GitHub - benmarten/dokku-multideploy: Deploy multiple applications to a single Dokku server with centralized configuration. Import existing apps from a server, migrate between servers, or set up fresh deployments.

Deploy multiple applications to a single Dokku server with centralized configuration. Import existing apps from a server, migrate between servers, or set up fresh deployments.

Features

• Multi-app orchestration - Deploy multiple independent repos from one config
• Hierarchical configuration - Parent settings cascade to deployments, child overrides parent
• Smart deployment - Skips unchanged apps (compares git commits)
• Secrets management - Hierarchical .env files, gitignored
• Pre/post deploy hooks - Run scripts before/after deployment
• Tag-based filtering - Deploy subsets by tag (--tag staging, --tag api)
• PostgreSQL auto-setup - Opt-in automatic database provisioning
• Let's Encrypt SSL - Opt-in automatic SSL certificate provisioning
• MySQL service exposure - Optional mysql:expose automation for local-only DB access
• Storage mounts, ports, domains - Full Dokku configuration support
• Server import/migration - Import all apps from existing server, migrate to new server
• Backup & Restore - Backup/restore PostgreSQL databases and storage mounts with xz compression

Prerequisites

Configure your Dokku server in ~/.ssh/config:

Host <ssh-alias>
  HostName <your-server-ip>
  User root
  IdentityFile ~/.ssh/<your-key>
  IdentitiesOnly yes

Then use <ssh-alias> as the ssh_alias in your config.json.

Global Install

You can run this from any project by symlinking it into your PATH:

mkdir -p ~/bin
ln -sf /absolute/path/to/dokku-multideploy/deploy.sh ~/bin/deploy
chmod +x /absolute/path/to/dokku-multideploy/deploy.sh

Then use it inside any app folder that has config.json:

deploy --dry-run
deploy --sync

Notes:

• If you use fish shell and deploy is not found, add ~/bin to fish paths: set -Ux fish_user_paths ~/bin $fish_user_paths Then restart your shell and verify with: command -v deploy
• By default, deploy looks for config.json next to the invoked script/symlink, then falls back to $PWD/config.json.
• You can always override explicitly: CONFIG_FILE=$PWD/config.json deploy --dry-run

Quick Start

Migrate existing Dokku server (most common)

# 1. Clone this repo
git clone https://github.com/benmarten/dokku-multideploy.git

# 2. Import all apps from your existing server to a separate directory
./dokku-multideploy/deploy.sh --import ./apps --ssh <ssh-alias>

# 3. Backup databases and storage mounts
cd apps
ln -s ../dokku-multideploy/deploy.sh .
./deploy.sh --backup

# 4. Update config.json with new server details
#    Change ssh_host and ssh_alias to new server

# 5. Deploy everything to new server
./deploy.sh --dry-run  # Preview first
./deploy.sh

# 6. Restore backups on new server
SSH_HOST=<new-server> ./restore.sh backups/<timestamp>

Fresh setup

# 1. Clone and set up your project
git clone https://github.com/benmarten/dokku-multideploy.git
cd <your-project>
ln -s <path-to>/dokku-multideploy/deploy.sh .
cp <path-to>/dokku-multideploy/config.example.json config.json
# Edit config.json with your apps

# 2. Add secrets (optional)
mkdir -p .env
echo "DATABASE_PASSWORD=secret" > .env/api.example.com

# 3. Deploy!
./deploy.sh

Import from Existing Server

Already have apps running on a Dokku server? Import everything:

# Import all apps from your Dokku server
./deploy.sh --import ./apps --ssh <ssh-alias>

# Import without secrets (env vars)
./deploy.sh --import ./apps --ssh <ssh-alias> --no-secrets

This will:

1. Clone all app git repos to ./apps/<domain>/
2. Generate config.json with settings (domains, ports, storage, postgres, letsencrypt)
3. Split env vars on import: sensitive keys go to .env/, non-sensitive keys go to deployment env_vars in config.json

Then symlink deploy.sh and you're ready:

cd ./apps
ln -s <path-to>/dokku-multideploy/deploy.sh .
./deploy.sh --dry-run

Server Migration

Migrate all apps to a new server:

# 1. Import from current server (if not already done)
./deploy.sh --import ./apps --ssh <old-server>

# 2. Set up new server with Dokku
ssh <new-server> "wget -NP . https://dokku.com/install/v0.34.4/bootstrap.sh && sudo bash bootstrap.sh"

# 3. Update SSH config for new server
#    Edit ~/.ssh/config to add <new-server> alias

# 4. Update config.json
#    Change ssh_host and ssh_alias to new server

# 5. Deploy everything to new server
./deploy.sh

The script will create all apps, configure domains, env vars, storage mounts, ports, postgres, and letsencrypt on the new server.

