Skip to main content

Docker Deployment Guide

Launch, verify, and operate the CTMS platform using Docker Compose.

Prerequisites

Before proceeding, complete the following:

  1. System Requirements — Ensure your server meets the minimum specs
  2. Installation — Clone the repo, configure .env.production, and build the API gateway
  3. Initial Setup & Configuration — Provision Frappe with DocTypes, RBAC, and seed data

1. Docker Compose Profiles​

Available Profiles​

ProfileServices
(default/core)Caddy, KrakenD, Zynexa, Sublink, ODM API
analyticsCube.js, Cubestore, MCP Server
lakehouseLakehouse PostgreSQL, Ingester, dbt
observabilityOpenObserve, OTEL Collector
initctms-init (one-shot)
allEverything except linux-logs

Start Commands​

# Core only
docker compose --env-file .env.production up -d

# Core + Analytics + AI
docker compose --env-file .env.production --profile analytics up -d

# Full stack
docker compose --env-file .env.production \
  --profile analytics --profile lakehouse --profile observability up -d

Production Overrides (EC2 / IP-Based Access)​

Exposes all ports directly and sets RUNTIME_* URL overrides pointing to http://EC2_PUBLIC_IP:port:

docker compose -f docker-compose.yml -f docker-compose.prod.yml \
  --env-file .env.production --profile analytics --profile lakehouse up -d

2. EC2 Deploy Script​

scripts/deploy-ec2.sh automates SSH-based deployment to an AWS EC2 instance:

./scripts/deploy-ec2.sh test-ssh   # Test connection
./scripts/deploy-ec2.sh setup      # Install Docker on the instance
./scripts/deploy-ec2.sh deploy     # Deploy the full stack
./scripts/deploy-ec2.sh update     # Pull latest images + restart
./scripts/deploy-ec2.sh status     # Container status
./scripts/deploy-ec2.sh logs       # View logs
./scripts/deploy-ec2.sh ssh        # SSH into the instance

SSH key: scripts/ssh-keys/your-ec2-key.pem | Deploy path: /opt/ctms-deployment

For the full EC2 setup guide (security groups, DNS, HTTPS), see AWS EC2 Deployment.

3. Verification​

Health Endpoints​

ServiceURLOK
Caddyhttp://localhost:8888/health200
API Gatewayhttp://localhost:9080/__health200
Zynexahttp://localhost:3000/api/health200
Sublinkhttp://localhost:3001/health200
Cube.jshttp://localhost:4000/readyz200
MCP Serverhttp://localhost:8006/health200
ODM APIhttp://localhost:8001/health200
OpenObservehttp://localhost:5080/healthz200

Cube Query Smoke Test​

curl -s http://localhost:4000/cubejs-api/v1/load \
  -H "Content-Type: application/json" \
  -d '{"query":{"measures":["Studies.count"]}}' | jq .

Database Table Counts​

docker compose --env-file .env.production exec lakehouse-db \
  psql -U ctms_user -d ctms_dlh -c \
  "SELECT schemaname, count(*) FROM pg_tables WHERE schemaname IN ('bronze','silver','gold') GROUP BY 1 ORDER BY 1;"

4. Production App URLs​

Replace <EC2_PUBLIC_IP> with the value of EC2_PUBLIC_IP from your .env.production.

ServicePortURL
Zynexa (CTMS App)3000http://<EC2_PUBLIC_IP>:3000
Sublink3001http://<EC2_PUBLIC_IP>:3001
KrakenD API Gateway9080http://<EC2_PUBLIC_IP>:9080
Cube.js (Semantic Layer)4000http://<EC2_PUBLIC_IP>:4000
MCP Server (AI Agent)8006http://<EC2_PUBLIC_IP>:8006
ODM API8001http://<EC2_PUBLIC_IP>:8001
OpenObserve5080http://<EC2_PUBLIC_IP>:5080
Lakehouse DB5433<EC2_PUBLIC_IP>:5433

Credentials for each service come from the corresponding env vars in .env.production (CUBEJS_API_SECRET, OPENOBSERVE_ROOT_EMAIL / PASSWORD, etc.).

