I am using jenkins-job-builder to create my pipeline project. But I have a problem with the variables values when I am trying to reuse or propagating.
It is my project configuration:
- project:
name: myproject
git_url: git#gitlabserver.cu:demos-products/myproject.git
jobs:
- '{name}-nfr-smoke-tests':
pipeline-next: '{name}-nfr-smoke-tests'
And here is my job-template:
- job-template:
name: "{name}-nfr-smoke-tests"
node: 'slave1'
scm:
- git:
skip-tag: false
url: 'git#gitlabserver.cu:test-products/{name}-nfr-tests.git'
branches:
- master
wipe-workspace: true
builders:
- shell: |
bundle install
bundle exec cucumber features/smoke.feature
publishers:
- trigger:
project: "{pipeline-next}"
threshold: SUCCESS
Ok, now when I run this configuration in jenkins and I check the job's construction, it says:
No such project ‘{name}-nfr-smoke-tests’. Did you mean ‘myproject-nfr-smoke-tests’?
Why the line: pipeline-next: '{name}-nfr-smoke-tests'doesn't propagates the value of variable name and just used it as a literal string? I am missing something.
You are missing 'name' under 'project' section in your job-template. Append the following lines:
- project:
name: project-name
The purpose of a project is to collect related jobs together, and provide values for the variables in a Job Template.
I found out that Jenkins Job Builder version 0.9.0-0.2 does not propagate the value, but for me version 1.3.0+2015.12.15.git136.959eb4b909-1 did. Perhaps updating your version of Jenkins Job Builder might help?
Related
Recently I've made some configuration on my team's github circleci. I needed to use a when statement to devide ci logics. I referenced this document(https://circleci.com/docs/2.0/configuration-reference/#logic-statements) but it seems the document not correct.
Below is my step definition:
...
image_build_step:
executor: golang_executor
steps:
- checkout
- setup_remote_docker:
version: 18.09.3
docker_layer_caching: true
- define_svc_name:
jobname: ${CIRCLE_JOB} # On this step set $SVC variable
- when:
conditon:
equal: ["${SVC}", "SVC_A" ]
- aws-ecr/build-and-push-image:
repo: SVC_A_REPO
dockerfile: ./Dockerfile
tag: "latest,${CIRCLE_SHA1},build-${CIRCLE_BUILD_NUM}"
...
Also I already tried this.
...
image_build_step:
executor: golang_executor
steps:
- checkout
- setup_remote_docker:
version: 18.09.3
docker_layer_caching: true
- define_svc_name:
jobname: ${CIRCLE_JOB} # On this step set $SVC variable
- when:
equal: ["${SVC}", "SVC_A" ]
- aws-ecr/build-and-push-image:
repo: SVC_A_REPO
dockerfile: ./Dockerfile
tag: "latest,${CIRCLE_SHA1},build-${CIRCLE_BUILD_NUM}"
...
I cannot figure out my mistake using when statement on circleci. Additionaly, I already passed circleci config validate .circleci/config.yaml command before I pushed this commit.
What is the correct usage of when statement in circleci? Joining circleci forum is also annoying me using github account, so I leave my question on stakeoverflow.
It's not possible to use environment variables in logic statements. The reason is that logic statements are evaluated at configuration compilation time, whereas environment variables are interpolated at run time.
The only workaround I know of is to use the CircleCI dynamic configuration functionality to set pipeline parameters' values in the "setup workflow" that you then pass to the "continuation" workflow.
Jenkins has a UI concept with dropdown lists, etc. to allow users to specify variables at run time. This has proven essential in our builds to make decisions in the pipeline (ie. which agent to run on, code base to choose, etc). By allowing parameters we are able to have a single pipeline/definition handle the same task for many clients/releases/environments.
I have been watching many people over the past year ask for this - to eliminate the number of almost identical build definitions - is there a best practice to handle this? Would be nice to have a single build definition for a specific task that can be smart enough to handle parameters.