Directory Structure

your-project/
├── deploy.sh              # Symlink to dokku-multideploy/deploy.sh
├── restore.sh             # Symlink to dokku-multideploy/restore.sh
├── config.json              # Your deployment configuration
├── .env/                    # Secret environment variables (gitignored)
│   ├── _api                 # Shared secrets for all "api" source_dir apps
│   ├── api.example.com      # Secrets specific to api.example.com
│   └── api-staging.example.com
├── certs/                   # Custom SSL certificates (optional)
│   └── api-example-com/
│       ├── server.crt
│       └── server.key
├── backups/                 # Backup files (gitignored)
│   └── 2026-01-06_143022/   # Timestamped backup folder
│       ├── api-example-com-db.dump.xz
│       └── api-example-com-storage-1.tar.xz
├── api/                     # Your API source code
│   ├── Dockerfile
│   ├── pre-deploy.sh        # Runs before deploy (e.g., migrations)
│   └── post-deploy.sh       # Runs after deploy (e.g., seed data)
└── web/                     # Your web app source code
    └── Dockerfile

Configuration

config.json

{
  "ssh_host": "dokku@<your-server-ip>",
  "ssh_alias": "<ssh-alias>",
  "global_domain": "example.com",
  "letsencrypt_email": "admin@example.com",
  "mysql_expose": {
    "mysql-prod": "127.0.0.1:3306",
    "mysql-staging": "127.0.0.1:3307",
    "mysql-test": "127.0.0.1:3308",
    "mysql-test-v2": "127.0.0.1:3309"
  },

  "api": {
    "source_dir": "api",
    "subtree_prefix": "services/api",
    "branch": "main",
    "builder": "dockerfile",
    "postgres": true,
    "letsencrypt": true,
    "env_vars": {
      "NODE_ENV": "production"
    },
    "deployments": {
      "api.example.com": {
        "tags": ["production", "api"],
        "env_vars": {
          "LOG_LEVEL": "warn"
        }
      },
      "api-staging.example.com": {
        "tags": ["staging", "api"],
        "env_vars": {
          "LOG_LEVEL": "debug"
        }
      }
    }
  }
}

Configuration Options

Root Level

Parent Level (e.g., "api", "web")

Deployment Level

Same options as parent level, plus:

Child settings override parent settings.

Dokku Settings Map

Use dokku_settings when you want config-driven dokku <plugin>:set behavior:

{
  "dokku_settings": {
    "nginx": {
      "client-max-body-size": "100m"
    }
  }
}

Notes:

• Parent and deployment dokku_settings are merged (deployment overrides parent per key).
• Keys are applied as dokku <plugin>:set <app> <key> <value> during deploy and --config-only.
• --sync now compares dokku_settings as part of drift detection.
• Import support currently captures nginx.client-max-body-size only; other plugin settings are still deploy-only unless you add custom import logic.

MySQL Service Exposure

Use root-level mysql_expose to declare local-only binds for Dokku MySQL services:

{
  "mysql_expose": {
    "mysql-prod": "127.0.0.1:3306",
    "mysql-staging": "127.0.0.1:3307"
  }
}

Behavior:

• Applied after deployment/config updates in full (unfiltered) deploy/config runs.
• Skipped for filtered runs (--tag, explicit app names, or --no-prod) unless --force-mysql-expose is provided.
• Skips the service if the desired bind is already active.
• If a different bind exists, unexposes all current bindings and re-exposes with the configured address.
• Accepts binds in the form 127.0.0.1:<port> or 0.0.0.0:<port> where port is 1–65535.
• If a listed service does not exist, deploy continues with a warning.
• In --dry-run mode, prints the command that would run without executing it.

For local clients like DBeaver, prefer 127.0.0.1:<port> and connect through SSH tunnel:

ssh -N -L 13306:127.0.0.1:3306 -L 13307:127.0.0.1:3307 <ssh-alias>

Then connect to 127.0.0.1 using local ports (13306, 13307, ...). Local port numbers are arbitrary — choose values that do not conflict with services already bound on your workstation.

Dokku Network Creation

Use root-level dokku_networks to ensure attachable Docker networks exist:

{
  "dokku_networks": ["vt-prod-internal", "vt-nonprod-internal"]
}

Behavior:

• Applied before deployment/config updates.
• Idempotent: existing networks are skipped.
• In --dry-run, commands are printed but not executed.

Secrets (.env files)

Secrets are loaded hierarchically:

1. .env/_<source_dir> - Shared secrets for all apps with that source_dir
2. .env/<domain> - Domain-specific secrets (overrides shared)

# .env/_api (shared by all api deployments)
DATABASE_PASSWORD=shared-secret
API_KEY=common-key

# .env/api.example.com (production-specific)
DATABASE_PASSWORD=production-secret

Usage

# Deploy all apps
./deploy.sh

# Deploy specific app(s)
./deploy.sh api.example.com
./deploy.sh api.example.com www.example.com

# Deploy by tag
./deploy.sh --tag staging
./deploy.sh --tag api
./deploy.sh --tag staging --tag api  # OR logic

# Skip production
./deploy.sh --no-prod

# Apply mysql_expose even on filtered deploys
./deploy.sh --tag staging --force-mysql-expose

# Dry run (see what would happen)
./deploy.sh --dry-run

# Force deploy (even if no code changes)
./deploy.sh --force

# Update config only (no code deploy, just env vars + restart)
./deploy.sh --config-only api.example.com

# Skip confirmation prompts
./deploy.sh --yes

Key options:

