Jenkins：从最后一个阶段将工件从一个代理复制到另一个代理

问题描述

我在声明性管道中使用了多个代理。有没有将工件（input.txt）从agent1复制到agent2？这是我的声明性管道，

pipeline {
    agent none

    stages {
        stage('Build') {
            agent {
                label 'agent1'
            }

            steps {
                sh 'echo arjun > input.txt'

            }
            post {
                always {
                    archiveArtifacts artifacts: 'input.txt',fingerprint: true
                }
            }

        }
        stage('Test') {
            agent {
                label 'agent2'
            }

            steps {
                sh 'cat input.txt'
            }
        }
        }
    }
 

解决方法

您可以使用 Copy Artifact Plugin 来做到这一点。

给定您的 Jenkinsfile，它会变成这样：

pipeline {
    agent none

    stages {
        stage('Build') {
            agent { label 'agent1' }
            steps {
                sh 'echo arjun > input.txt'
            }
            post {
                always {
                    archiveArtifacts artifacts: 'input.txt',fingerprint: true
                }
            }
        }

        stage('Test') {
            agent { label 'agent2' }
            steps {
                // this brings artifacts from job named as this one,and this build
                step([
                    $class: 'CopyArtifact',filter: 'input.txt',fingerprintArtifacts: true,optional: true,projectName: env.JOB_NAME,selector: [$class: 'SpecificBuildSelector',buildNumber: env.BUILD_NUMBER]
                ])

                sh 'cat input.txt'
            }
        }
    }
}

,

使用 stash 和 unstash。

示例来自： https://www.jenkins.io/doc/pipeline/examples/

// First we'll generate a text file in a subdirectory on one node and stash it.
stage "first step on first node"

// Run on a node with the "first-node" label.
node('first-node') {
    // Make the output directory.
    sh "mkdir -p output"

    // Write a text file there.
    writeFile file: "output/somefile",text: "Hey look,some text."

    // Stash that directory and file.
    // Note that the includes could be "output/","output/*" as below,or even
    // "output/**/*" - it all works out basically the same.
    stash name: "first-stash",includes: "output/*"
}

// Next,we'll make a new directory on a second node,and unstash the original
// into that new directory,rather than into the root of the build.
stage "second step on second node"

// Run on a node with the "second-node" label.
node('second-node') {
    // Run the unstash from within that directory!
    dir("first-stash") {
        unstash "first-stash"
    }

    // Look,no output directory under the root!
    // pwd() outputs the current directory Pipeline is running in.
    sh "ls -la ${pwd()}"

    // And look,output directory is there under first-stash!
    sh "ls -la ${pwd()}/first-stash"
}