Day-to-day development is a big part of DevOps and release automation. If we can't trust that our development processes will result in successful production deployment, we'll find ourselves doubling the effort needed to create and deploy new services.
Architect Components help ensure consistency between local debugging and remote deployments by capturing infrastructure-agnostic architectural details. However, there remains important cases for differing build and run processes locally that are not suitable for remote deployments. In order to enable these different use cases, Architect services can specify a
debug configuration block – a blanket way to override service/task configuration exclusively for debugging scenarios. In this guide, you'll learn how you can use this
debug block to enable different key development features for your Component.
The most common debugging requirement is the ability for a running application to respond live to changes made to the underlying source code, commonly referred to as "hot-reloading". Achieving hot-reloading in a containerized setting requires two key steps: mounting the code from your host machine with a directory inside the container, and running a version of your start command that will automatically rebuild the running application whenever it detects code changes.
Each language and framework has its own tools to detect and react to code changes. Below is an example of how to do so for a Node.js application:
name: component-namedescription: An example component that includes debugging details to faciliate hot-reloadingservices:api:build:context: ./interfaces:api: 8080# Local configuration for api servicedebug:build:# The main difference between the main dockerfile and the debug one is the need# to install `devDependencies`, which in this case includes the `nodemon` utility# used for hot-reloading.dockerfile: Dockerfile.debug# With `nodemon` installed, we can now use that to bootup our app and enable hot-reloadingcommand: nodemon index.js# Lastly, we'll need to sync our local source code with the expected directory inside the# container. Otherwise, nodemon won't detect any changes after container startup.volumes:src:host_path: ./src/mount_path: /usr/src/app/
Another common need for teams working on rapid code changes is the use of an IDE attached debugger and breakpoints. Two things are needed to accomplish this: exposure of the debugger port, and execution of a start command that includes a debugger. An example component specification is as follows:
name: component-nameservices:api:build:context: ./interfaces:api: 8080# Local configuration for api servicedebug:# First thing we need to do is attach the debugger. Node has a handy built-in flag, `--inspect`.command: node --brk-inspect=0.0.0.0:9229 index.js# Next, we need to expose the port the debugger is listening on.interfaces:debug: 9229
Once the debugger port is exposed, you'll need to attach it to your IDE. Each IDE has different instructions for doing this, but you can see an example using VS Code in the following article: How to debug a Node.js app in a docker container.
Containers are designed to be self-contained, but that means that when they disappear so do their filesystems. This can be a major nuisance for local dev since your database and other stateful services will lose their contents forcing you to go through the same workflows over and over.
In order to avoid this, you'll have to mount the data volume for the database to your host machine. Fortunately, this can be done easily with Architect:
name: component-nameservices:db:image: postgres:13interfaces:postgres: 5432# Local configuration for db servicedebug:environment:PGDATA: /var/lib/postgresql/data/pgdatavolumes:data:mount_path: /var/lib/postgresql/data/host_path: ./.data/
The above will store the contents in a
.data directory relative to the architect.yml file. We recommend you add this directory to your gitignore so it doesn't get checked into version control.
The debug block is used whenever local source code is leveraged instead of built and published Component artifacts. This can happen either because you're deploying a component directly from source, or if you've used linking to use local code to fulfill dependency references:
# Deploying a component from source$ architect dev ./architect.yml# Linking a component and using by name$ architect link ./architect.yml$ architect deploy <name>
Sometimes you might want to test out the production deployment process before registering the component. To simulate the regular deployment flow, use the
$ architect dev ./architect.yml --debug=false
This set the environment variable
PGDATA or mount the volume.