5. Data Pipeline: Ingester and dbt​

Sample Data Required

The data pipeline ingests data from Frappe Cloud into the analytics database. Before running it, ensure your Frappe instance has been populated with sample clinical data — e.g. Studies, Sites, Subjects, Practitioners, Vital Signs, Drug Prescriptions, etc. Without this data, the pipeline will produce empty tables and the dashboards/analytics will have nothing to display.

After verifying the platform is running, populate the analytics database by running the data pipeline.

Frappe Cloud --> Ingester (DLT) --> Lakehouse DB (PostgreSQL) <-- dbt
                Bronze layer       Bronze, Silver, Gold          197 tests

For detailed documentation on each stage, see Data Pipeline — Ingester and Data Pipeline — dbt.

5.1 Start Lakehouse DB and Run Pipeline​

# Start DB and wait for healthy
docker compose --env-file .env.production --profile lakehouse up -d lakehouse-db

# Run Ingester: Bronze (~46 tables)
docker compose --env-file .env.production --profile lakehouse run --rm lakehouse-ingester

# Run dbt: Silver + Gold (bronze ~63, silver ~7, gold ~28 tables)
docker compose --env-file .env.production --profile lakehouse run --rm lakehouse-dbt daily

The daily command runs: dbt deps, dbt build, dbt run --select elementary. Expected dbt output: PASS=197 WARN=5 ERROR=0

5.2 Remote Database Alternative​

TARGET_DB_HOST=your-db-host.example.com
TARGET_DB_PORT=5432
TARGET_DB_SSLMODE=require

5.3 Schedule with Cron​

0 2 * * * cd /opt/ctms-deployment && \
  docker compose --env-file .env.production --profile lakehouse run --rm lakehouse-ingester && \
  docker compose --env-file .env.production --profile lakehouse run --rm lakehouse-dbt daily \
  >> /var/log/ctms-pipeline.log 2>&1

6. External / Vendor Service Dashboards​

Supabase (Auth Provider)​

Supabase can run as a cloud instance or self-hosted via the supabase/ folder in ctms.devops:

DeploymentDashboard URLSUPABASE_URL
Cloudhttps://supabase.com/dashboard/project/<ref>https://<ref>.supabase.co
Self-hostedhttp://localhost:8000 (Studio)http://localhost:8000
FieldEnv Var
Project URLSUPABASE_URL
Anon KeySUPABASE_ANON_KEY
Service KeySUPABASE_KEY
DB Credentialsctms_user / ctms_pwd (self-hosted)

Self-hosted Supabase ships with CTMS-specific init scripts: profiles table, handle_new_user() trigger, devices, medication_consumption_logs, notification_logs, and get_medication_status(). These are applied by the ctms-supabase-seed container — see Self-Hosted Vendor Stacks for the full setup guide.

Frappe (Clinical Data Backend)​

Frappe can run as Frappe Cloud or self-hosted via the frappe-marley-health/ folder in ctms.devops. For self-hosted Frappe, the setup service automates wizard completion, admin user creation, and API token generation — see Self-Hosted Vendor Stacks.

DeploymentDashboard URLFRAPPE_URL
CloudFrappe Cloud dashboardhttps://<site>.frappe.cloud
Self-hostedhttp://localhost:8080/apphttp://localhost:8080
FieldEnv Var
Site URLFRAPPE_URL
API TokenFRAPPE_API_TOKEN
Dashboard<FRAPPE_URL>/app

To regenerate the API token on a self-hosted instance:

docker exec -w /home/frappe/frappe-bench frappe-marley-health-backend-1 \
  bash -c 'source env/bin/activate && python3 /setup/frappe-generate-token.py'

OpenAI (AI / MCP Server)​

FieldEnv Var
API KeyOPENAI_API_KEY
ModelOPENAI_MODEL

All values are configured in .env.production. See Installation for the full list.

