Hello,
I have a long-running job that uses a single agent with a custom workspace, and it uses the disableConcurrentBuilds() option because two jobs would interfere with each other when using the same workspace. I need to use a custom workspace because its size is several hundred gigabytes, and checking it out from SCM takes a long time. To avoid this, I reuse the same workspace for all runs of this job.
Recently, I added another agent with its own custom workspace (which has the same path as the original agent’s workspace), and I want to be able to run this job concurrently on both agents, but not on the same agent. So, I removed disableConcurrentBuilds() and added resource locking in the options {} block.
I am testing this setup with the following pipeline:
pipeline {
agent {
node {
label "${params.Agent}"
}
}
options {
lock "label-${env.JOB_BASE_NAME}-${env.NODE_NAME}"
}
parameters {
choice(name: 'Agent', choices: ['agents-label', 'agent01', 'agent02'], description: 'Choose a specific agent or leave the default label for automatic selection.')
}
stages {
stage('Running') {
steps {
echo "Starting ${env.JOB_BASE_NAME} #${currentBuild.number} on ${env.NODE_NAME}."
sleep time: 1, unit: 'MINUTES'
echo "Finished ${env.JOB_BASE_NAME} #${currentBuild.number} on ${env.NODE_NAME}."
}
}
}
}
When I start the job with the agents-label parameter, it tries to lock label-mytestjob-null, which isn’t correct. I believe this happens because the agent is not assigned when the lock is executed in the options block.
In my original long-running pipeline, there are several stages, and placing steps inside the lock {} block is not an option because the lock would be released between stages.
How can I achieve the desired behavior? Is it possible to lock the entire job, for example, in a script after the agent is assigned but before the stages start?
Thank you in advance for your help.