Deploying Dependencies

RootCause.ai requires several infrastructure services to run. This guide covers deploying the complete dependency stack using Helm charts.

The dependencies chart deploys everything your RootCause.ai platform needs: databases, message queues, workflow orchestration, identity management, and LLM integration.

---

Prerequisites

Before deploying dependencies:

1. Kubernetes Cluster – A running cluster (v1.30.0+, AWS EKS, GKE, AKS, or similar)

2. kubectl – Configured to access your cluster

3. Helm 3.0+ – Installed and configured

4. Namespace – Decide on a namespace (e.g., rootcause)

# Verify cluster access
kubectl cluster-info

# Verify Helm
helm version

---

Services Deployed

The dependencies chart installs:

Service	Purpose	Version
PostgreSQL	Relational database for Temporal, FusionAuth, LiteLLM	15.0+
MongoDB	Document database for application data	8.0+
Redis	Sync layer	7.2.4
RabbitMQ	Message queue for async processing	3.12.2+
Temporal	Durable workflow execution	1.27.2
FusionAuth	Identity and access management	1.57.1
LiteLLM	LLM API proxy and management	1.77.5

---

Installation from OCI Registry

Step 1: Authenticate with GitLab Registry

helm registry login registry.gitlab.com \
  --username <your-gitlab-username> \
  --password <your-personal-access-token>

Step 2: Install Required Operators

MongoDB and RabbitMQ use Kubernetes operators:

# Add MongoDB Helm repository
helm repo add mongodb https://mongodb.github.io/helm-charts
helm repo update

# Install MongoDB Community Operator
helm install community-operator mongodb/community-operator \
  --namespace rootcause \
  --create-namespace

# Install RabbitMQ Cluster Operator
kubectl apply -f "https://github.com/rabbitmq/cluster-operator/releases/latest/download/cluster-operator.yml"

Step 3: Verify Operators

# Check MongoDB operator
kubectl get pods -n rootcause -l name=mongodb-kubernetes-operator

# Check RabbitMQ operator
kubectl get pods -n rabbitmq-system -l app.kubernetes.io/name=rabbitmq-cluster-operator

Step 4: Install Dependencies Chart

# Install latest version
helm install perceptura-platform \
  oci://registry.gitlab.com/perceptura/client-deployments/helm-charts-dependencies/helm/perceptura-dependencies \
  --namespace rootcause \
  --create-namespace

# Or install specific version
helm install perceptura-platform \
  oci://registry.gitlab.com/perceptura/client-deployments/helm-charts-dependencies/helm/perceptura-dependencies \
  --version 0.1.5 \
  --namespace rootcause \
  --create-namespace

---

Component Configuration

Each component can be customised via values.yaml:

# PostgreSQL authentication settings
postgres:
  auth:
    username: "postgres"
    password: ""
    existingSecret: ""
    database: "appdb"

# Redis authentication settings
redis:
  auth:
    enabled: true
    password: ""
    existingSecret: ""
    existingSecretKey: "redis-password"

# LiteLLM settings
litellm:
  secrets:
    masterKey: "" # it will be generated if not provided
    saltKey: "" # it will be generated if not provided

  # API Keys for LLM providers - set these values for your deployment
  apiKeys:
    openaiApiKey: "" # Set your OpenAI API key here
    anthropicApiKey: "" # Set your Anthropic API key here
    googleAiStudioApiKey: "" # Set your Google AI Studio API key here
    bedrockAwsAccessKey: "" # Set your AWS access key for Bedrock here
    bedrockAwsSecretKey: "" # Set your AWS secret key for Bedrock here

fusionauth:
  database:
    protocol: "postgresql"
    host: "postgres.rootcause.svc.cluster.local"
    port: 5432
    name: "fusionauth"
    tls: false
    tlsMode: require

    existingSecret: "perceptura-postgres-credentials"
    user: "postgres"
    root:
      user: "postgres"

  search:
    engine: "database"

---

Custom Values File

Create a values-production.yaml for your environment:

# Production values for dependencies

postgres:
  auth:
    postgresPassword: "${POSTGRES_PASSWORD}"
  primary:
    persistence:
      storageClass: "gp3"
      size: 100Gi
    resources:
      requests:
        cpu: 2
        memory: 4Gi
      limits:
        cpu: 4
        memory: 8Gi

mongodb:
  replicaSet:
    members: 3
  persistence:
    storageClass: "gp3"
    size: 200Gi

redis:
  master:
    resources:
      requests:
        cpu: 1
        memory: 2Gi

temporal:
  server:
    replicaCount: 3

Install with custom values:

helm install perceptura-dependencies \
  oci://registry.gitlab.com/perceptura/client-deployments/helm-charts-dependencies/helm/perceptura-dependencies \
  --namespace rootcause \
  -f values-production.yaml

---

Verifying Installation

Check all components are running:

# View all pods
kubectl get pods -n rootcause

# Check services
kubectl get svc -n rootcause

# Verify MongoDB replica set
kubectl get mongodbcommunity -n rootcause

# Verify RabbitMQ cluster
kubectl get rabbitmqclusters -n rootcause

Expected output shows all pods in Running state.

---

Persistence

All stateful services use Persistent Volume Claims:

Service	Default Size	Recommended Production
PostgreSQL	1Gi	10Gi+
MongoDB	50Gi	100Gi+
Redis	5Gi	10Gi
RabbitMQ	10Gi	25Gi

Storage Class

Use SSD-backed storage classes for best performance:

postgres:
  primary:
    persistence:
      storageClass: "gp3" # AWS EKS

---

Networking

Services communicate via internal Kubernetes DNS:

Service	Internal DNS
PostgreSQL	postgres-service.rootcause.svc.cluster.local
MongoDB	perceptura-mongo-svc.rootcause.svc.cluster.local
Redis	redis-master.rootcause.svc.cluster.local
RabbitMQ	perceptura-rabbitmq.rootcause.svc.cluster.local
Temporal	temporal-frontend.rootcause.svc.cluster.local

Uninstalling

# Remove dependencies chart
helm uninstall perceptura-platform -n rootcause

# Remove operators (if no longer needed)
helm uninstall community-operator -n rootcause
kubectl delete -f "https://github.com/rabbitmq/cluster-operator/releases/latest/download/cluster-operator.yml"

---

Next Steps

With dependencies running:

• Configure LiteLLM

• Deploy the Platform Components

• Configure Secrets and Environment

• Set up Scaling for production loads