Edit : example of possible pseudo-code to build on levi-lu#MSFT's suggestion.
parameters:
- name: ClientName
displayName: Pool Image
default: Select client
values: powershell
valuesScript : [
assemble curl request to http://myUrl.com/Clients/GetAll
]
- name: TargetEnvironment
displayName: Client Environment
type: string
values: powershell
valuesScript: [
assemble curl request using above parameter value to
https://myUrl.com/Clients/$(ClientName)/GetEnvironments
]
trigger: none
jobs:
- job: build
displayName: Run pipeline job
pool:
vmImage: windows-latest
parameters:
ClientName : $(ClientName)
TargetEnvironment : $(TargetEnvironment)
steps:
- script: echo building $(Build.BuildNumber)
Runtime parameters is available now. You can now set runtime parameters at the beginning of your pipepline YAML using parameters. For below example:
parameters:
- name: image
displayName: Pool Image
default: ubuntu-latest
values:
- windows-latest
- vs2017-win2016
- ubuntu-latest
- ubuntu-16.04
- macOS-latest
- macOS-10.14
- name: test
displayName: Run Tests?
type: boolean
default: false
trigger: none
jobs:
- job: build
displayName: Build and Test
pool:
vmImage: ${{ parameters.image }}
steps:
- script: echo building $(Build.BuildNumber)
- ${{ if eq(parameters.test, true) }}:
- script: echo "Running all the tests"
Above example is from Microsoft official document. Click here to learn more about runtime parameters.
When you run above Yaml pipeline, You will be able to select the parameter's value from the dropdown list. See below screenshot.
Update: To set variables dynamically at runtime.
You can use the task.setvariable logging command to set variables dynamically in scripts.
For below example: $resultValue is the value from rest api call. And its value is assigned to variable VariableName
- powershell: |
$resultValue = call from Rest API
echo "##vso[task.setvariable variable=VariableName]$resultValue"
Check document here for more information.
I have a Jenkins JJB template, where I would like to customize triggers per job instance. This approach would have worked great if triggers was a string parameter, but it requires an object. How can I pass an object (in this case - an array) to the job template?
- job-template:
name: 'my-template-{proj}'
trigger_overrides: []
# The obj: syntax should have worked,
# but it doesn't handle the default from above
triggers: '{obj:trigger_overrides}'
- project:
name: my-project
jobs:
# for this job, customize triggers
- 'my-template':
proj: A
trigger_overrides:
- github
- timed: "#daily"
# this job should use the default triggers
- 'my-template':
proj: B
Note that in case triggers is always passed in, the empty triggers doesn't work. The global default overrides template parameter:
- defaults:
name: global
triggers:
- github
- job-template:
name: 'my-template-{proj}'
triggers: '{obj:trigger_overrides}'
- project:
name: my-project
jobs:
- 'my-template':
proj: A
trigger_overrides: []
Neither does this work (empty array is given as a default):
triggers: '{obj:trigger_overrides|[]}
P.S. This issue was also filed with jjb devs
You want to refer to how gerrit_release_trigger_file_paths is handled here in the releng/builder repository for passing a list:
https://github.com/opendaylight/releng-builder/blob/master/jjb/lf-infra/lf-infra-jobs.yaml#L120-L130
The original template is implemented in releng-global-jjb repository here: https://github.com/lfit/releng-global-jjb/blob/master/jjb/lf-docker-jobs.yaml#L229
JJB Best Practices:
https://docs.releng.linuxfoundation.org/projects/global-jjb/en/latest/best-practices.html
- project:
name: my-project
jobs:
- 'my-template':
proj: A
trigger_overrides:
- trigger_a
- trigger_b
- trigger_c
Within JJB, you can define project-level variables like this:
- defaults:
name: global
git_url: "git#....."
- project
name: some-test
jobs:
- test-{name}
- job-template
name: test-{name}
scm:
- git:
url: "{git_url}"
branches:
- master
My question, must I hardcode the value of git_url at the default level or can I use some JJB mechanism to bring that in at job load/execution?
The reason I ask is that the yaml script that contains these JJB jobs can be used to define TEST, QA and PROD. It would be nice to just point at a properties file that contains the value for git_url and any other global variable values. I took a look at: http://docs.openstack.org/infra/jenkins-job-builder/definition.html?highlight=default#defaults and I did not see any mechanism.
If I understand your question correctly, there are two other approaches available within the context of a single yaml file
Approach 1: Set git_url at the project level
- project
name: some-test
git_url: "git#dogs.net:woof/bark.git"
jobs:
- test-{name}:
- job-template
name: test-{name}
scm:
- git:
url: "{git_url}"
branches:
- master
Here git_url is set at the project level. This approach allows you to define a second project with a different value for git_url, ie
- project
name: some-other-test
git_url: "git#cats.net:meow/meow.git"
jobs:
- test-{name}:
Approach 2: Set git_url at the job-template instance level
- project
name: some-test
jobs:
- test-{name}:
git_url: "git#....."
