Job Content

Zuul jobs are implemented as Ansible playbooks. Zuul prepares the repositories used for a job, installs any required Ansible roles, and then executes the job’s playbooks. Any setup or artifact collection required is the responsibility of the job itself. While this flexible arrangement allows for almost any kind of job to be run by Zuul, batteries are included. Zuul has a standard library of jobs upon which to build.

Working Directory

Before starting each job, the Zuul executor creates a directory to hold all of the content related to the job. This includes some directories which are used by Zuul to configure and run Ansible and may not be accessible, as well as a directory tree, under work/, that is readable and writable by the job. The hierarchy is:

work/: The working directory of the job.
work/src/: Contains the prepared git repositories for the job.
work/logs/: Where the Ansible log for the job is written; your job may place other logs here as well.

Git Repositories

The git repositories in work/src contain the repositories for all of the projects specified in the required-projects section of the job, plus the project associated with the queue item if it isn’t already in that list. In the case of a proposed change, that change and all of the changes ahead of it in the pipeline queue will already be merged into their respective repositories and target branches. The change’s project will have the change’s branch checked out, as will all of the other projects, if that branch exists (otherwise, a fallback or default branch will be used). If your job needs to operate on multiple branches, simply checkout the appropriate branches of these git repos to ensure that the job results reflect the proposed future state that Zuul is testing, and all dependencies are present.

The git repositories will have a remote origin with refs pointing to the previous change in the speculative state. This means that e.g. a git diff origin/<branch>..<branch> will show the changes being tested. Note that the origin URL is set to a bogus value (file:///dev/null) and can not be used for updating the repository state; the local repositories are guaranteed to be up to date.

The repositories will be placed on the filesystem in directories corresponding with the canonical hostname of their source connection. For example:

work/src/git.example.com/project1
work/src/github.com/project2

Is the layout that would be present for a job which included project1 from the connection associated to git.example.com and project2 from GitHub. This helps avoid collisions between projects with the same name, and some language environments, such as Go, expect repositories in this format.

Note that these git repositories are located on the executor; in order to be useful to most kinds of jobs, they will need to be present on the test nodes. The base job in the standard library (see zuul-base-jobs documentation for details) contains a pre-playbook which copies the repositories to all of the job’s nodes. It is recommended to always inherit from this base job to ensure that behavior.

Variables

There are several sources of variables which are available to Ansible: variables defined in jobs, secrets, and site-wide variables. The order of precedence is:

Site-wide variables
Job extra variables
Secrets
Job variables
Project variables
Parent job results

Meaning that a site-wide variable with the same name as any other will override its value, and similarly, secrets override job variables of the same name which override data returned from parent jobs. Each of the sources is described below.

Site-wide Variables

The Zuul administrator may define variables which will be available to all jobs running in the system. These are statically defined and may not be altered by jobs. See the Administrator’s Guide for information on how a site administrator may define these variables.

Job Extra Variables

Any extra variables in the job definition (using the job.extra-vars attribute) are available to Ansible but not added into the inventory file.

Secrets

Secrets also appear as variables available to Ansible. Unlike job variables, these are not added to the inventory file (so that the inventory file may be kept for debugging purposes without revealing secrets). But they are still available to Ansible as normal variables. Because secrets are groups of variables, they will appear as a dictionary structure in templates, with the dictionary itself being the name of the secret, and its members the individual items in the secret. For example, a secret defined as:

- secret:
    name: credentials
    data:
      username: foo
      password: bar

Might be used in a template as:

{{ credentials.username }} {{ credentials.password }}

Secrets are only available to playbooks associated with the job definition which uses the secret; they are not available to playbooks associated with child jobs or job variants.

Job Variables

Any variables specified in the job definition (using the job.vars attribute) are available as Ansible host variables. They are added to the vars section of the inventory file under the all hosts group, so they are available to all hosts. Simply refer to them by the name specified in the job’s vars section.

Project Variables

Any variables specified in the project definition (using the project.vars attribute) are available to jobs as Ansible host variables in the same way as job variables. Variables set in a project-template are merged into the project variables when the template is included by a project.

- project-template:
    name: sample-template
    description: Description
    vars:
      var_from_template: foo
    post:
      jobs:
        - template_job
    release:
      jobs:
        - template_job

- project:
    name: Sample project
    description: Description
    templates:
      - sample-template
    vars:
      var_for_all_jobs: value
    check:
      jobs:
        - job1
        - job2:
            vars:
              var_for_all_jobs: override

Parent Job Results

A job may return data to Zuul for later use by jobs which depend on it. For details, see Return Values.

