Control Layer Helm Chart

A standalone Helm chart for deploying the Doubleword control layer service.

Overview

This chart deploys the control layer service, which includes:

• Deployment with configurable replicas
• Service for internal cluster communication
• ConfigMap for application configuration
• Secret for database credentials
• Optional ServiceMonitor for Prometheus metrics

Installation

Prerequisites

• Kubernetes 1.19+
• Helm 3.0+

Install from OCI Registry

helm install my-control-layer oci://ghcr.io/doublewordai/charts/control-layer

You can also provide custom values:

helm install my-control-layer oci://ghcr.io/doublewordai/charts/control-layer -f custom-values.yaml

Configuration

Externally managed runtime Secret

By default the chart creates <release>-control-layer-secret from secrets.controlLayer.data. To use a Secret managed by another controller, configure its name instead:

secrets:
  controlLayer:
    existingSecret: externally-managed-runtime

The named Secret must exist in the release namespace and contain DATABASE_URL plus any other runtime keys required by the deployment. The chart does not render or mutate a Secret when existingSecret is set, and both the application and Fusillade workloads consume the same name.

For a small credential that rotates independently, keep the chart-owned Secret and load one or more overlays after it:

secrets:
  controlLayer:
    extraExistingSecrets:
      - rotated-database

Kubernetes resolves duplicate envFrom keys from the later Secret, so overlays should contain only the keys they intentionally replace.

The following table lists the configurable parameters and their default values. See values.yaml for all available options.

Core Configuration

Parameter	Description	Default
replicaCount	Number of replicas	1
image.repository	Container image repository	ghcr.io/doublewordai/control-layer
image.tag	Container image tag	Chart appVersion
image.pullPolicy	Image pull policy	IfNotPresent
imagePullSecrets	Image pull secrets	[]

Service Configuration

Parameter	Description	Default
service.type	Kubernetes service type	ClusterIP
service.port	Service port	3001

Service Account

Parameter	Description	Default
serviceAccountName	Name of existing service account to use	""

If you want to use a specific service account, set serviceAccountName. Otherwise, pods will use the default service account.

Database Configuration

Parameter	Description	Default
secrets.controlLayer.create	Create secret for database credentials	true
secrets.controlLayer.name	Name of existing secret (if not creating)	""
secrets.controlLayer.data.DATABASE_URL	Database connection string	"" (auto-generated if postgresql.enabled)
postgresql.enabled	Whether PostgreSQL is enabled in parent chart	true
secrets.postgres.data.POSTGRES_DB	PostgreSQL database name	clay
secrets.postgres.data.POSTGRES_USER	PostgreSQL user	clay
secrets.postgres.data.POSTGRES_PASSWORD	PostgreSQL password	clay_password

The chart supports both internal and external PostgreSQL databases:

• Internal PostgreSQL: If postgresql.enabled: true, the DATABASE_URL will be auto-generated using the postgres secret values
• External PostgreSQL: Set secrets.controlLayer.data.DATABASE_URL to your external connection string and postgresql.enabled: false

Monitoring

Parameter	Description	Default
serviceMonitor.enabled	Enable Prometheus ServiceMonitor	false
serviceMonitor.path	Metrics endpoint path	/metrics
serviceMonitor.interval	Scrape interval	30s
serviceMonitor.scrapeTimeout	Scrape timeout	10s
serviceMonitor.labels	Additional labels for ServiceMonitor	{}

Application Configuration

Parameter	Description	Default
config	Control layer application configuration (YAML)	See values.yaml
env	Additional environment variables	{}

Resources and Probes

Parameter	Description	Default
resources	CPU/Memory resource requests/limits	{}
livenessProbe	Liveness probe configuration	HTTP GET /healthz
readinessProbe	Readiness probe configuration	HTTP GET /healthz

Pod Configuration

Parameter	Description	Default
podAnnotations	Annotations to add to pods	{}
podLabels	Labels to add to pods	{}
podSecurityContext	Pod security context	{}
securityContext	Container security context	{}
nodeSelector	Node selector	{}
tolerations	Tolerations	[]
affinity	Affinity rules	{}
volumes	Additional volumes	[]
volumeMounts	Additional volume mounts	[]

Fusillade Daemon Configuration

The fusillade daemon handles background batch processing tasks. By default, it runs within the control layer pods based on leader election. You can optionally deploy it as a separate deployment for better resource isolation and independent scaling.

