint"
  sourcetype: "edge:endpoint:logs"
  
  # OpenTelemetry Collector configuration
  collector:
    image:
      repository: otel/opentelemetry-collector-contrib
      tag: "0.91.0"
    resources:
      limits:
        cpu: 200m
        memory: 256Mi
      requests:
        cpu: 100m
        memory: 128Mi

# OpenTelemetry Collector dependency configuration
# This configures the official OpenTelemetry Helm chart as a dependency
opentelemetry-collector:
  # Enable the OpenTelemetry collector when logging modes require it
  # This gets set dynamically based on loggingMode
  enabled: false  # Will be overridden via --set at deployment time
  
  # DaemonSet mode for collecting logs from all pods on each node
  mode: daemonset
  
  namespaceOverride: "edge"
  
  nameOverride: "opentelemetry-collector"
  fullnameOverride: "opentelemetry-collector"
  
  # Use the contrib image which includes all receivers/processors
  image:
    repository: otel/opentelemetry-collector-contrib
    tag: "0.91.0"
  
  # Resource configuration
  resources:
    limits:
      cpu: 200m
      memory: 256Mi
    requests:
      cpu: 100m
      memory: 128Mi
  
  # Preset configurations
  presets:
    kubernetesAttributes:
      enabled: true
    logsCollection:
      enabled: false  # Disabled due to unsupported 'container' operator in v0.91.0
    kubeletMetrics:
      enabled: false
  
  config:
    receivers:
      filelog:
        include:
          - /var/log/pods/*/*/*.log
        exclude:
          - /var/log/pods/edge_opentelemetry-collector*/*/*.log
        start_at: end
        include_file_path: true
        include_file_name: false
        operators:
          # Parse CRI format logs manually
          - type: regex_parser
            id: extract_metadata_from_filepath  
            regex: '^.*\/(?P<namespace>[^_]+)_(?P<pod_name>[^_]+)_(?P<uid>[a-f0-9\-]{36})\/(?P<container_name>[^\._]+)\/(?P<restart_count>\d+)\.log$'
            parse_from: attributes["log.file.path"]
          # Parse CRI log format: timestamp stream partial_flag log_message
          - type: regex_parser
            id: parse_cri_format
            regex: '^(?P<timestamp>[^\s]+) (?P<stream>stdout|stderr) (?P<partial>P|F) (?P<message>.*)$'
            parse_from: body
            timestamp:
              parse_from: attributes.timestamp
              layout: '%Y-%m-%dT%H:%M:%S.%LZ'
          # Extract log level from the message content
          - type: regex_parser
            id: extract_log_level
            regex: '(?:^|.*\s)(?P<level>DEBUG|INFO|WARN|WARNING|ERROR|CRITICAL|FATAL)(?:[:\s]|$)'
            parse_from: attributes.message
            on_error: 'send'
          # Parse JSON logs that are already structured
          - type: json_parser
            id: parse_json_logs
            parse_from: attributes.message
            on_error: 'send'  # Continue if not JSON
          # Parse FastAPI/Uvicorn HTTP access logs: "INFO: 10.42.0.1:50126 - "GET /health/ready HTTP/1.1" 200 OK"
          - type: regex_parser
            id: extract_fastapi_access_log
            regex: '(?P<client_ip>\d+\.\d+\.\d+\.\d+):(?P<client_port>\d+)\s+-\s+"(?P<http_method>\w+)\s+(?P<http_path>[^?\s]+)(?:\?(?P<http_query>[^"\s]*?))?(?:\s+[^"]*)?"\s+(?P<status_code>\d+)'
            parse_from: attributes.message  
            on_error: 'send'
          # Parse NGINX access logs: "NGINX: 10.42.0.1 - - [24/Sep/2025:17:16:24 +0000] "GET /health/ready HTTP/1.1" 200 18 "-" "kube-probe/1.33""
          - type: regex_parser
            id: extract_nginx_access_log
            regex: 'NGINX:\s+(?P<client_ip>\d+\.\d+\.\d+\.\d+)\s+-\s+-\s+\[(?P<nginx_timestamp>[^\]]+)\]\s+"(?P<http_method>\w+)\s+(?P<http_path>[^?\s]+)(?:\?(?P<http_query>[^"\s]*?))?(?:\s+[^"]*)?"\s+(?P<status_code>\d+)\s+(?P<response_size>\d+)\s+"(?P<referer>[^"]*)"\s+"(?P<user_agent>[^"]*)"'
            parse_from: attributes.message
            on_error: 'send'
    
    processors:
      # Add resource attributes for custom fields
      resource:
        attributes:
          - key: service.name
            value: "edge-endpoint"
            action: upsert
          - key: deployment.environment
            value: "edge"
            action: upsert
          - key: balena.device.uuid
            value: "${env:BALENA_DEVICE_UUID}"
            action: upsert
    
    exporters:
      # Splunk HEC exporter configuration
      splunk_hec:
        endpoint: "http://splunk.edge.svc.cluster.local:8088/services/collector"
        token: "abcd1234-5678-90ef-ghij-klmnopqrstuv"
        index: "edge_app"
        source: "edge-endpoint"
        sourcetype: "edge:endpoint:logs"
        disable_compression: false
        timeout: 10s
        retry_on_failure:
          enabled: true
          initial_interval: 5s
          max_interval: 30s
          max_elapsed_time: 120s
    
    service:
      pipelines:
        logs:
          receivers: [filelog]
          processors: [k8sattributes, resource, memory_limiter, batch]
          exporters: [splunk_hec]
  
  # Mount host directories for log collection
  extraVolumes:
    - name: varlogpods
      hostPath:
        path: /var/log/pods
    - name: varlibdockercontainers
      hostPath:
        path: /var/lib/docker/containers
  
  extraVolumeMounts:
    - name: varlogpods
      mountPath: /var/log/pods
      readOnly: true
    - name: varlibdockercontainers
      mountPath: /var/lib/docker/containers
      readOnly: true
  
  # Environment variables for custom attributes  
  extraEnvs:
    - name: BALENA_DEVICE_UUID
      valueFrom:
        configMapKeyRef:
          name: balena-config
          key: device-uuid
  
  # Security context to read log files  
  securityContext:
    runAsUser: 0
    runAsGroup: 0

# Network healer - automatically recovers service networking after host IP changes
#
# An alternative experimental approach is implemented on the `tim/network-hooks` branch:
# a NetworkManager dispatcher hook that restarts k3s when network/IP changes are detected.
# That approach has worked in testing, but it likely deserves deeper consideration before use
# in production deployments (operational risk, platform assumptions, and restart frequency).
networkHealer:
  enabled: true
  checkIntervalSeconds: 10