Binci is a utility that allows you to easily containerize your development workflow using Docker. Simply put, it's like having a cleanroom for all of your development processes which contain services (like databases) without needing to setup and maintain these environments manually.
FAQ: Why Binci over Docker-Compose?
The best way to install Binci and keep it updated is through NPM, included with Node.js, or Yarn.
yarn global add binci or npm install binci -g
Obvious Note: You need to have Docker installed as well.
Important Note: In order to run the tasks, Binci creates a temp file (binci.sh). The tool will do its best to determine the best location (usually /tmp), but this can be explicitly set by specifying the environment variable BINCI_TMP.
After you have Binci installed you can initialize a project by moving to the project directory and running the following:
binci init
The above will prompt you to enter a base image; this should be a valid Docker image.
Once the configuration is generated you can run tasks. The default template includes several, for example:
binci env
The above will load your project via Binci & Docker, then echo the environment variables available.
Binci is controlled by a binci.yml file in the root of your project. A basic example is shown below:
from: node:6
services:
- mongo:
from: mongo:3.0
env:
- DB_ROOT_PASSWORD=foo
expose:
- 27017:27017
env:
- TMP=${TMP}
- HOST=${HOST:-localhost}
expose:
- 8080:8080
volumes:
- ${HOME}/.ssh:/root/.ssh
user: nobody
hosts:
- google.com:127.0.0.1
before: npm install
after: echo "done"
tasks:
env: env | sort
start: node index.js
lint: npm run lint
test: npm test
run: node index.jsThe above can then be executed via the binci <task> command from within the same directory as your project and binci.yml. For example, binci run would perform the following:
- Pull and start
mongowithDB_ROOT_PASSWORDenvironment variable and port27017exposed - Sets the following on the container:
- Set the primary container environment variable
TMPto the same as the host machine - Expose port
8080to the host system - Mount the host machine's
.sshdirectory in the container - Set a host entry for
google.comto127.0.0.1
- Set the primary container environment variable
- Run
npm installinside the container before running the task - Run
node index.jstask inside the container - Echo
doneafter the task has completed
Binci also allows for executing tasks not predefined in the configuration file using the -e flag. For example:
binci -e "/bin/sh"
The above would start the container using the configuration, call the before task, then start the sh shell. The container will then remain in the shell until an exit command is sent by the user.
The dockerfile configuration property can be specified to point to this project's Dockerfile, which will be auto-built for task execution. This image will be rebuilt any time the Dockerfile is edited. Defaults to ./Dockerfile.
The from configuration property causes Binci to use the specified image to run tasks, rather than building a new image from a local Dockerfile.
For testing different images easily, the either the -b <build-dockerfile> or -f <from-alternate-image> arguments can be passed on execution.
If a dockerfile is being used rather than the from property, tags can be specified to define one or more tags to apply to any newly built container in addition to the default Binci tag. They can be specified in any format accepted by the docker build command's -t flag, such as repo/imageName:version.
dockerfile: ./Dockerfile
tags:
- myorg/myrepo:latest
- myorg/myrepo:5.1.0The above binci.yml will, in the event that Binci needs to build a new container, tag the build with Binci's own tag in
addition to the two listed tags. When Binci is done running, the command docker push myorg/myrepo:latest would work
as expected.
When a dockerfile is used, Binci will automatically rebuild it if the dockerfile changes. However, it may be necessary to trigger a rebuild when other files change that may impact the build. To list these files, specify rebuildOnChange:
dockerfile: ./Dockerfile
rebuildOnChange:
- ./Gemfile
- ./Gemfile.lock
- ./RakefileWhen any of those files are changed, the next Binci run will rebuild the image from the dockerfile before executing the task.
Services add links into the primary container, exposing the services for utilization. For the most part, services utilize the same format for definition as the primary container.
During execution, service containers are named in 2 ways:
- Ephemeral (non-persisted):
bc_<NAME>_<INSTANCE-ID> - Persisted:
<NAME>
The above naming convention allows for persisted services to be shared with other Binci instances, or manually run docker containers, via the --link argument.
At startup Binci will ensure any persisted or already running containers are not started again.
After completion, Binci will run a detached process which will execute docker stop and docker rm on any non-persisted, ephemeral services.
Services which need to persist between runs can be set by providing persist: true in their configurations.
Persisted services will not stop after the primary container finishes its task and can be used by the same project, other projects, or independently.
