Multiple Jenkins Nodes in Pipeline

Question

Currently we have a jenkins pipeline with 4 stages. Setup, Build, Deploy, Teardown. Deploy and Teardown prompt for manual user input. Because of this, we don`t want manual user input to take up an executor. So, we want to use agent none. However, when resuming, there is no guarentee we get the same jenkins workspace. Stash/unstash says it uses alot of resources, so if you have large files not to use it. Is there a way to get the exact slave, and when resuming, run back on that same slave?

I have something like this now I also tried agent gcp at top level, and putting agent none in manual input

pipeline {
agent none

 environment {
    userInput = false
 }

stages {
    stage('Setup') {
        agent { node { label 'gcp' } }
        steps {
            deleteDir()
            dir('pipelines') {
                checkout scm
            }
            dir('deployment_pipelines'){
                git branch: __deployment_scripts_code_branch, credentialsId: 'jenkins', url: __deployment_scripts_code_repo
            }
            dir('gcp_template_core'){
                git branch: __gcp_template_code_branch, credentialsId: 'jenkins', url: __gcp_template_code_repo
            }
            dir('control_repo'){
                 git branch: _control_repo_branch, credentialsId: 'jenkins', url: _control_repo
            }

            // Copy core templates to the project
            sh('bash deployment_pipelines/deployment/setup.sh gcp_template_core/gcp_foundation/ control_repo')
        }
    }

    stage('Build') {
        agent { node { label 'gcp' } }
        steps {
            sh('printenv') //TODO: Remove. Debug only
            sh('python deployment_pipelines/deployment/build.py control_repo --env ${_env_type_long}')
        }
    }

    stage('Deploy') {
        agent { node { label 'gcp' } }
        steps {
            sh('python deployment_pipelines/deployment/deploy.py control_repo --env ${_env_type_short}')
        }
    }

     stage('Release') {
        steps {
            agent none
            script {
                sh('python deployment_pipelines/deployment/set_manual_approvers.py deployment_pipelines/config/production-release-approvers.yaml -o approver.txt')
                def approvers = readFile('approver.txt')

                try {
                    userInput = input(
                        message: 'Do you want to proceed with Release?',
                        submitter: approvers)
                } catch(err) { // input false
                    //def user = err.getCauses()[0].getUser() //need script approval for getUser()
                    userInput = false
                    // echo "Aborted by [${user}]"
                }
                agent { node { label 'gcp' } }
                if(userInput)
                {
                    sh("echo 'Do Release'")
                }
            }
        }
    }
    stage('Teardown'){
        agent { node { label 'gcp' } }
        steps {
             script {
                def approvers = readFile('approver.txt')

                try {
                    userInput = input(
                        message: 'Do you want to proceed with Teardown?',
                        submitter: approvers)
                } catch(err) { // input false
                    //def user = err.getCauses()[0].getUser() //need script approval for getUser()
                    userInput = false
                    // echo "Aborted by [${user}]"
                }
                if(userInput)
                {
                    sh("echo 'Do Teardown'")
                }
            }
        }
    }
}

post {
    always {
        echo 'DO TEARDOWN REGARDLESS'
    }
}
}

slashpai slashpai · Accepted Answer · 2017-12-18T13:43:09

agent none should be above step block in stage('Release'). You can refer https://jenkins.io/doc/book/pipeline/syntax/#agent for syntax and flow

Multiple Jenkins Nodes in Pipeline

1 Answers