I am trying to setup a Jenkins declarative pipeline to use two different agents during its execution. The agents are dynamically spawned by the Kubernetes plugin. For sake of argument and simplicity, let's assume I want to do this:
On Agent 1 (Cloud name: "ubuntu"):
On Agent 2 (Cloud name: "fedora"):
The problem I have is that if if I use a global agent declaration:
pipeline {
agent {
kubernetes {
cloud 'ubuntu'
label "ubuntu-agent"
containerTemplate {
name 'support'
image 'blenderfox/support'
ttyEnabled true
command 'cat'
}
}
}
...
}
Then that is used across all the stages if I don't declare an agent on each of the stages.
If I use agent none:
pipeline {
agent none
...
}
Then I have to declare an agent spec for each stage, for example:
stage ("apt update") {
agent {
kubernetes {
cloud 'ubuntu'
label "ubuntu-agent"
containerTemplate {
name 'support'
image 'blenderfox/support'
ttyEnabled true
command 'cat'
}
}
}
steps {
sh """
apt update
"""
}
}
While this would work for me in that I can declare per stage which agent I want, the problem this method causes, is that it spins up a new agent for each stage, meaning the state isn't carried between, for example, these two stages:
stage ("apt-update") {
agent {
....
}
steps {
sh """
apt update
"""
}
}
stage ("apt-install") {
agent {
....
}
steps {
sh """
apt install -y ....
"""
}
}
Can I reuse the same agent across stages? For example, something like this:
stage ("provision agent") {
agent {
...
label "ubuntu-agent"
...
}
steps {
sh """
echo "Provisioning agent"
"""
}
}
stage ("apt-update") {
agent {
label "ubuntu-agent" //reuse agent from previous stage
}
steps {
sh """
apt update
"""
}
}
stage ("apt-install") {
agent {
label "ubuntu-agent" //reuse agent from previous stage
}
steps {
sh """
apt install -y ....
"""
}
}
Found a solution. Very hacky but it works:
pipeline {
agent none
stages {
stage ("Provision dev agent") {
agent {
kubernetes {
cloud 'dev-cloud'
label "dev-agent-${env.BUILD_NUMBER}"
slaveConnectTimeout 300
idleMinutes 5
yamlFile "jenkins-dev-agent.yaml"
}
}
steps {
sh """
## Do any agent init steps here
"""
}
}
stage ("Do something on dev agent") {
agent {
kubernetes {
label "dev-agent-${env.BUILD_NUMBER}"
}
}
steps {
sh """
## Do something here
"""
}
}
stage ("Provision production agent") {
agent {
kubernetes {
cloud 'prod-cloud'
label "prod-agent-${env.BUILD_NUMBER}"
slaveConnectTimeout 300
idleMinutes 5
yamlFile "jenkins-prod-agent.yaml"
}
}
steps {
sh """
## Do any agent init steps here
"""
}
}
stage ("Do something on prod agent") {
agent {
kubernetes {
label "prod-agent-${env.BUILD_NUMBER}"
}
}
steps {
sh """
## Do something here
"""
}
}
}
}
The agent yamls vary, but you can do something like this:
spec:
containers:
- name: docker
image: docker:18.06.1
command: ["tail", "-f", "/dev/null"]
imagePullPolicy: Always
volumeMounts:
- name: docker
mountPath: /var/run/docker.sock
volumes:
- hostPath:
path: "/var/run/docker.sock"
name: "docker"And then use the agent like so:
stage ("docker build") {
agent {
kubernetes {
label "dev-agent-${env.BUILD_NUMBER}"
}
}
steps {
container('docker') {
sh """
## docker build....
"""
}
}
}