Reputation: 677
I have a Jenkins pipeline job called "TestPipeline". I want to trigger a build on 2 different slaves which labeled "tester1' and "tester2". And the pipeline script is quite simple here:
node('tester1') {
build 'test_job'
}
node('tester2') {
build 'test_job'
}
However when I run the TestPipeline job, the "test_job" won't run on the nodes which I assigned. But run on random node instead.
I'm wondering if I should set "Restrict where this project can be run" on my "test_job". So I set it to "tester" (The "tester" label contains both node "tester1" and "tester2"). But when I run the pipeline job again, the "test_job" runs on "tester2" twice. I should expect the job to run on "tester1" first and then run on "tester2". Why is that? Is it because the "node" step doesn't matter when it comes to which node the build step should be built on?
Upvotes: 26
Views: 89348
Reputation: 151
Here is how i got it working
create a 'job' 'test_job' with parameter type 'label' , name 'node', value can be any string. (this is the job to be triggered) .Set 'Restrict where this project can be run' to the label value
create a 'pipeline' with parameter type "Node". From the pipeline section, use the following script
Note the label for test_job is ${env.NODE_NAME} which will be set by the pipeline based on user's choice
pipeline {
agent any
stages {
stage('Build') {
steps {
echo 'Building..'
script {
build job: 'test_job', parameters: [
[$class: 'LabelParameterValue', name: 'node', label: "${env.NODE_NAME}" ]
]
}}}}}
Upvotes: 8
Reputation: 846
I did the same, but using Node parameter, so I can use nodes list running job directly, or running pipeline (in pipeline I use Choice parameter to have list of available nodes). So in my case:
node_name
node
, put there possible choices, so running pipeline you can choose on which node it should be run.Code to insert into pipeline script:
build job: 'my_job',
parameters: [[$class: 'NodeParameterValue', name: 'node_name', labels: ["$node"], nodeEligibility: [$class: 'AllNodeEligibility']]]
Upvotes: 9
Reputation: 677
Please see the bug here. The solution is as follows.
build job: 'test_job', parameters: [[$class: 'LabelParameterValue', name: 'node', label: 'tester1']]
build job: 'test_job', parameters: [[$class: 'LabelParameterValue', name: 'node', label: 'tester2']]
And the job will be built as I wanted.
However, I think it is only a workaround. I still believe this is a bug. Because the node step should do its job instead of letting other plugins to do for it.
Upvotes: 31
Reputation: 322
I just tested this on my installation, and it correctly ran each script on each node. You might want to check that you've configured your slaves correctly. I believe the documentation says they need to have the Launch slave agents via Java Web Start setting, you could verify that.
Upvotes: 0