How can I have unique build numbers across branches with Jenkins & the Pipeline Multibranch Plugin

Question

We are using Jenkins Pipeline Multibranch Plugin with Blue Ocean.

Through my reading, I believe it is quite common to tie your project's build number to the Jenkins run, as this allows traceability from an installed application through to the CI system, then to the change in source control, and then onto the issue that prompted the change.

The problem is that for each branch, the run number begins at 0. For a project with multiple branches, it seems impossible to guarantee a unique build number.

Answer 1

You can get the Git branch name from $GIT_BRANCH and add this to $BUILD_NUMBER to make an ID that's unique across branches (as long as your company doesn't do something like get themselves taken over by a large corporation that migrates you to another Jenkins server and resets all the build numbers: to protect against that, you might want to use $BUILD_URL).

Only snag is $GIT_BRANCH contains the / character, plus any characters you used when naming the branch, and these may or may not be permitted in all the places where you want an ID. ($BUILD_URL is also going to contain characters like : and /) If this is an issue, one workaround would be to delete unwanted characters with tr:

export MY_ID=$(echo $GIT_BRANCH-$BUILD_NUMBER | tr -dc [A-Za-z0-9-])

(-dc means delete the complement of these characters, so A-Z, a-z, 0-9 and - are the characters you want to keep.)

Answer 2

Maybe instead of a unique (global numeric) build number you might want to try a unique (global) build display name?

According to "pipeline syntax: global variables reference" currentBuild.displayName is a writable property. So you could e.g. add additional information to the build number (in order to make it globally unique) and use that string in subsequent artifact/application build steps (to incorporate that in the application's version output for your desired traceability), e.g. something like:

currentBuild.displayName = "${env.BRANCH_NAME}-${currentBuild.id}"

Using the build's schedule or start time formatted (currentBuild.timeInMillis) as a readable date, or using the SCM revision might be also useful, e.g. resulting in "20180119-091439-rev149923".

See also:

• https://groups.google.com/forum/#!msg/jenkinsci-users/CDuWAYLz2zI/NLxwOku4AwAJ
• https://support.cloudbees.com/hc/en-us/articles/220860347-How-to-set-build-name-in-Pipeline-job

Answer 3

One way is to have a Job that is being called from all branches and using it's build number. That job can be just a normal pipeline job with a dummy Jenkinsfile like echo "hello". Then just call it like this

def job = build job: 'build number generator', quietPeriod: 0, parameters: [
        string(value: "${BRANCH_NAME}-${BUILD_NUMBER}", name: 'UID')  
]
def BNUMBER = job.getNumber().toString()
currentBuild.displayName = "build #"+BNUMBER
echo BNUMBER

Not sure if that UID parameter is needed but it forces all calls into "build number generator" job to be unique so Jenkins wouldn't optimize builds that happen at same time to use same "build number generator" job.