- job-template
name: test-{name}
scm:
- git:
url: "{git_url}"
branches:
- master
Here git_url is set on the actual instance of the job-template where it is specified. If your job-template had more than just {name} in its name, this would allow you to create multiple instances of it in the list of jobs at the project level, ie
- project
name: some-test
git_url: "git#....."
jobs:
- test-{name}-{type}:
type: 'cat'
- test-{name}-{type}:
type: 'dog'
- job-template
name: test-{name}-{type}
display-name: 'Test for {type} projects'
scm:
- git:
url: "{git_url}"
branches:
- master
Thoughts on TEST vs QA vs PROD
You also mentioned that you would like some kind of external properties file to differentiate between TEST, QA, and PROD environments. To address this let's consider four different files, project.yaml, defaults/TEST.yaml, defaults/QA.yaml, defaults/PROD.yaml whose contents are enumerated below.
project.yaml
- project
name: some-test
jobs:
- test-{name}:
defaults/TEST.yaml
- defaults:
name: global
git_url: "git#dogs.net:woof/test.git"
defaults/QA.yaml
- defaults:
name: global
git_url: "git#dogs.net:woof/qa.git"
defaults/PROD.yaml
- defaults:
name: global
git_url: "git#dogs.net:woof/prod.git"
Okay so these aren't great examples because you probably wouldn't have a different git repository for each environment, but I don't want to complicate things by straying too far from your original example.
With JJB you can specify more than one YAML file on the command line (I don't want to complicate the example or its explanation, but you can also specify directories full of JJB yaml). To differentiate between TEST, QA, and PROD deployments of your Jenkins job you can then do something like:
jenkins-jobs project.yaml:defaults/TEST.yaml
For your test environment.
jenkins-jobs project.yaml:defaults/QA.yaml
For your qa environment.
jenkins-jobs project.yaml:defaults/PROD.yaml
For your prod environment.
Hope that helps.
I am investigating using Jenkins Job-Builder (from OpenStack) as our means of managing jenkins job configurations. In doing so I am trying to figure out the right (best?) way to include a job-template from an external file using the !include custom tag.
In the current use case we will basically have one template that is going to be used by a LOT of job. Each job is going to need to exist in its own file for reason that are out of scope here.
So far I have gotten this to work:
job-template.yml
name: 'pre-build-{proj}-{repo}'
project-type: freestyle
... etc ...
job-1.yml
- job-template:
!include job-template.yml
- project:
name: job-1
proj: my-proj
repo: my-repo
jobs:
- 'build-config-{proj}-{repo}'
This seem wrong because the template definition gets split across both files and require needless duplication of the -job-template: line in every job file. I would like to get the following to work instead:
job-template.yml
- job-template:
name: 'pre-build-{proj}-{repo}'
project-type: freestyle
... etc ...
job-1.yml
!include job-template.yml
- project:
name: job-1
proj: my-proj
repo: my-repo
jobs:
- 'build-config-{proj}-{repo}'
The latter unfortunately results in a yaml parse error on the - project: line:
yaml.scanner.ScannerError: mapping values are not allowed here
in "job-1.yml", line 3, column 10
Is there way to get the entire template definition into the template file? This will become particularly annoying if ever we need to pull in multiple templates from multiple files.
Jenkins-jobs takes a path argument which can be a directory holding your files (job-template.yaml, job-1.yaml and job-2.yaml. It will assemble them as a single YAML document, so you do not need to use !include. So you can write:
job-template.yaml
- job-template:
name: 'pre-build-{proj}-{repo}'
builders:
- shell: 'echo building {proj} for {repo}'
job1.yaml
- project:
name: job-1
proj: my-proj
repo: my-repo
jobs:
- 'pre-build-{proj}-{repo}'
job2.yaml
- project:
name: job-2
proj: my-other-proj
repo: my-other-repo
jobs:
- 'pre-build-{proj}-{repo}'
That will generates two jobs with the following shell commands:
pre-build-my-other-proj-my-other-repo:
<command>echo building my-other-proj for my-other-repo</command>
pre-build-my-proj-my-repo:
<command>echo building my-proj for my-repo</command>
Assuming the files are in a directory config/ you can generate them all with:
jenkins-jobs test config/ -o /tmp/myjobs
Or use the name argument to filter the jobs that will be realized:
jenkins-jobs test config/ -o /tmp/myjobs '*my-proj*'
# Creates pre-build-my-proj-my-repo
# Skips pre-build-my-other-proj-my-other-repo