Implementing Elastic Jenkins Agents on Kubernetes Using ScriptConsole and Shared Library

The article continues a CI/CD series focused on building an elastic build pool on Kubernetes, demonstrating how to automatically provision a Jenkins agent for a pipeline run and recycle it after the job finishes.

The approach stores the Groovy scripts that create and delete agents in the resources/scripts directory of a shared library, loads them in the pipeline with libraryResource , replaces the placeholder __AGENT_NAME__ with a generated name, writes the script to a local file, and then executes it via Jenkins ScriptConsole API after obtaining a CSRF crumb.

Example of loading the shared library and setting environment variables:

@Library("mylib@feature-k8s") _      // load shared library
import org.devops.*
env.namespace = "jenkins"
env.agentName = "jenkinsagent${UUID.randomUUID().toString()[0..7]}"

The following Groovy helper methods perform the actual API calls. RunScriptConsole sends the script content to /scriptText using curl with the crumb header, and GetCrumb retrieves the crumb value from /crumbIssuer/api/json :

// ScriptConsole runs a script
def RunScriptConsole(scriptContent, crumb){
response = sh returnStdout: true,
script: """
curl -s -d \"script=$(cat ${scriptContent})\" \
/>            --header \"Jenkins-Crumb:${crumb}\" \
/>            -X POST http://admin:[email protected]:8080/scriptText\n/>        """
try {
response = readJSON text: response - "Result: "
} catch(e){
println(e)
}
return response
}
// Get Crumb value
def GetCrumb(){
response = sh returnStdout: true,
script: """
curl -s -u admin:admin \
/>            --location \
/>            --request GET 'http://192.168.1.200:8080/crumbIssuer/api/json'\n/>        """
response = readJSON text: response
return response.crumb
}

Pipeline stages use these helpers to create and delete the agent pod:

stage("CreateBuildENV"){
steps{
script{
// Get agent create script
scriptContent = libraryResource 'scripts/create_jenkins_agent.groovy'
scriptContent = scriptContent.replaceAll("__AGENT_NAME__", "${env.agentName}")
writeFile file: 'create_jenkins_agent.groovy', text: scriptContent
// Run agent create script
jenkinsCrumb = GetCrumb()
agentConfig = RunScriptConsole("create_jenkins_agent.groovy", jenkinsCrumb)
}
}
}
stage("DeleteBuildENV"){
steps{
script{
// Get agent delete script
scriptContent = libraryResource 'scripts/delete_jenkins_agent.groovy'
scriptContent = scriptContent.replaceAll("__AGENT_NAME__", "${env.agentName}")
writeFile file: 'delete_jenkins_agent.groovy', text: scriptContent
// Run agent delete script
jenkinsCrumb = GetCrumb()
agentConfig = RunScriptConsole("delete_jenkins_agent.groovy", jenkinsCrumb)
}
}
}

The current implementation uses plain credentials in the curl command, which should be moved to Jenkins credentials for better security. It also notes that the created agent is not yet connected to the Jenkins master, so further configuration is required.

In summary, when the pipeline runs, a Jenkins agent pod is automatically created and later destroyed, demonstrating elastic build pool capabilities; a live session is advertised for deeper debugging of the Kubernetes API and full elastic build pool setup.