The generated config XML for the branch source contains an <id> element. Job DSL generates a random ID each time the seed job runs. I think the Multibranch plugin uses this ID to manage the repository state. So it's not OK to generate a random ID with each run of the seed job.

You can use the Automatically Generated DSL to set the ID:

multibranchPipelineJob('job-dsl-plugin_multibranch') {
  branchSources {
    branchSource {
      source {
        git {
          id('must-be-unique')
          remote('https://github.com/jenkinsci/job-dsl-plugin.git')
          credentialsId(null)
          remoteName('origin')
          rawRefSpecs('+refs/heads/*:refs/remotes/origin/*')
          includes('*')
          excludes('')
          ignoreOnPushNotifications(false)
        }
      }
    }
  }
}

HUZZAH!  I've been fighting with this bug for a month or two now at work. daspilker Thank you that fixed the issue!

Code changed in jenkins
User: Daniel Spilker
Path:
docs/Home.md
job-dsl-core/src/main/groovy/javaposse/jobdsl/dsl/helpers/workflow/BranchSourcesContext.groovy
job-dsl-core/src/main/groovy/javaposse/jobdsl/dsl/helpers/workflow/GitBranchSourceContext.groovy
job-dsl-core/src/main/groovy/javaposse/jobdsl/dsl/helpers/workflow/GitHubBranchSourceContext.groovy
job-dsl-core/src/test/groovy/javaposse/jobdsl/dsl/helpers/workflow/BranchSourcesContextsSpec.groovy
http://jenkins-ci.org/commit/job-dsl-plugin/1bd2d95eaae1356de6b3f92613766527280f41ca
Log:
enhanced support for the Pipeline Multibranch Plugin

[FIXES JENKINS-43693]

Code changed in jenkins
User: Daniel Spilker
Path:
docs/Home.md
job-dsl-core/src/main/groovy/javaposse/jobdsl/dsl/helpers/workflow/BranchSourcesContext.groovy
job-dsl-core/src/main/groovy/javaposse/jobdsl/dsl/helpers/workflow/GitBranchSourceContext.groovy
job-dsl-core/src/main/groovy/javaposse/jobdsl/dsl/helpers/workflow/GitHubBranchSourceContext.groovy
job-dsl-core/src/test/groovy/javaposse/jobdsl/dsl/helpers/workflow/BranchSourcesContextsSpec.groovy
http://jenkins-ci.org/commit/job-dsl-plugin/f8b51912215c545d702b6e8c8fbe6fc129fb3130
Log:
Merge branch 'JENKINS-43693'

Compare: https://github.com/jenkinsci/job-dsl-plugin/compare/c722b8bb538c...f8b51912215c

I believe I am still hitting this issue. After updating to jobDSL 1.66, after running my jobDSL job, if I try to build one of the updated multibranch projects it fails with `ERROR: Could not determine exact tip revision of master`. Is it possible this is still unfixed?

I face the same issue.

Jenkins - v2.60.1

Job DSL - v1.64

Pipeline Multibranch - v2.16

I have just commited to test branch but scanning has triggered build for all branches which had Jenkinsfile in it.

Checking branch develop
      ‘Jenkinsfile’ found
    Met criteria
Takeover for example_project » develop by source #1 from source that no longer exists
Branch reopened: develop (11617f79c6408e1d5fd54ea60a64d31b787a9471)
Scheduled build for branch: develop

Checking branch test
      ‘Jenkinsfile’ found
    Met criteria
Takeover for example_project » test by source #1 from source that no longer exists
Branch reopened: test (f84bcda107da005f9363759a68af2008a135a84e)
Scheduled build for branch: test

Checking branch master
      ‘Jenkinsfile’ found
    Met criteria
Takeover for example_project » master by source #1 from source that no longer exists
Branch reopened: master (2a2521fe4ea2ab71edb51b3279b5e49c4c8e77d5)
Scheduled build for branch: master

let me know if you need more information.

EDIT:

Jenkins is integrated with GitLab v. 10.0.3 CE

So the changes made to fix this one seem to still use a random UUID - I'm not sure whether or not generating it based off the name or the repo being cloned (or why not both) would fix this - If I'm reading the problem correctly you need a static UUID being generated for the job for it to be able to cope  - https://github.com/jenkinsci/job-dsl-plugin/compare/c722b8bb538c...f8b51912215c#diff-f468c06e75a4e86856c95107995aaa0fR6

(and yes I'm also still getting this issue, and have been wondering what is causing it for a while - going to try setting the id manually for a bit instead)

I have created a fresh environment from 2.121 and installing:

• Pipeline: Multibranch 2.18
• Github Branch Source Plugin 2.3.4
• Job DSL 1.69

I was able to reproduce the issue using the steps in the description. I can still see

Takeover for job-dsl-plugin_multibranch » master by source #1 from source that no longer exists
Branch reopened: master (5045094d82daa475fe79e7f5e52eba58c6df9020)
Scheduled build for branch: master

after I re-run the Seed Job.

Sorry, daspilker . It looks like I misunderstood the resolution. Is the resolution to manually add unique Ids to the source definitions? Like this:

multibranchPipelineJob('job-dsl-plugin_multibranch') {
  branchSources {
    def repo = 'https://github.com/jenkinsci/job-dsl-plugin.git'
    git {
      id = UUID.nameUUIDFromBytes(repo.getBytes())
      remote(repo)
    }
  }
}

duemir yes, it worked for me and I use repository name for id.

Should this documentation be updated to include setting the id?

https://jenkinsci.github.io/job-dsl-plugin/#path/multibranchPipelineJob

I'm with rkivisto on this.

I don't think this issue should be "Closed/Fix" if the default functionality still doesn't work as expected. 

The ability to set the `id` does solve the problem, but in my opinion this is a workaround. And if a workaround is required, this should be clearly documented. Or even better, the default functionality should be the expected functionality. 

Re-opening so we can track the discussion about if this is truly fixed or not, or if we should just update the plugin documentation with the `id` setting.

PR for improving the documentation and removing support for randomly generated identifiers:

https://github.com/jenkinsci/job-dsl-plugin/pull/1187

To sum things up, you need to set a constant and unique identifier for each branch source to make things work. That's a delicacy of the SCM API plugin. There must be unique identifier for each branch source, no two branch sources either in the same or different jobs may share the same ID. It's your responsibility to keep the ID unique when using Job DSL. The ID must not change over time, e.g. on the next run of the seed job.

daspilker is there a way in Job DSL to resemble the automatic id creation as done when the job is created via the UI? Or can the id automatically be set only on job creation but not changed when updating the job?

renescheibe Job DSL should not replicate any plugin logic. The behavior will change over time, will most likely fail silently and cause more problems.

I'll close this issue since there no more known problems.

Still an issue: Enforcing to manually set the id is only possible for branch sources explicitly implemented in Job DSL. For branch sources that are provided via "dynamic" DSL that's not enforced.

That's for example all the sources under branchSources - branchSource which are for example:

• bitbucket
• git
• github
• mercurialSCMSource
• subversionSCMSource

Do you think these can also be covered somehow?

I faced the same issue even though I don't use the Job DSL plugin to configure a GitHub Organization pipeline. It was created in UI. Any workarounds?

Ditto here!  Would love to see resolution using the Job DSL plugin to configure a Gitlab Group pipeline