Zuul Variables

Zuul supplies not only the variables specified by the job definition to Ansible, but also some variables from Zuul itself.

When a pipeline is triggered by an action, it enqueues items which may vary based on the pipeline’s configuration. For example, when a new change is created, that change may be enqueued into the pipeline, while a tag may be enqueued into the pipeline when it is pushed.

Information about these items is available to jobs. All of the items enqueued in a pipeline are git references, and therefore share some attributes in common. But other attributes may vary based on the type of item.

zuul

All items provide the following information as Ansible variables under the zuul key:

zuul.artifacts Type: list

If the job has a job.requires attribute, and Zuul has found changes ahead of this change in the pipeline with matching job.provides attributes, then information about any artifacts returned from those jobs will appear here.

This value is a list of dictionaries with the following format:

zuul.artifacts[].project: The name of the project which supplied this artifact.

zuul.artifacts[].change: The change number which supplied this artifact.

zuul.artifacts[].patchset: The patchset of the change.

zuul.artifacts[].job: The name of the job which produced the artifact.

zuul.artifacts[].name: The name of the artifact (as supplied to Returning artifact URLs).

zuul.artifacts[].url: The URL of the artifact (as supplied to Returning artifact URLs).

zuul.artifacts[].metadata: The metadata of the artifact (as supplied to Returning artifact URLs).

zuul.build: The UUID of the build. A build is a single execution of a job. When an item is enqueued into a pipeline, this usually results in one build of each job configured for that item’s project. However, items may be re-enqueued in which case another build may run. In dependent pipelines, the same job may run multiple times for the same item as circumstances change ahead in the queue. Each time a job is run, for whatever reason, it is acompanied with a new unique id.

zuul.buildset: The build set UUID. When Zuul runs jobs for an item, the collection of those jobs is known as a buildset. If the configuration of items ahead in a dependent pipeline changes, Zuul creates a new buildset and restarts all of the jobs.

zuul.child_jobs: A list of the first level dependent jobs to be run after this job has finished successfully.

zuul.ref: The git ref of the item. This will be the full path (e.g., refs/heads/master or refs/changes/…).

zuul.override_checkout: If the job was configured to override the branch or tag checked out, this will contain the specified value. Otherwise, this variable will be undefined.

zuul.pipeline: The name of the pipeline in which the job is being run.

zuul.post_review Type: bool: Whether the current job is running in a post-review pipeline or not.

zuul.job: The name of the job being run.

zuul.event_id: The UUID of the event that triggered this execution. This is mainly useful for debugging purposes.

zuul.voting: A boolean indicating whether the job is voting.

zuul.attempts: An integer count of how many attempts have been made to run this job for the current buildset. If there are pre-run failures or network connectivity issues then previous attempts may have been cancelled, and this value will be greater than 1.

zuul.project

The item’s project. This is a data structure with the following fields:

zuul.project.name: The name of the project, excluding hostname. E.g., org/project.

zuul.project.short_name: The name of the project, excluding directories or organizations. E.g., project.

zuul.project.canonical_hostname: The canonical hostname where the project lives. E.g., git.example.com.

zuul.project.canonical_name: The full canonical name of the project including hostname. E.g., git.example.com/org/project.

zuul.project.src_dir: The path to the source code relative to the work dir. E.g., src/git.example.com/org/project.

zuul.projects Type: dict

A dictionary of all projects prepared by Zuul for the item. It includes, at least, the item’s own project. It also includes the projects of any items this item depends on, as well as the projects that appear in job.required-projects.

This is a dictionary of dictionaries. Each value has a key of the canonical_name, then each entry consists of:

zuul.projects{}.name: The name of the project, excluding hostname. E.g., org/project.

zuul.projects{}.short_name: The name of the project, excluding directories or organizations. E.g., project.

zuul.projects{}.canonical_hostname: The canonical hostname where the project lives. E.g., git.example.com.

zuul.projects{}.canonical_name: The full canonical name of the project including hostname. E.g., git.example.com/org/project.

zuul.projects{}.src_dir: The path to the source code, relative to the work dir. E.g., src/git.example.com/org/project.

zuul.projects{}.required: A boolean indicating whether this project appears in the job.required-projects list for this job.

zuul.projects{}.checkout: The branch or tag that Zuul checked out for this project. This may be influenced by the branch or tag associated with the item as well as the job configuration.