• --tag <tag>: deploy only apps with matching tag(s)
• --no-prod: exclude apps tagged production
• --force-mysql-expose: apply root mysql_expose even when using filtered deploy selection

Sync Check

Compare local config.json against live Dokku state without deploying:

# Check all selected deployments
./deploy.sh --sync

# Check only a subset
./deploy.sh --sync --tag staging
./deploy.sh --sync api.example.com

# Re-import live state before checking
./deploy.sh --sync --refresh-sync

# Clear cache and re-import
./deploy.sh --sync --reset-sync

# Patch local config.json from live Dokku values
./deploy.sh --sync --sync-apply

# Preview patch changes without writing
./deploy.sh --sync --sync-apply --dry-run

--sync behavior:

1. Imports current Dokku app config to .sync-cache/ (no git clone, no env secret export)
2. Compares local vs remote by domain
3. Reports:
   • ✓ In sync
   • ✗ Missing on Dokku
   • ⚠ Drift with per-field differences
4. branch is ignored in drift detection (local branch is source selection; Dokku deploy target is standardized to master)

Cache options:

• --refresh-sync: refresh .sync-cache/config.json before comparing
• --reset-sync: clear the sync cache directory, then import fresh
• --sync-dir <dir>: use a custom cache directory
• --sync-apply: write drifted fields from Dokku into deployment-level entries in local config.json

Exit codes:

• 0: all selected deployments are in sync
• 1: drift/missing detected or sync check failed

Backup

Backup PostgreSQL databases and storage mounts to compressed .xz files:

# Backup all apps
./deploy.sh --backup

# Backup to custom directory
./deploy.sh --backup --backup-dir ~/dokku-backups

# Backup specific app
./deploy.sh --backup api.example.com

# Backup by tag
./deploy.sh --backup --tag production

# Dry run (see what would be backed up)
./deploy.sh --backup --dry-run

This creates timestamped backup folders:

./backups/2026-01-06_143022/
├── api-example-com-db.dump.xz       # PostgreSQL dump (pg_dump custom format)
├── api-example-com-storage-1.tar.xz # Storage mount #1 contents
└── api-example-com-storage-2.tar.xz # Storage mount #2 contents

Backups are saved to ./backups/<timestamp>/ by default (gitignored).

Storage backup notes:

• storage_mounts supports object entries like {"mount":"<host>:<container>","backup":false}.
• Mounts marked backup:false are skipped in backup mode and explicitly listed in CLI output.
• Mounts larger than 100MB are skipped by default and explicitly listed in CLI output.
• Override threshold with BACKUP_MAX_STORAGE_MB (0 disables size-based skipping).
• Backup mode prints a final consolidated "Manual rsync required" list across all apps.
• Use direct host-to-host rsync for skipped large volumes during migrations.

Restore

Restore PostgreSQL databases and storage mounts from a backup:

# Restore to default server (from config.json ssh_alias)
./restore.sh backups/2026-01-31_085506

# Restore to specific server
SSH_HOST=co2 ./restore.sh backups/2026-01-31_085506

# Dry run (see what would be restored)
./restore.sh backups/2026-01-31_085506 --dry-run

The restore script will:

1. Install postgres plugin if needed
2. Create databases if they don't exist, or import into existing ones
3. Extract storage archives to /var/lib/dokku/data/storage/<app>/
4. Fix permissions for Dokku

Workflow for server migration:

# 1. Deploy apps to new server (creates apps, empty DBs, storage mounts)
CONFIG_FILE=config-newserver.json ./deploy.sh

# 2. Restore data from backup
SSH_HOST=newserver ./restore.sh backups/2026-01-31_085506

Deploy Hooks

Create pre-deploy.sh or post-deploy.sh in your source directory:

# api/pre-deploy.sh
#!/bin/bash
echo "Running migrations..."
npm run db:migrate

# api/post-deploy.sh
#!/bin/bash
echo "Seeding database..."
npm run db:seed

The APP_NAME environment variable is available in hooks.

How It Works

1. Parse config - Reads config.json, merges parent/child settings
2. Filter - Applies tag filters and deployment selection
3. For each app:
   • Sync with git origin
   • Check if deployment needed (compare commits)
   • Create Dokku app if needed
   • Configure PostgreSQL if enabled
   • Set domains
   • Mount storage
   • Set port mappings
   • Load secrets from .env files
   • Set env vars and build args
   • Run pre-deploy hook
   • Git push to Dokku
   • Run post-deploy hook
   • Enable Let's Encrypt if configured

Requirements

• bash 4.0+
• jq - JSON processor (brew install jq or apt install jq)
• git
• ssh access to your Dokku server
• xz (for backup mode - usually pre-installed)

SSH Configuration

Add to ~/.ssh/config:

Host <ssh-alias>
    HostName <your-server-ip>
    User dokku
    IdentityFile ~/.ssh/<your-key>

Then set "ssh_alias": "<ssh-alias>" in config.json.

License

MIT