Test Harness: create agent with multiple executors?

Using Jenkins Ask a question
,
1

I have this test that uses jenkins-test-harness:

class JenkinsUtilsTest extends spock.lang.Specification {

    @spock.lang.Shared
    @org.junit.ClassRule
    JenkinsRule harness = new JenkinsRule()

    void setupSpec() {
        hudson.model.Slave node = harness.createSlave('an-agent', null, null)
        node.numExecutors = 2
        harness.waitOnline(node)
    }

    void 'check parallel works'() {
        given:
        def job = harness.createProject(WorkflowJob)
        job.definition = new org.jenkinsci.plugins.workflow.cps.CpsFlowDefinition('''
        parallel([
            sleeper: {
                node('an-agent') {
                    echo '--- going to sleep'
                    sleep time: 15, unit: 'SECONDS'
                    echo '--- done sleeping'
                }
            },
            waiter: {
                echo '--- pause in favor of sleeper'
                sleep time: 5, unit: 'SECONDS'
                node('an-agent') {
                    echo '--- done waiting'
                }
            }
        ])
        ''', true)

        when:
        WorkflowRun run = harness.buildAndAssertSuccess(job)

        then:
        harness.assertLogContains('show me the logs', run)
    }

When I run it, the logs say:

    [Pipeline] Start of Pipeline
    [Pipeline] parallel
    [Pipeline] { (Branch: sleeper)
    [Pipeline] { (Branch: waiter)
    [Pipeline] node
    [Pipeline] echo
    --- pause in favor of sleeper
    [Pipeline] sleep
    Sleeping for 5 sec
    Running on an-agent in /tmp/j h3074825386784972163/agent-work-dirs/an-agent/workspace/test0
    [Pipeline] {
    [Pipeline] echo
    --- going to sleep
    [Pipeline] sleep
    Sleeping for 15 sec
    [Pipeline] node
    [Pipeline] echo
    --- done sleeping
    [Pipeline] }
    Running on an-agent in /tmp/j h3074825386784972163/agent-work-dirs/an-agent/workspace/test0
    [Pipeline] // node
    [Pipeline] {
    [Pipeline] }
    [Pipeline] echo
    --- done waiting
    [Pipeline] }
    [Pipeline] // node
    [Pipeline] }
    [Pipeline] // parallel
    [Pipeline] End of Pipeline

, so clearly there is only one Executor on the agent. I can confirm this asserting node.computer.executors.size() == 2 in the setup - it is actually 1, just like the hudson.model.Slave#getNumExecutors() docs state:

This may be different from getExecutors().size() because it takes time to adjust the number of executors.

But how do I actually ensure that the number of executors has adjusted? Or better yet, force the adjustion - as I tried Thread.sleep(60000) and it did nothing.

2

you will need to do

node.toComputer().setNumExecutors(2)
3

Thank you for chiming in @mawinter69 . There does not seem to be setNumExecutors in the hudson.slaves.SlaveComputer object returned by .toComputer() or .getComputer() – only in hudson.model.Slave itself…

JenkinsUtilsTest > classMethod FAILED
    groovy.lang.MissingMethodException: No signature of method: hudson.slaves.SlaveComputer.setNumExecutors() is applicable for argument types: (java.lang.Integer) values: [2]
    Possible solutions: getNumExecutors(), getAllExecutors(), getExecutors()
4

ah sorry setNumExecutors is private in Computer class

5

@mawinter69 do you maybe have some other ideas by any chance, or maybe know someone else we can try pinging here? :folded_hands:

6

You can use the following trick

node.numExecutors = 2
harness.jenkins.getNodesObject().replaceNode(node, node);
1 Like
7

Thank you very much Markus, that did it! Was afraid it would incur some additional waits for reinitialization to complete, but seems to go almost instantly. :tada:

1 Like