Depler

Easily deploy your docker app to single remote host without dependencies. No registry. No swarm (actually, optional). No kubernetes. Only hardcore. Only single remote machine. What you only need: it is docker on both - local and remote machines.

Installation

0. Install docker

This package requires docker on both: local and remote machines. Just google for install docker %os%.

1. Install the package

Install as local package to link with ./node_modules/.bin.

npm i --save-dev depler

Install globally to use depler deploy from everywhere.

npm i --save-dev -g depler

2. Expose port 80 in your dockerfile

By default depler thinks that you expose port 80 in your dockerfile like below. In the future, we will improve this feature and will implement ability to bind to another port via argument. But now, please, expose this port in your dockerfile.

# run the app
CMD ["node", "server.js", "--port", "80"]
EXPOSE 80

3. [optional] Add scripts to package.json

Add script to your package.json because commands usually too long to type them every time in console.

{
  "scripts": {
    "deploy": "./node_modules/.bin/depler deploy ..."
  }
}

Examples

Copy-and-paste source code to remote host → build docker image remotely → run docker container remotely

Take a look at this command.

depler deploy --code my-nodejs-app --host web@192.168.1.22 --as source --run-as container .

Bro, deploy the app (deploy) for me with repository name my-nodejs-app (--code) to host 192.168.1.22 as user web (--host) using source strategy (--as source) and run it on port 8080 (default config).

The flag ---as source tells tool that we should copy source code to remote host → build image → run container on remote host.

Build docker image locally → transfer image to remote host → run docker container remotely

Almost the same command, but instead of --as source we should pass --as image flag.

depler deploy --code my-nodejs-app --host web@192.168.1.22 --as image --run-as container .

This flag tells tool that we should build image locally first → transfer it to remote host → run container on remote host.

Deploy your app to raspberry pi from something else

I run into some problems when I tried to deploy my app from OS X to Raspberry PI (both have different architectures). So, my workaround is easy: deploy only source code of the app and build image remotely from different docker image.

Take a look at this command. It says to the tool: bro, deploy the app (deploy) with repository name my-nodejs-app (--code) to host 192.168.1.22 as user web (--host) using source strategy (--as source) and build the image with arg FROM=arm64v8/node:10.16.1-buster-slim (--build-arg) and run the container on port 8080 (default config).

depler deploy --code my-nodejs-app --host web@192.168.1.22 --as source --run-as container --build-arg FROM=arm64v8/node:10.16.1-buster-slim .

My dockerfile for the app looks like below.

#!/bin/bash
ARG FROM=mhart/alpine-node:10.16
FROM ${FROM}
# ...

As you see, I have default value for base image name (mhart/alpine-node:10.16), but I override this value with --build-arg and set this arg to arm64v8/node:10.16.1-buster-slim on docker build stage.

Scenario

Deploy tells tool to run the image as a single detached container with te following scenario:

1. Stop all containers with matched name.
2. Run new container with the tag.

Commands

Deploy

Basically the tool has only one command: deploy. This command performs deploy to remote host. And it has a lot of options.

Options

--as source vs. --as image

Which to chose?

The first one (--as source) tells tool to use the next flow:

1. Copy-and-paste source code from local folder to remote host into /tmp/... folder.
2. Build docker image on remote host via ssh from source code from /tmp/... folder.
3. Run docker container on remote host.

The second one (--as image) tells tool to use the another flow:

1. Build docker image locally.
2. Archive image to .tar file and put it into local /tmp/... folder.
3. Transfer archive to remote host and put it into /tmp/... remote folder.
4. Load archive to docker from /tmp/... folder.
5. Run docker container on remote host.

I prefer to use the second scenario (--as image).

Other commands

All other commands from this tool are not usable as standalone commands. But, if something went wrong, you can just rerun failed command with logged arguments.

Overrides for commands

There is an ability to override some commands like docker run or docker build. You can pass custom arguments to this commands through .json settings file with --config argument. Section default will be used for all commands. Section run:container will be used to pass variables to docker run script.

A real configuration file to deploy yii2 api project.

{
  "commands": {
    "run": {
      "container": {
        "publish": "8013:80",
        "volume": "/data/ceater/uploads:/app/web/uploads",
        "network": "webulla-ceater",
        "env": [
          "MYSQL_HOST=webulla-ceater-database",
          "MYSQL_DATABASE=data",
          "MYSQL_ROOT_USER=USER",
          "MYSQL_ROOT_PASSWORD=PASSWORD"
        ]
      }
    }
  }
}

To be done

• Automatically run docker swarm init on remote host.
• Additional overrides for the commands via settings.json.