Report failed stage name not working with parallel stages

[elavaud]

Hi there,

Reporting failed stage in the post section is quite a handy trick. Thanks for that! However I struggle to make it work in a more complex scenario using parallel stages. Let's consider this:

pipeline {
    agent none // Using 2 different slaves
    stages {
        stage('For parallel stuff') {
            failFast true
            parallel {
                stage('Slave A with sequential stages') {
                    agent  { node { label "os-linux" } }
                    stages {
                        stage ('Stuff 1 on slave A') {
                            steps {
                                script {
                                    last_started = env.STAGE_NAME
                                    // stuff
                                }
                            }
                        }
                        stage ('Stuff 2 on slave A') { ... }
                        stage ('Stuff 3 on slave A') { ... }
                    }
                }
                stage('Slave B with sequential stages') {
                    agent  { node { label "os-windows" } }
                    stages {
                        stage ('Stuff 1 on slave B') {
                            steps {
                                script {
                                    last_started = env.STAGE_NAME
                                    // stuff
                                }
                            }
                        }
                        stage ('Stuff 2 on slave B') { ... }
                        stage ('Stuff 3 on slave B') { ... }
                    }
                }
            }
        }
    }
    post { // Using STAGE_NAME variable }
}

If, for example, "Stuff 2 on slave B" started after "Stuff 1 on slave A", although that the later one fails, the first one is reported.

Then people throw rocks at you beause you reported the wrong failed stage ... :(

Some ideas on how to handle that? How come Jenkins pipeline UI or Blue Ocean manage to get the right one that failed?

Cheers,

Edouard

[elavaud]

Clearly not as handy / beautiful, but I solved my issue assigning the variable "last_started" in a post block for the stage:

pipeline {
    agent none // Using 2 different slaves
    stages {
        stage('For parallel stuff') {
            failFast true
            parallel {
                stage('Slave A with sequential stages') {
                    agent  { node { label "os-linux" } }
                    stages {
                        stage ('Stuff 1 on slave A') {
                            steps {
                                // stuff
                            }
                            post {
                                failure {
                                    last_started = env.STAGE_NAME
                                }
                                unstable { ... } // Adding this one on stages that could be unstable
                            }
                        }
                        ...
                    }
                }
                ....
            }
        }
    }
    post { // Using STAGE_NAME variable }
}