7. Docker Network Architecture​

All three stacks — Supabase, Frappe, and CTMS core — share a single Docker bridge network called ctms-network. This allows every container to discover and communicate with any other container by name, without port mapping.

Shared Network: ctms-network​

┌─────────────────────────── ctms-network (bridge) ──────────────────────────────┐
│                                                                                │
│  ┌─ Supabase ──────────┐  ┌─ Frappe ──────────┐  ┌─ CTMS Core ─────────────┐  │
│  │ supabase-kong :8000  │  │ frontend :8080    │  │ caddy, api-gateway      │  │
│  │ supabase-db   :5432  │  │ backend  :8000    │  │ zynexa, sublink         │  │
│  │ supabase-auth :9999  │  │ db (MariaDB)      │  │ cube, mcp-server        │  │
│  │ supabase-rest :3000  │  │ redis-cache/queue │  │ lakehouse-db, odm-api   │  │
│  │ supabase-realtime    │  │ websocket         │  │ otel, openobserve       │  │
│  │ supabase-storage     │  │ scheduler         │  │                         │  │
│  │ supabase-studio      │  │ queue-short/long  │  │                         │  │
│  │ ...13 services       │  │ ...11 services    │  │ ...16+ services         │  │
│  └──────────────────────┘  └───────────────────┘  └─────────────────────────┘  │
│                                                                                │
└────────────────────────────────────────────────────────────────────────────────┘

How It Works​

The network is defined as name: ctms-network with driver: bridge in all three docker-compose.yml files. Whichever stack starts first creates the network; the remaining stacks reuse it.

Recommended startup order:

# 1. Supabase (auth provider – needed by CTMS apps)
cd supabase && make up

# 2. Frappe (clinical data backend – needed by CTMS apps)
cd frappe-marley-health && make up

# 3. CTMS core services
docker compose --env-file .env.production --profile all up -d

Cross-Stack Container Discovery​

Because all containers share one network, they can reach each other directly:

FromToInternal URL
Zynexa / KrakenDSupabase APIhttp://supabase-kong:8000
Zynexa / KrakenDFrappehttp://frontend:8080
Supabase Edge FunctionsFrappe APIhttp://frontend:8080
IngesterFrappe (via Caddy)https://api.localhost
Any containerSupabase DBpostgresql://ctms_user:ctms_pwd@supabase-db:5432/postgres
Any containerLakehouse DBpostgresql://ctms_user:ctms_pwd@lakehouse-db:5432/ctms_dlh

Network Commands​

# Inspect the shared network
docker network inspect ctms-network

# List containers on the network
docker network inspect ctms-network --format '{{range .Containers}}{{.Name}} {{end}}'

# Create the network manually (not normally needed)
docker network create --driver bridge ctms-network

# Check if the network exists
docker network ls | grep ctms-network

Stopping Stacks​

When stopping, note that the last stack to stop removes the shared network. If another stack is still running, docker compose down will skip network removal automatically.

# Stop in reverse order (CTMS → Frappe → Supabase)
docker compose --env-file .env.production down
cd frappe-marley-health && make down
cd supabase && make down

8. Database Connection (Lakehouse)​

FieldEnv VarDefault
Host (external)EC2_PUBLIC_IP or localhost—
Host (Docker internal)—lakehouse-db
Port (external)LAKEHOUSE_DB_PORT5433
Port (internal)—5432
DatabaseTARGET_DB_NAMEctms_dlh
UserTARGET_DB_USERctms_user
PasswordTARGET_DB_PASSWORD(set in .env.production)

Connection string pattern:

postgresql://<TARGET_DB_USER>:<TARGET_DB_PASSWORD>@<HOST>:<PORT>/<TARGET_DB_NAME>
# From your machine (EC2)
psql -h <EC2_PUBLIC_IP> -p 5433 -U <TARGET_DB_USER> -d <TARGET_DB_NAME>

