How to get a list of all Jenkins nodes assigned with label including master node?

Question

I'm creating a Jenkins pipeline job and I need to run a job on all nodes labelled with a certain label.

Therefore I'm trying to get a list of node names assigned with a certain label. (With a node I can get the labels with getAssignedLabels())

The nodes-list in jenkins.model.Jenkins.instance.nodes seems not contain the master-node which I need to include in my search.

My current solution is to iterate over the jenkins.model.Jenkins.instance.computers and use the getNode()-method to get the node. This works, but in the javadoc of Jenkins I'm reading the this list might not be up-to-date.

In the long-run I will add (dynamically) cloud-nodes and I'm afraid that I won't be able to use computers then.

What is the right way to get the list of all current nodes?

This is what I'm doing right now:

@NonCPS
def nodeNames(label) {
    def nodes = []
    jenkins.model.Jenkins.instance.computers.each { c ->
        if (c.node.labelString.contains(label)) {
            nodes.add(c.node.selfLabel.name)
        }
    }   
    return nodes
}

Answer 1

Updated answer: in a pipeline use nodesByLabel to get all nodes assigned to a label.

Answer 2

This is the way I'm doing it right now. I haven't found anything else:

@NonCPS
def hostNames(label) {
  def nodes = []
  jenkins.model.Jenkins.get().computers.each { c ->
    if (c.node.labelString.contains(label)) {
      nodes.add(c.node.selfLabel.name)
    }
  }
  return nodes
}

jenkins.model.Jenkins.get.computers contains the master-node and all the slaves.

Answer 3

Update to @patrick-b answer : contains can be buggy if you have labels containing same string, I've added a split step do check every label separated with spaces.

@NonCPS
def hostNames(label) {
    def nodes = []
    jenkins.model.Jenkins.get.computers.each { c ->
        c.node.labelString.split(/\s+/).each { l ->
            if (l != null && l.equals(label)) {
                nodes.add(c.node.selfLabel.name)
             }
        }
    }

    return nodes
}

Answer 4

Here is a functional solution which is more readable and concise:

def nodes = jenkins.model.Jenkins.get().computers
  .findAll{ it.node.labelString.contains(label) }
  .collect{ it.node.selfLabel.name }

You can verify it in the Jenkins Script Console.

Answer 5

Try using for (aSlave in hudson.model.Hudson.instance.slaves) {} and aSlave.getLabelString()); to get all the labels for all of your nodes. You can construct a list of nodes per label this way.

Answer 6

I think that you can do this with:

def nodes = Jenkins.get.getLabel('my-label').getNodes()
for (int i = 0; i < nodes.size(); i++) {
    node(nodes[i].getNodeName()) {
        // on node
    }
}

I don't know for sure whether this works with cloud nodes.

Answer 7

using nodesByLabel as pointed out by @towel is probably the solution in most cases. One limitation I found with nodesByLabel is that there is no way to indiscriminately select all nodes. I can't use any of the other solutions either because of script security, some of these can be pretty dangerous, so I preferred not to approve them for usage.

As an alternatively, you can add a function as a pipeline library, which will allow usage of these functions. Since pipeline libraries can be set up so they are fully under the administrator's control, it's safer to go with this route. To do so, set up a pipeline library (I don't think it matters if it's global or not, but for me it is). Then add the following contents to the file vars/parallelRunOnNodes.groovy:

def call(Closure callback) {
    parallel jenkins.model.Jenkins.get().computers.collectEntries { agent ->
        def nodeLabel = agent.node.selfLabel.name
        ["${nodeLabel}": {
            node("${nodeLabel}") {
                stage("${nodeLabel}") {
                    callback(nodeLabel)
                }
            }
        }]
    }
}

Which can then be used as follows:

pipeline {
    agent none
    stages {
        stage('Parallel on all nodes') {
            steps {
                parallelRunOnNodes { nodeLabel ->
                    println(nodeLabel)
                }
            }
        }
    }
}

Obviously adjust as you see fit, e.g. you can add additional parameters to filter, maybe you don't care about parallel etc.