For example, to access the source directory of a single known project, you might use:

{{ zuul.projects['git.example.com/org/project'].src_dir }}

To iterate over the project list, you might write a task something like:

- name: Sample project iteration
  debug:
    msg: "Project {{ item.name }} is at {{ item.src_dir }}
  with_items: {{ zuul.projects.values() | list }}

zuul.tenant: The name of the current Zuul tenant.

zuul.timeout: The job timeout, in seconds.

zuul.post_timeout: The post-run playbook timeout, in seconds.

zuul.jobtags: A list of tags associated with the job. Not to be confused with git tags, these are simply free-form text fields that can be used by the job for reporting or classification purposes.

zuul.items Type: list

Note

zuul.items conflicts with the items() builtin so the variable can only be accessed with python dictionary like syntax, e.g: zuul['items']

A list of dictionaries, each representing an item being tested with this change with the format:

zuul.items[].project

The item’s project. This is a data structure with the following fields:

zuul.items[].project.name: The name of the project, excluding hostname. E.g., org/project.

zuul.items[].project.short_name: The name of the project, excluding directories or organizations. E.g., project.

zuul.items[].project.canonical_hostname: The canonical hostname where the project lives. E.g., git.example.com.

zuul.items[].project.canonical_name: The full canonical name of the project including hostname. E.g., git.example.com/org/project.

zuul.items[].project.src_dir: The path to the source code on the remote host, relative to the home dir of the remote user. E.g., src/git.example.com/org/project.

zuul.items[].branch: The target branch of the change (without the refs/heads/ prefix).

zuul.items[].change: The identifier for the change.

zuul.items[].change_url: The URL to the source location of the given change. E.g., https://review.example.org/#/c/123456/ or https://github.com/example/example/pull/1234.

zuul.items[].patchset: The patchset identifier for the change. If a change is revised, this will have a different value.

zuul.items[].resources

zuul.items[].:type: dict

A job using a container build resources has access to a resources variable that describes the resource. Resources is a dictionary of group keys, each value consists of:

zuul.items[].:type: dict.namespace: The resource’s namespace name.

zuul.items[].:type: dict.context: The kube config context name.

zuul.items[].:type: dict.pod: The name of the pod when the label defines a kubectl connection.

Project or namespace resources might be used in a template as:

- hosts: localhost
    tasks:
    - name: Create a k8s resource
      k8s_raw:
        state: present
        context: "{{ zuul.resources['node-name'].context }}"
        namespace: "{{ zuul.resources['node-name'].namespace }}"

Kubectl resources might be used in a template as:

- hosts: localhost
    tasks:
    - name: Copy src repos to the pod
      command: >
        oc rsync -q --progress=false
            {{ zuul.executor.src_root }}/
            {{ zuul.resources['node-name'].pod }}:src/
        no_log: true

zuul_success

Post run playbook(s) will be passed this variable to indicate if the run phase of the job was successful or not. This variable is meant to be used with the bool filter.

tasks:
  - shell: echo example
    when: zuul_success | bool

Change Items

A change to the repository. Most often, this will be a git reference which has not yet been merged into the repository (e.g., a gerrit change or a GitHub pull request). The following additional variables are available:

zuul.branch: The target branch of the change (without the refs/heads/ prefix).

zuul.change: The identifier for the change.

zuul.patchset: The patchset identifier for the change. If a change is revised, this will have a different value.

zuul.change_url: The URL to the source location of the given change. E.g., https://review.example.org/#/c/123456/ or https://github.com/example/example/pull/1234.

zuul.message

The commit or pull request message of the change base64 encoded. Use the b64decode filter in ansible when working with it.

- hosts: all
  tasks:
    - name: Dump commit message
      copy:
        content: "{{ zuul.message | b64decode }}"
        dest: "{{ zuul.executor.log_root }}/commit-message.txt"

Branch Items

This represents a branch tip. This item may have been enqueued because the branch was updated (via a change having merged, or a direct push). Or it may have been enqueued by a timer for the purpose of verifying the current condition of the branch. The following additional variables are available:

zuul.branch: The name of the item’s branch (without the refs/heads/ prefix).

zuul.oldrev: If the item was enqueued as the result of a change merging or being pushed to the branch, the git sha of the old revision will be included here. Otherwise, this variable will be undefined.

zuul.newrev: If the item was enqueued as the result of a change merging or being pushed to the branch, the git sha of the new revision will be included here. Otherwise, this variable will be undefined.