# From your machine (local Docker)
psql -h localhost -p 5433 -U <TARGET_DB_USER> -d <TARGET_DB_NAME>

# From inside Docker network
psql -h lakehouse-db -p 5432 -U <TARGET_DB_USER> -d <TARGET_DB_NAME>

9. Local Development URLs​

Add to /etc/hosts:

127.0.0.1 api.localhost zynexa.localhost sublink.localhost cube.localhost mcp.localhost odm.localhost observe.localhost
ServiceURL
Zynexahttps://zynexa.localhost
Sublinkhttps://sublink.localhost
API Gatewayhttps://api.localhost
Cube.jshttps://cube.localhost
MCP Serverhttps://mcp.localhost
ODM APIhttps://odm.localhost
OpenObservehttps://observe.localhost
Lakehouse DBlocalhost:5433

10. Operations​

# Logs
docker compose --env-file .env.production logs -f cube
docker compose --env-file .env.production logs --tail 100 zynexa

# Restart a service
docker compose --env-file .env.production restart cube

# Stop all
docker compose --env-file .env.production --profile analytics --profile lakehouse down

# Update (pull + recreate)
docker compose --env-file .env.production --profile analytics --profile lakehouse pull
docker compose --env-file .env.production --profile analytics --profile lakehouse up -d

# Pull latest KrakenD image + restart
docker compose --env-file .env.production pull api-gateway && \
docker compose --env-file .env.production up -d api-gateway

# Clean restart (WARNING: destroys data)
docker compose --env-file .env.production --profile analytics --profile lakehouse down -v

11. Quick Start Checklist​

  1. Complete System Requirements, Installation, and Initial Setup
  2. Start platform: docker compose --env-file .env.production --profile analytics up -d
  3. Verify health endpoints (Section 3)
  4. Start lakehouse: docker compose --env-file .env.production --profile lakehouse up -d lakehouse-db
  5. Run ingester: docker compose --env-file .env.production --profile lakehouse run --rm lakehouse-ingester
  6. Run dbt: docker compose --env-file .env.production --profile lakehouse run --rm lakehouse-dbt daily
  7. Log in to Zynexa

12. Troubleshooting​

ProblemCauseFix
Cube unhealthyGold tables missingRun ingester + dbt (section 5)
MCP can't reach CubeWrong CUBE_API_URLMust be http://cube:4000/cubejs-api/v1
Cubestore crash (ARM)amd64-only imageSet CUBEJS_DEV_MODE=true, unset CUBEJS_CUBESTORE_HOST
Duplicate CORS headersCaddy + backend both add headerCaddy handle @options handles preflight; backends should not add own CORS
Ingester can't reach FrappeFRAPPE_BASE_URL unresolvableNeeds Caddy + KrakenD running; uses extra_hosts: api.localhost:host-gateway
Continue Wait on queriesCubestore unreachableStart cubestore or comment out CUBEJS_CUBESTORE_HOST

Architecture​

                     +----------------------+
                     |    Frappe Cloud       |
                     +----------+-----------+
              +-----------------+------------------+
              v                 v                  v
        +----------+    +------------+     +------------+
        |  Caddy   |    |  Ingester  |     | ctms-init  |
        |(Rev Proxy)|   |(DLT>Bronze)|     |(Provision) |
        +----+-----+    +-----+------+     +------------+
             |                 |
        +----+-----+   +------v-----+
        | KrakenD  |   |Lakehouse DB|<-- dbt (Bronze>Silver>Gold)
        |(API GW)  |   |(PostgreSQL)|
        +----+-----+   +------+-----+
             |                 |
  +------+---+---+      +-----+------+
  v      v       v      |  Cube.js   |-- Cubestore
Zynexa Sublink  ODM     |(Semantic)  |
                        +-----+------+
                              v
                        +------------+    +----------+
                        | MCP Server |    | Supabase |
                        |(AI Agent)  |    |  (Auth)  |
                        +------------+    +----------+