Download the PHP package bedezign/dchelper without Composer

Docker Compose Helper

(STILL IN DEVELOPMENT, use at your own risk)

Please note that the network related functionality in this helper (alias/proxy) only works on macOS. I'm really sorry about that. I would love to help Windows people as well, but fact is that it simply does not support most of the required functionality.

If you use docker compose and you want to be able to develop using a decent url for your project like http://myproject.test instead of http://127.0.0.1:32721 with nearly no extra setup, then read on!

This tool wraps around docker-compose and will be ran instead of it so it can add some exiting new functionality to your development environment.

Please note that this script will need sudo rights for some of the commands.

Installing

You can simply add it via packagist/composer:

After that you might want to setup an alias in your .bashrc file:

It's as easy as that!

Notes

• Currently dchelper assumes that there will be "one of each service". It is mainly for development purposes and until someone requests support for multiple containers per service I'm not going to think about the added complexity.
• Mapping IP & Hostname is currently for the first found IP only (with a preference for aliased instead of proxied).

What does it do?

In short:

• Provides you with a "dedicated IP" for your project (instead of 127.0.0.1)
• Get rid of having to specify ports
• Automatically create/maintain a hostname for your project. Use myproject.test in your browser instead of an IP!
• Easily "shell" into a container (docker-compose shell and you have a login-shell into the default container).
• Built-in helpers for docker-compose v3.4 (or higher) configurations (there's "envsubst" and "scriptrunner" at this moment, suggestions welcome)
• All of this via environment variables (.env) or docker-compose.yml settings.

Ports, ports, ports!

When publishing a port from a container, by default the docker VM maps this onto your localhost (127.0.0.1).

If you don't specify a remote port, you won't get port collisions but you'll end up with some obscure port like 32122. I don't know about you, but hosting my development site at http(s)://127.0.0.1:32122 just doesn't seem right. Even assigning a host to it doesn't make it any prettier when you always have to specify a port.

The alternative would be to map it to the default remote port, but then you'll probably end up with "port in use" etc.

Docker actually has its own solution for this and DCHelper adds a second one.

The premise is that we can add an extra virtual IP per docker-compose-project. That would mean that heavily used ports like 80 and 443 are always available on "your" IP.

There is already a solution built-in in your OS: Linux (and by extension macOS) allow you to add "virtual" IPs to your loopback network interface (aka IP alias). Both docker and dchelper can use this functionality to simplify your life a bit.

Remote IPs (docker)

Docker (and Docker Compose) define a port specification as follows: [[remote_ip:]remote_port[-remote_port]:]port[/protocol]

This means that, when you are adding a published port, you can also specify the IP on which you want the port to be added:

So the above tells docker to link service port 80 to port 80 on IP 172.99.0.6 (instead of 127.0.0.1). Pretty awesome no? There is a small gotcha however: docker-compose will not add the aliased IP to your network interface.

DCHelper will.

It interprets the docker-compose config in advance (as well as your .env if one was found) and determines what IPs need to be aliased for your configuration to work.

If it finds any that don't exist yet, it will register them for you before passing control to docker-compose.

Using this functionality is easy: The only modification needed to your configuration is what was shown in the example above.

More extensive example:

.env:

docker-compose.yml:

By using a configured value you don't have much work to update the IP later on. Since the variable will be replaced by docker-compose, DCHelper will still pick up the actual IPs.

COMPOSE_ALIAS_IP has a special meaning: DCHelper will always alias this IP, even if it wasn't used further in your configuration. The variable is also used by the hosts functionality.

This means slightly more in your configuration file, but docker takes care of the rest, no external utilities needed.

TCP Proxying/Tunneling (via socat)

This functionality uses socat (SOcket CAT), which is a pretty awesome tool that describes itself as a "Multipurpose relay". (It can be easily installed on a mac via HomeBrew)

For this variant your ports-configuration remains as is:

The trick here is that you either add a global COMPOSE_PROXY_IP environment variable, or specify a PROXY_IP in the containers' environment:

(Specifying PROXY_IP allows you to use different IPs in your setup. Use the COMPOSE_PROXY_IP globally if you only need the one IP.)

When parsing the configuration, DCHelper will detect all published ports and where they are linked to. It will then launch a bunch of socat instances to establish a tunnel between the proxy IP and the randomly assigned docker port on 127.0.0.1.

The net effect will be the same as the "Remote IP"-method, so it depends on what you prefer. The disadvantage here is that this only works if you detach from the containers when running the up command. The tunneling can only be done after your "up" command finalises (containers need to be running to detect the configuration), which doesn't happen without -d

This example yields an identical result:

.env:

docker-compose.yml:

Both port 3306 and 80 will be available via 172.99.0.1.

As mentioned above: It is possible to use different IPs per service, just use PROXY_IP instead.

Hostnames

By adding a COMPOSE_HOSTNAME to your environment, you tell DCHelper to check your /etc/hosts file and if needed, add a new entry to it. The IP used will be either COMPOSE_ALIAS_IP or, if that wasn't set, COMPOSE_PROXY_IP.

There is currently no support for a per service/container host, just the global one. (I'll add this if someone requests it, I currently have no need for it)

Shell

DCHelper adds a shell-command. By doing dchelper shell php for example, it will trigger a (login) shell for in the related container. This actually runs docker exec -it <container> <command> in the background, but it will allow you to specify the compose service name instead of having to figure out the docker container name.

Due to this and this (issues with the pseudo terminal size not being set correctly) the default command currently executed is actually '/bin/bash -c "export COLUMNS=`tput cols`; export LINES=`tput lines`; exec bash -l"'.

