릴리스 버전 관리 및 배포

When you build an Erlang application for production, it's bundled into a release. Just like any software, releases need clear version numbers.

Versioning helps you:

• Track changes over time.
• Identify specific builds.
• Manage compatibility with other systems.
• Plan for upgrades and rollbacks.

In Erlang, versioning happens at different levels. Each individual application (an .app file) has its own version. The overall release (the .rel file) also has a version that typically reflects the entire bundle.

The vsn key in an .app file defines an application's version.

A widely adopted standard is Semantic Versioning (SemVer), typically formatted as MAJOR.MINOR.PATCH. It's a great practice to follow for your Erlang applications.

• MAJOR: Incremented for incompatible API changes.
• MINOR: Incremented for new functionality in a backward-compatible way.
• PATCH: Incremented for backward-compatible bug fixes.

This helps users understand the impact of an upgrade.

You'll typically apply SemVer to your application's vsn in the .app file. When you create a release, its version (relvsn) often matches the main application's version or follows a similar scheme.

Tools like rebar3 manage these versions when building releases.

Once you have a versioned Erlang release, how do you get it running in production? There are several strategies, depending on your system's requirements for uptime and complexity.

• Cold Deployment: Simplest, involves downtime.
• Hot Upgrades: Zero-downtime (covered in previous lesson).
• Blue/Green or Canary Deployments: More advanced, often used with orchestration tools.

A cold deployment is the most straightforward method. You stop your currently running Erlang system completely, replace its files with the new release, and then start the new version.

This method always incurs downtime, making it suitable for systems where brief interruptions are acceptable.

Recall from the previous lesson that hot code loading allows you to upgrade a running Erlang system without stopping it.

While powerful for applying patches or minor feature additions, hot upgrades are specific to changing *parts* of a running system. A full deployment often refers to the initial setup or a complete replacement of the entire release, sometimes involving a cold restart or more advanced traffic management.

Even with careful testing, new deployments can introduce unexpected issues. A robust deployment strategy always includes a plan for rollbacks.

A rollback means reverting your system to a previous, stable version if the new deployment fails or causes problems. This minimizes the impact of failures and restores service quickly.

Erlang releases are designed to support rollbacks. When you perform a hot upgrade, Erlang generates an .relup script that describes how to move between versions. Crucially, it also knows how to reverse that script to go back.

Even with cold deployments, you can switch back to a previous release by managing your release directories and restarting the desired version.

Order the steps for a typical cold deployment of an Erlang release.

You've learned about the critical aspects of release versioning and deployment in Erlang:

• Versioning: Use vsn in .app files, ideally following SemVer (MAJOR.MINOR.PATCH), to track and manage changes.
• Deployment: Understand cold deployment (with downtime) and its distinction from hot upgrades.
• Rollbacks: Always plan for reverting to a stable previous version to ensure system resilience.

These practices are key to managing robust Erlang applications in production!