Tag Items

This represents a git tag. The item may have been enqueued because a tag was created or deleted. The following additional variables are available:

zuul.tag: The name of the item’s tag (without the refs/tags/ prefix).

zuul.oldrev: If the item was enqueued as the result of a tag being deleted, the previous git sha of the tag will be included here. If the tag was created, this variable will be undefined.

zuul.newrev: If the item was enqueued as the result of a tag being created, the new git sha of the tag will be included here. If the tag was deleted, this variable will be undefined.

Ref Items

This represents a git reference that is neither a change, branch, or tag. Note that all items include a ref attribute which may be used to identify the ref. The following additional variables are available:

zuul.oldrev: If the item was enqueued as the result of a ref being deleted, the previous git sha of the ref will be included here. If the ref was created, this variable will be undefined.

zuul.newrev: If the item was enqueued as the result of a ref being created, the new git sha of the ref will be included here. If the ref was deleted, this variable will be undefined.

Working Directory

Additionally, some information about the working directory and the executor running the job is available:

zuul.executor

A number of values related to the executor running the job are available:

zuul.executor.hostname: The hostname of the executor.

zuul.executor.src_root: The path to the source directory.

zuul.executor.log_root: The path to the logs directory.

zuul.executor.work_root: The path to the working directory.

zuul.executor.inventory_file

The path to the inventory. This variable is needed for jobs running without a nodeset since Ansible doesn’t set it for localhost; see this porting guide.

The inventory file is only readable by jobs running in a trusted execution context.

SSH Keys

Zuul starts each job with an SSH agent running and at least one key added to that agent. Generally you won’t need to be aware of this since Ansible will use this when performing any tasks on remote nodes. However, under some circumstances you may want to interact with the agent. For example, you may wish to add a key provided as a secret to the job in order to access a specific host, or you may want to, in a pre-playbook, replace the key used to log into the assigned nodes in order to further protect it from being abused by untrusted job content.

A description of each of the keys added to the SSH agent follows.

Nodepool Key

This key is supplied by the system administrator. It is expected to be accepted by every node supplied by Nodepool and is generally the key that will be used by Zuul when running jobs. Because of the potential for an unrelated job to add an arbitrary host to the Ansible inventory which might accept this key (e.g., a node for another job, or a static host), the use of the add-build-sshkey <https://zuul-ci.org/docs/zuul-jobs/roles.html#role-add-build-sshkey> role is recommended.

Project Key

Each project in Zuul has its own SSH keypair. This key is added to the SSH agent for all jobs running in a post-review pipeline. If a system administrator trusts that project, they can add the project’s public key to systems to allow post-review jobs to access those systems. The systems may be added to the inventory using the add_host Ansible module, or they may be supplied by static nodes in Nodepool.

Zuul serves each project’s public SSH key using its build-in webserver. They can be fetched at the path /api/tenant/<tenant>/project-ssh-key/<project>.pub where <project> is the canonical name of a project and <tenant> is the name of a tenant with that project.

Return Values

A job may return some values to Zuul to affect its behavior and for use by dependent jobs. To return a value, use the zuul_return Ansible module in a job playbook. For example:

tasks:
  - zuul_return:
      data:
        foo: bar

Will return the dictionary {'foo': 'bar'} to Zuul.

Optionally, if you have a large supply of data to return, you may specify the path to a JSON-formatted file with that data. For example:

tasks:
  - zuul_return:
      file: /path/to/data.json

Normally returned data are provided to dependent jobs in the inventory file, which may end up in the log archive of a job. In the case where sensitive data must be provided to dependent jobs, the secret_data attribute may be used instead, and the data will be provided via the same mechanism as job secrets, where the data are not written to disk in the work directory. Care must still be taken to avoid displaying or storing sensitive data within the job. For example:

tasks:
  - zuul_return:
      secret_data:
        password: foobar

Any values other than those in the zuul hierarchy will be supplied as Ansible variables to dependent jobs. These variables have less precedence than any other type of variable in Zuul, so be sure their names are not shared by any job variables. If more than one parent job returns the same variable, the value from the later job in the job graph will take precedence.