By using COMPOSE_SHELL_DEFAULT=service-name in your environment you can indicate what service to use if none was specified. This can also be done by adding SHELL_DEFAULT=1 to one of your service environment definitions.

If you have a terminal application that understands escape sequences (like iTerm2), DCHelper can also change the tab title for you. You can either specify a title per service in the services environment using SHELL_TITLE or specify a global format via COMPOSE_SHELL_TITLE. In this case {CONTAINER} will be replaced by the containers' name.

For example: COMPOSE_SHELL_TITLE="${COMPOSE_HOSTNAME: {CONTAINER}". The hostname replacement will be taken care of by docker compose, the {CONTAINER} will be replace by DCHelper.

Helpers

This only works if you use compose file format v3.4 or later. This is the first version that allows for (ignored) vendor-specific root entries (x-...)

Using helpers

Single helper

You can specify a single command to run:

Multiple helpers

If you want to run the same helper multiple times, you can add some "junk" to create different keys:

What is behind the dot is discarded, so use whatever you want.

Root

You can specify root as the helper name and it will set the 'base path' for every relative source/origin directory on the local system:

This will use the template from /generic/docker/folder/nginx.template and store it in the .docker/nginx.conf relative to where docker-compose.yml lives. It will also translate ~-entries for you, in the source, target and root directives.

If you have a global root set and want it back to the project directory for a specific section, just set it to empty in that section:

script.sh will be loaded from working directory rather than the /generic/docker/folder

Stages

Currently the helpers can run at 2 stages: pre.<command> and post.<command>. (as in: before and after the docker-compose up command). By specifying at in the configuration you can override this behavior. post.up can only run after the up command completes, so if you do not detach it will only run after you terminate the containers. Other possibilities are for example pre.shell etc.

EnvSubst

If your compose project needs configuration files with values based on your environment, the trick so far was splice in an envsubst call somewhere that takes care of this for you. DCHelper supports this natively and in a simple manner.

Important: The helper is written in a way that it will only replace whatever variables are known in the environment. This will not touch anything except that. So proxy_set_header Host $http_host; will be safe, as long as you don't have a http_host defined.

Note that this is internal functionality. You don't need to install the envsubst command for this to work.

To generate a configuration file, you just add an entry for envsubst:

This runs @ pre.up by default.

environment-file

Which files to load as environment files. The default (if not specified) .env in the working directory.

environment

The docker-container environments to load.

In the example it will use the .env file with the environment from the nginx service added on top.

files

List of files to do the replacement on. The source (or even target) files do not need to be in the project directory.

This allows you to create a number of templates and then generate a per-project config whenever you run docker-compose up.

By using the multiple commands syntax you can run envsubst multiple times if you want different environments.

By prepending the name of the output file with a service name the configuration will be created in the relevant container. This path has to be absolute. docker cp is used for this functionality.

Example for the above:

dchelper will make sure the target folder gets created for you if needed and then copies the file into it.

Result

The envsubst helper runs before anything else, so the generated files are available in all your services:

ScriptRunner

scriptrunner allows you to run shell scripts within the container. It doesn't matter if the script is mapped into the container or not, it executes it as a set of bash commands.

This runs @ post.up by default if nothing was specified.

An example:

Note that scriptrunner supports EnvSubst.

The functionality works both for the command lines (always enabled), so ${SCRIPT_FOLDER}/script.sh will work). Substituting in the scripts' content also enabled by default, but can be turned of by specifying envsubst: false. As with EnvSubst you can specify the environment-file and environment-key as well for more configuration.

service

What service/container to run against. If you want to run the script on the host itself, use localhost. The once behavior is currently not supported for scripts running on localhost, only for containers.

lock-file

This file is added to the container and used to remember which scripts already ran. Used by the "once" command. If not specified it will default to /.dchelper.scripts

once vs always

You can specify scripts under either once or always. I think the difference is self-explanatory.

direct

For scripts running on localhost (with always) you can specify direct: true. This forces dchelper to run the shell script as opposed to loading its content and executing that. Please note that envsubst on the file content will not work like this. Since it is running on localhost you can always just source the .env file on those scripts.

Full example

Below is an example of a full docker-compose.yml:

Part of .env:

What this does: This config will give you a working http://myproject.test that has a port 80 and 3306. It runs on a 3 container structure, all using unmodified images from the docker hub.

The root entry signifies that all relative source locations will be mapped in there, the target locations do not use this.
~ is translated for both target and source.

Note: if the root is also relative it will use the current working directory as its base (or home if ~ is used). It can be overwritten locally in a helper configuration.

We use 2 envsubst commands here:

• The nginx-one creates a locally stored file based on the configuration at the pre.up stage. The resulting file is mounted into the nginx container so that it can be used when it boots.
• The aws-one will create a file directly in the container. Since that functionality only works after the container is up, we use at: post.up in the configuration.

This also shows that you can use multiple envsubst entries just by appending something random (preferably something that makes sense to you ;) ).

The scriptrunner command runs some of my utility scripts to install things in the containers and it keeps track of what was installed within that container. once indicates that these will only be ran once, if not found in the lock-file yet.

Since these will be executed "passthrough", you'll see what is being done in your console without having to look in the container logs to see when things are wrapped up. (which is something that annoyed me personally when using an alternate entry script that installs things).

The APP_URL in the example was added to show that you can reuse the variables there as well. The goal here is to have to modify as little as possible to setup a new project.