Parameter	Description	Default
fusillade.enabled	Deploy fusillade as a separate deployment	false
fusillade.replicaCount	Number of fusillade replicas	1
fusillade.mode	Optional daemon mode for the standard fusillade deployment (both, request_only, batch_only); empty uses the application default	""
fusillade.split.enabled	Render separate request-only and batch-only daemon deployments	false
fusillade.split.request.enabled	Render the request-only daemon deployment	true
fusillade.split.request.replicaCount	Number of request daemon replicas	inherits fusillade.replicaCount
fusillade.split.request.resources	CPU/Memory requests/limits for request daemon pods	inherits fusillade.resources
fusillade.split.request.env	Additional environment variables for request daemon pods	merged after env and fusillade.env
fusillade.split.request.database	Database pool overrides for request daemon pods	merged over fusillade.database
fusillade.split.batch.enabled	Render the batch-only daemon deployment	true
fusillade.split.batch.replicaCount	Number of batch daemon replicas	inherits fusillade.replicaCount
fusillade.split.batch.resources	CPU/Memory requests/limits for batch daemon pods	inherits fusillade.resources
fusillade.split.batch.env	Additional environment variables for batch daemon pods	merged after env and fusillade.env
fusillade.split.batch.database	Database pool overrides for batch daemon pods	merged over fusillade.database
fusillade.image.repository	Override image repository	(uses main image)
fusillade.image.tag	Override image tag	(uses main image)
fusillade.resources	CPU/Memory resource requests/limits	{}
fusillade.podAnnotations	Annotations for fusillade pods	{}
fusillade.podLabels	Labels for fusillade pods	{}
fusillade.nodeSelector	Node selector for fusillade pods	{}
fusillade.tolerations	Tolerations for fusillade pods	[]
fusillade.affinity	Affinity rules for fusillade pods	{}
fusillade.env	Additional environment variables	{}

When fusillade.enabled: true:

• The control layer pods will have background_services.batch_daemon.enabled set to never
• The fusillade pods will have background_services.batch_daemon.enabled set to always
• The standard fusillade deployment uses the application's default daemon mode unless fusillade.mode is set
• With fusillade.split.enabled: true, request pods use mode=request_only and batch pods use mode=batch_only

Keystore Persistence with a Managed StorageClass

The internal keystore can create and use a dedicated StorageClass for its Redis PVC. On GKE, set disk-encryption-kms-key to provision the backing Persistent Disk with a customer-managed Cloud KMS key:

keystore:
  enabled: true
  persistence:
    managedStorageClass:
      enabled: true
      parameters:
        type: pd-balanced
        disk-encryption-kms-key: projects/PROJECT_ID/locations/REGION/keyRings/KEY_RING/cryptoKeys/KEY

Existing PVCs keep the StorageClass they were created with. Recreate or migrate the keystore volume to move an existing install onto the managed StorageClass.

Example Configurations

With External Database

# custom-values.yaml
replicaCount: 2

image:
  tag: "v1.2.3"

postgresql:
  enabled: false

secrets:
  controlLayer:
    create: true
    data:
      DATABASE_URL: "postgres://user:[email protected]:5432/controldb"

resources:
  limits:
    cpu: 500m
    memory: 512Mi
  requests:
    cpu: 250m
    memory: 256Mi

With Prometheus Monitoring

# monitoring-values.yaml
serviceMonitor:
  enabled: true
  interval: 15s
  labels:
    prometheus: kube-prometheus

With Separate Fusillade Deployment

# fusillade-values.yaml
replicaCount: 3

fusillade:
  enabled: true
  replicaCount: 2
  resources:
    limits:
      cpu: 1000m
      memory: 1Gi
    requests:
      cpu: 500m
      memory: 512Mi

With Split Fusillade Daemons

# split-fusillade-values.yaml
fusillade:
  enabled: true
  split:
    enabled: true
    request:
      replicaCount: 4
      resources:
        requests:
          cpu: 500m
          memory: 512Mi
    batch:
      replicaCount: 1
      resources:
        requests:
          cpu: 1000m
          memory: 1Gi
      database:
        fusillade_pool:
          max_connections: 80

License

This project is licensed under the Apache License 2.0 - see the LICENSE.md file for details.