The values in the zuul hierarchy are special variables that influence the behavior of zuul itself. The following paragraphs describe the currently supported special variables and their meaning.

Returning the log url

To set the log URL for a build, use zuul_return to set the zuul.log_url value. For example:

tasks:
  - zuul_return:
      data:
        zuul:
          log_url: http://logs.example.com/path/to/build/logs

Returning artifact URLs

If a build produces artifacts, any number of URLs may be returned to Zuul and stored in the SQL database. These will then be available via the web interface and subsequent jobs.

To provide artifact URLs for a build, use zuul_return to set keys under the zuul.artifacts dictionary. For example:

tasks:
  - zuul_return:
      data:
        zuul:
          artifacts:
            - name: tarball
              url: http://example.com/path/to/package.tar.gz
              metadata:
                version: 3.0
            - name: docs
              url: build/docs/

If the value of url is a relative URL, it will be combined with the zuul.log_url value if set to create an absolute URL. The metadata key is optional; if it is provided, it must be a dictionary; its keys and values may be anything.

If zuul_return is invoked multiple times (e.g., via multiple playbooks), then the elements of zuul.artifacts from each invocation will be appended.

Skipping dependent jobs

Note

In the following section the use of ‘child jobs’ refers to dependent jobs configured by job.dependencies and should not be confused with jobs that inherit from a parent job.

To skip a dependent job for the current build, use zuul_return to set the zuul.child_jobs value. For example:

tasks:
  - zuul_return:
      data:
        zuul:
          child_jobs:
            - dependent_jobA
            - dependent_jobC

Will tell zuul to only run the dependent_jobA and dependent_jobC for pre-configured dependent jobs. If dependent_jobB was configured, it would be now marked as SKIPPED. If zuul.child_jobs is empty, all jobs will be marked as SKIPPED. Invalid dependent jobs are stripped and ignored, if only invalid jobs are listed it is the same as providing an empty list to zuul.child_jobs.

Leaving warnings

A job can leave warnings that will be appended to the comment zuul leaves on the change. Use zuul_return to add a list of warnings. For example:

tasks:
  - zuul_return:
      data:
        zuul:
          warnings:
            - This warning will be posted on the change.

If zuul_return is invoked multiple times (e.g., via multiple playbooks), then the elements of zuul.warnings from each invocation will be appended.

Leaving file comments

To instruct the reporters to leave line comments on files in the change, set the zuul.file_comments value. For example:

tasks:
  - zuul_return:
      data:
        zuul:
          file_comments:
            path/to/file.py:
              - line: 42
                message: "Line too long"
                level: info
              - line: 82
                message: "Line too short"
              - line: 119
                message: "This block is indented too far."
                level: warning
                range:
                  start_line: 117
                  start_character:    0
                  end_line:   119
                  end_character:  37

Not all reporters currently support line comments (or all of the features of line comments); in these cases, reporters will simply ignore this data. The level is optional, but if provided must be one of info, warning, error.

Zuul will attempt to automatically translate the supplied line numbers to the corresponding lines in the original change as written (they may differ due to other changes which may have merged since the change was written). If this produces erroneous results for a job, the behavior may be disabled by setting the zuul.disable_file_comment_line_mapping variable to true in zuul_return.

If zuul_return is invoked multiple times (e.g., via multiple playbooks), then the elements of zuul.file_comments from each invocation will be appended.

Pausing the job

A job can be paused after the run phase by notifing zuul during the run phase. In this case the dependent jobs can start and the prior job stays paused until all dependent jobs are finished. This for example can be useful to start a docker registry in a job that will be used by the dependent job. To indicate that the job should be paused use zuul_return to set the zuul.pause value. You still can at the same time supply any arbitrary data to the dependent jobs. For example:

tasks:
  - zuul_return:
      data:
        zuul:
          pause: true
        registry_ip_address: "{{ hostvars[groups.all[0]].ansible_host }}"

Build Status

A job build may have the following status:

SUCCESS: Nominal job execution.
FAILURE: Job executed correctly, but exited with a failure.
RETRY: The pre-run playbook failed and the job will be retried.
RETRY_LIMIT: The pre-run playbook failed more than the maximum number of retry attempts.
POST_FAILURE: The post-run playbook failed.
SKIPPED: One of the build dependencies failed and this job was not executed.
NODE_FAILURE: The test instance provider was unable to fullfill the nodeset request. This can happen if Nodepool is unable to provide the requested node(s) for the request.