A lot has been written about the advantages of microservices architecture over the monolith approach to app development, and the case for employing microservices is strong. The issue is: how do you pull it off?
One method that we’ve found extremely beneficial at ClearScale is the use of the Twelve-Factor App methodology. It was created by developers at Heroku who observed that successful apps share some common principles.
While the original principles were based on apps developed on Heroku platform, they’re technology- and language-agnostic. They work well for breaking monolithic apps into microservices and for developing cloud-native apps from scratch using microservices.
The following provides an overview of using the Twelve-Factor App methodology with some information specific to microservices. You can view the Heroku team’s complete document here.
One codebase tracked in revision control, many deploys
There is only one codebase per app ─ in microservices architecture, it’s one codebase per service ─ but there can be numerous deploys. A deploy is a running instance of the app. This is usually a production site and one or more staging sites. A copy of an app running in a developer’s local development environment is also considered a deploy.
The codebase must be the same across all deployments, although different versions may be active in each deploy. For this reason, it’s essential to track the app in a version control system, such as Subversion and Git.
Explicitly declare and isolate dependencies
Declare all dependencies precisely via a dependency declaration manifest. Use a dependency isolation tool during execution to ensure that no implicit dependencies enter from the surrounding system. This applies to both production and development.
Don’t rely on the implicit existence of any system-wide site packages or site tools. While these tools may exist on most systems, there’s no guarantee they’ll exist on the systems where the app runs in the future. Nor is it a sure thing that the version found on a future system will be compatible with the app.
Store config in the environment
All configuration data should be stored separately from the code – in the environment as variables and not in the code repository — and read in by the code at runtime. A separate config file makes it easy to update the config values without touching the actual codebase, eliminating the need for redeploying apps when config values are changed.
The following are some additional best practices to consider:
- Use an environment variable for anything that can change at runtime, and for secrets that shouldn’t be committed to a shared repository.
- Use non version-controlled .env files for local development.
- Keep all .env files in a secure storage system so they’re available to the development teams, but not committed to the code repository being used.
- Once an app is deployed to a delivery platform, use the platform’s mechanism for managing environment variables.
Configs stored as variables are unlikely to be checked into the repository accidentally. Another bonus: then your configs are independent of language and OS.
4. Backing services
Treat backing services as attached resources
Backing services are usually managed locally by the same systems administrators who deploy the app’s runtime. The app may also have services provided and managed by third parties.
For microservices, anything external to a service is treated as an attached resource. This ensures that every service is completely portable and loosely coupled to the other resources in the system. In addition, the strict separation increases flexibility during development; developers only need to run the services they’re modifying.
5. Build, release, run
Strictly separate build and run stages
Use a continuous integration/continuous delivery (CI/CD) tool to automate builds and support strict separation of build, release, and run stages, Docker images make it easy to separate the build and run stages. Images should be created from every commit and treated as deployment artifacts.
Start the build process by storing the app in source control and then build out its dependencies. Separating the config information enables you to combine it with the build for the release stage. It’s then ready for the run stage. Each release should have a unique ID.
Execute the app as one or more stateless processes
Ensuring the app is stateless makes it easy to scale a service horizontally by simply adding more instances of that service. Store any stateful data or data that needs to be shared between instances, in a backing service such as a database. The process’ memory space or filesystem can be used as a brief, single-transaction cache. Session state data is a good candidate for a datastore that offers time expiration.
Never assume that anything cached in memory or on disk will be available on a future request or job.
7. Port binding
Export services via port binding
The app should be completely self-contained and not rely on the runtime injection of a webserver into the execution environment to create a web-facing service. The web app exports HTTP as a service by binding to a port, and listening to requests coming in on that port. Nearly any kind of server software can be run via a process binding to a port and awaiting incoming requests.
To make the port binding factor more useful for microservices, allow access to the persistent data owned by a service only by way of the service’s API. This prevents implicit service contracts between microservices. It also ensures they can’t become tightly coupled. In addition, data isolation allows the developer to choose the type of data store for each service, that best suits its needs.
Scale-out via the process model
Supporting concurrency means that different parts of an app can be scaled up to meet the need at hand. When you develop the app to be concurrent, you can spin up new instances to the cloud effortlessly.
This principle draws from the UNIX model for running service daemons, enabling an app to be architected to handle diverse workloads by assigning each type of work to a process type. To do this, organize processes according to their purpose and then separate them so that they can be scaled up and down according to need.
Docker and other containerized services provide service concurrency.
Maximize robustness with fast startup and graceful shutdown
An app’s processes should be disposable so they can be started, stopped, and redeployed quickly with no loss of data. This facilitates fast elastic scaling, rapid deployment of code and config changes, and robustness of production deploys.
Services deployed in Docker containers do this automatically, as it’s an inherent feature of containers that they can be stopped and started instantly.
The concept of disposable processes means that an app can die at any time, but it won’t affect the user—the app can be replaced by other apps, or it can start right up again.
Building disposability into your app ensures that the app shuts down gracefully: it should clean up all utilized resources and shut down smoothly.
When designed this way, the app comes back up again quickly. Likewise, when processes terminate, they should finish their current request, refuse any incoming request, and exit.
10. Dev/prod parity
Keep development, staging, and production as similar as possible
Continuous deployment requires continuous integration based on matching environments to limit deviation and errors. As such, dev, staging, and production should be as similar as possible.
Containers work well for this as they enable you to run the exact same execution environment all the way from local development through production. Note: the differences in the underlying data can still cause differences at runtime.
Don’t use different backing services between development and production, even when adapters theoretically abstract away any differences. Even minor deviations can cause incompatibilities to crop up that can cause code that worked in development or staging to fail in production.
Treat logs as event streams
Stream logs to a chosen location rather than dumping them into a log file. Logs can be directed anywhere. For example, they could be directed to a database in NoSQL, to another service, to a file in a repository, to a log-indexing-and-analysis system, or to a data-warehousing system.
Use a log-management solution for routing or storing logs.
The process for routing log data needs to be separate from processing log data. Define your logging strategy as part of the architecture standards, so all services generate logs in a similar fashion.
12. Admin processes
Run admin/management tasks as one-off processes
The idea here is to separate administrative tasks from the rest of the app to prevent one-off tasks from causing issues with your running apps. Containers make this easy, as you can spin up a container just to run a task and then shut it down.
Examples include doing data cleanup, running analytics for a presentation, or turning on and off features for A/B testing.
Though the admin processes are separate, you must continue to run them in the same environment and against the base code and config of the app itself. Shipping the admin tasks code alongside the app prevents drift.
ClearScale is well versed in helping organizations across a wide range of industries develop solutions ranging from cloud migrations to the development of machine learning-powered apps. We employ a variety of methodologies and best practices in developing cloud-native apps with the approach often driven by our customers’ specific needs and goals.
You can also contact us to find out how ClearScale can help you achieve your business objectives related to app development. We can take on projects for you or work in conjunction with your in-house team.