-
Bug
-
Resolution: Fixed
-
Minor
-
None
-
Jenkins 2.190.1
InfluxDB Plugin 2.0.2
-
-
2.5
I just tried out the Global Listener feature and it seems to work great, except build_agent_name is always blank. This works fine for me if I disable Global Listener and use influxDbPublisher() in my Jenkinsfile so at least there is a workaround. I'm guessing this has something to do with the global listener not seeing the NODE_NAME environment variable.
[JENKINS-60284] build_agent_name not working with Global Listener
Digging a bit deeper into this I noticed that the `build_agent_name` is correctly set in freestyle jobs, but not in pipelines. In `GlobalRunListener.java` there's a part
EnvVars env; try { env = build.getEnvironment(listener); } catch (IOException | InterruptedException e) { env = new EnvVars(); }
So, basically pipelines cannot use build.getEnvironment(listener) and I need to figure a way for pipelines to access environment variables.
andytaylorzeetta as by the PR mentioned to fix https://issues.jenkins.io/browse/JENKINS-62753, you need to separately make your pipeline job see your environment variables. If you add something like this to your pipeline:
public class PublishEnvVarAction extends InvisibleAction implements EnvironmentContributingAction { private final Map<String, String> envOverrides; public PublishEnvVarAction(final Map<String, String> env) { this.envOverrides = env; } @NonCPS @Override public void buildEnvironment(Run<?, ?> run, EnvVars env) { if (env != null && envOverrides != null){ env.putAllNonNull(envOverrides); } } } def publishEnv(Map config) { currentBuild.getRawBuild().addAction(new PublishEnvVarAction(config.collectEntries{ k, v -> [(k.toString()) : v.toString()] })) }
you can then call it from the pipeline like this
pipeline {
agent { label "my_label" }
stages {
stage("Do stuff") {
steps {
script {
myVars = [:]
myVars["NODE_NAME"] = NODE_NAME
publishEnv(myVars)
}
}
}
}
}
This makes the global listener find the NODE_NAME environment variable and the field shows correctly.
Quick update: I've noticed that build_agent_name is correctly set to 'master' when a job is running on the master, but if a job runs on any of the slaves it is null.