Skip to content

Latest commit

 

History

History
74 lines (51 loc) · 2.6 KB

README.md

File metadata and controls

74 lines (51 loc) · 2.6 KB
Raw

Marathon CI pipeline

In order to be CI tool agnostic and provide the benefit of running all CI tasks on a "local" machine, the marathon team has moved to using Ammonite for CI pipelining tasks. Ammonite is a Scala based scripting tool and is easiest to install on a Mac with brew install ammonite-repl. Other platforms please read the Ammonite site. The ci folder in the root project contains the ammonite scripts. The script requires the project build requirements such as a JDK, Scala and sbt in the path.

Using Ammonite

The ammonite script is a runnable script. To get a list of functions to invoke execute: ./ci/pipeline

To execute a particular function just invoke the script with function appended as in: ./ci/pipeline compileAndTest which will compile and test the marathon project.

Pipeline Targets

The ci/pipeline script defines two primary targets

  1. jenkins
  2. pr
  3. release

The jenkins target is executed on every release branch, on master, and on each Github pull request. The jenkins target will check if a build is for a pull-request, and if so, run forward to the pr target.

The pr target runs the test pipeline followed by Github PR review reporting. It is triggered with each diff update.

The release target runs the build and packages Marathon in a Docker image, native packages for various Linux distributions and a big jar.

The test pipeline involves the following steps:

  • Compile
  • Run unit tests
  • Run integration tests
  • Build and upload snapshot artifacts

Additionally, for master builds, we:

  • Update the DCOS packages for Marathon by pushing an update to the snapshot Marathon DCOS branch (for OSS and Enterprise).

Sub Targets

The provision.* targets prepare the Jenkins AWS nodes by killing leak processes from older test runs and updating Mesos. You can run them locally but be careful. The killStaleTestProcesses might kill process you don't want to be gone.

The compileAndTest target basically runs sbt clean test integration/test. This is the main compilation step.

The build target assembles Marathon binary packages and generates the sha1 checksums for the zip and tarball packages. See createPackageSha1s in the code base for details.

There are several targets in githubClient:

  • reject
  • comment
  • reportSuccess
  • reportFailure

Method reject and reportFailure rejects a GitHub pull request review while reportSuccess approves it. The comment method posts a comment to the pull request.

One can comment from the CLI with

amm ./ci/githubClient.sc comment 777 "test from pipeline"