By default, all services in the configuration will be linked on any run. To disable services for specific tasks, you can define them like this:
tasks:
lint:
disable:
- mongo
cmd: npm run lint
start: npm startAlternatively, you can disable all services for a task with '*':
tasks:
lint:
disable: "*"
cmd: npm run lint
start: npm startFor one-off cases, individual services can also be disabled via the command line:
binci lint -d mongo
or all services:
binci lint -d '*'
binci lint --disable-all
Binci will automatically stop services after any run (success or fail). However, if this fails or some other fringe-case causes this process to stop responding the system can leave orphaned containers running.
In order to mitigate this issue Binci will run a check for any bc_ prefixed containers on each run. If orphaned services are identified a warning message will appear at the beginning of the process to indicate the orphaned service(s) and commands to remedy/exit these containers.
The following commands can be run to cleanup any running containers:
Stop and Remove Binci Containers:
binci --cleanup
Stop and Remove ALL Containers:
binci --cleanup-all
Setting env array items will expose environment variables in the primary instance or services. These entries can be raw strings or use ${VAR} notation, where VAR is an environment variable on the host machine to use. Entries should use the format <ENV_VAR>=<VALUE>.
You can provide default values for environment variables by using the standard bash syntax. For example, VAR=${FOO:-foobar} will first look for the FOO environment variable, and if it is not defined VAR will be set to the string foobar (i.e. VAR=foobar).
Setting expose array items will expose ports to the host machine from the primary or service containers. Entries should use the format <CONTAINER_PORT>:<HOST_PORT>
Setting volumes will mount volumes on the host machine to designated paths on the primary or service containers. Entries should use the format <HOST_PATH>:<CONTAINER_PATH>
The current working directory will automatically mount to the same path on the container instance by default. To change its mount point and the working directory, specify the workDir parameter at the top level.
Setting user will instruct the container to run under the username or UID specified.
Setting hosts will update the hosts configuration for the container. Entries should use the format <HOST_NAME>:<ADDRESS>
By default binci uses Docker's --privileged flag to run, you can disable by setting this in the config or passing --privileged=false in the binci command.
The standard procedure for stopping a Docker container is the stop command which sends SIGTERM and allows a grace period (default: 10) for the container to exit on its own.
Some containers may not exit via SIGTERM (or may hang). In this case, the service container can utilize the stopTimeSecs property:
services:
- mongo:
from: mongo:3.0
stopTimeSecs: 3The stopTimeSecs above would forcibly stop the container after 3 seconds using Docker's stop command's -t option.
Global Setting:
In addition to setting the stopTimeSecs per service, this property can be set in the root of the binci.yml configuration and will be applied to any services that don't have an explicit stopTimeSecs property.
In the event that a service's default command needs to be overridden, the command property can do that. Specify the command with a string or array of strings.
Binci can be run via yarn/npm scripts, but is also setup to run development tasks using Binci.
Ensure you have the latest version installed then run:
binci install test or yarn install && yarn test.
To run end-to-end tests run yarn run e2e. This works by fully emulating a run inside the /test/project directory and executing /test/system/run.js with the /test/system/tests.json definitions file.
To test binary builds:
1. Build Binary:
yarn run build:linux
2. Run (Ubuntu) Docker in Docker:
docker run -it --rm -v /var/run/docker.sock:/var/run/docker.sock -v $PWD:/app -w /app ubuntu sh -c "apt-get update && apt-get install docker.io -y && bash"
3. Create Binci Alias:
alias binci=$PWD/bin/linux/binci
Once the above steps are completed the binci executable will be available.
First off, we like Docker Compose, and definitely think it's a powerful tool. However, Binci was built because Compose is more about long-running, containerized environment and what we set out to build was a way to run ephemeral, limited-lifespan tasks without having to manage cleanup between each run.
Compose takes the approach of spinning up containers that run, almost like a virtual machine, while you need them. Binci looks at things from a point of view of abstracting docker run command chains to create a single-run instance only for that task, then shutting down and doing cleanup so each run is clean and running off a consistent base.
Some more comparisons:
- With Binci you don't need a Dockerfile for local development, thus you can use it whether or not your project will be deployed in Docker or to bare metal.
- Binci only builds Docker images if you want it to. Specifying an image in the config will run all tasks off of that image without ever building a local one first.
- When you install local dependencies in your project folder, run a build, execute your coverage tool, or write any local files, that just happens on your hard disk, not locked away in some container. They'll be available to every other task you run.
- With Binci you don't need to run tasks in a containerized shell, you simply define the tasks and run them. You can kick tasks off with any local script, build tool, or IDE run configuration without building a container first.
- Tasks don't need to be defined at runtime via arguments or flags, you just tell Binci which predefined task to run.
Binci is licensed under the MIT license. Please see LICENSE.txt for full details.
Binci was originally created at TechnologyAdvice in Nashville, TN.