[JENKINS-56766] Allow preserved stashes to be used in new run of the job

Ref.: https://jenkins.io/doc/book/pipeline/running-pipelines/#preserving-stashes-for-use-with-restarted-stages

Currently we can request that a stash be kept after a job finishes, even for more than one job (why? it is not clear from the text above how this would be useful. Multiple restarts perhaps?). But it is only available if we restart a stage of a job (so I guess it must have failed). At least is what the above text seems to imply and some empirical data confirms.

As we are storing the stash tar balls anyway, why not make them available for the next job that starts regularly (as opposed to with a restart)? That way we could store some data to be tested on the next run and make a decision based on it.

Possible use cases would be to store something like the last commitId tested so you don't run the tests again for the same code and any other things about the last run that would be useful for the next run.

E.g.:

pipeline {

  agent {

    node {

      label 'centos'

    }

  }

  options {

    preserveStashes(buildCount: 5)

  }

  environment {

    CHANGED = 'False'

  }

  stages {

    stage('Check for changes'){

      agent { 

node { 

  label 'rhel'

}

      }

      steps {

script {

    def cmd = "git ls-remote https://xxx.yyyy.zzz.com/mycode.git master | awk '{print \$1}' | tee last-image-build.txt.new"

    echo "cmd: ${cmd}"

    def newcommitid = sh (returnStdout: true, script: "${cmd}")

    echo "newcommitid: ${newcommitid}"

    try {

unstash 'last-image-build'

    } catch (Exception e) {

echo "last-image-build stash not found, assume changed"

CHANGED = 'True'

    }

    if (fileExists('last-image-build.txt')) {

def oldcommitid = readFile 'last-image-build.txt'

echo "oldcommitid: ${oldcommitid}"

if (oldcommitid != newcommitid) {

    CHANGED = 'True'

    echo "CHANGED new value: ${CHANGED}"

}   

    }

    sh 'mv -f last-image-build.txt.new last-image-build.txt'

    stash name: 'last-image-build', includes: 'last-image-build.txt', allowEmpty:true

    echo "CHANGED: ${CHANGED}"

}

      }

    }

    stage('Build only if changed'){

      when {

expression { environment name: 'CHANGED', value: 'True' }

      }

      steps{

echo "BUILDING"

      }

    }

  }

}

 

 

---
Originally reported by fnasser, imported from: Allow preserved stashes to be used in new run of the job

• status: Open
  
• priority: Minor
  
• component(s): workflow-basic-steps-plugin
  
• resolution: Unresolved
  
• votes: 1
  
• watchers: 3
  
• imported: 20251215-220547
  

Raw content of original issue

Ref.: https://jenkins.io/doc/book/pipeline/running-pipelines/#preserving-stashes-for-use-with-restarted-stages


Currently we can request that a stash be kept after a job finishes, even for more than one job (why? it is not clear from the text above how this would be useful. Multiple restarts perhaps?). But it is only available if we restart a stage of a job (so I guess it must have failed). At least is what the above text seems to imply and some empirical data confirms.


As we are storing the stash tar balls anyway, why not make them available for the next job that starts regularly (as opposed to with a restart)? That way we could store some data to be tested on the next run and make a decision based on it.


Possible use cases would be to store something like the last commitId tested so you don't run the tests again for the same code and any other things about the last run that would be useful for the next run.


E.g.:



pipeline {

  agent {

    node {

      label 'centos'

    }

  }

  options {

    preserveStashes(buildCount: 5)

  }

  environment {

    CHANGED = 'False'

  }

  stages {

    stage('Check for changes'){

      agent { 

        node { 

          label 'rhel'

        }

      }

      steps {

        script {

            def cmd = "git ls-remote https://xxx.yyyy.zzz.com/mycode.git master | awk '{print \$1}' | tee last-image-build.txt.new"

            echo "cmd: ${cmd}"

            def newcommitid = sh (returnStdout: true, script: "${cmd}")

            echo "newcommitid: ${newcommitid}"

            try {

                unstash 'last-image-build'

            } catch (Exception e) {

                echo "last-image-build stash not found, assume changed"

                CHANGED = 'True'

            }

            if (fileExists('last-image-build.txt')) {

                def oldcommitid = readFile 'last-image-build.txt'

                echo "oldcommitid: ${oldcommitid}"

                if (oldcommitid != newcommitid) {

                    CHANGED = 'True'

                    echo "CHANGED new value: ${CHANGED}"

                }   

            }

            sh 'mv -f last-image-build.txt.new last-image-build.txt'

            stash name: 'last-image-build', includes: 'last-image-build.txt', allowEmpty:true

            echo "CHANGED: ${CHANGED}"

        }

      }

    }

    stage('Build only if changed'){

      when {

        expression { environment name: 'CHANGED', value: 'True' }

      }

      steps{

        echo "BUILDING"

      }

    }

  }

}





 


 

Start with the linked Jenkins documentation on preserving stashes for restarted stages and the example using preserveStashes and unstash. Clarify how a preserved stash should be selected for a regular new run, including retention and naming behavior. Done means a new job run can use the preserved data, with the behavior documented and verified by appropriate plugin tests.

Vom Indexierungsmodell aus dem Issue-Text verfasst.