Pipelines

Overview

A Pipeline is a collection of Tasks that you define and arrange in a specific order of execution as part of your continuous integration flow. Each Task in a Pipeline executes as a Pod on your Kubernetes cluster. You can configure various execution conditions to fit your business needs.

Configuring a Pipeline

A Pipeline definition supports the following fields:

  • Required:
    • apiVersion - Specifies the API version, for example tekton.dev/v1beta1.
    • kind - Identifies this resource object as a Pipeline object.
    • metadata - Specifies metadata that uniquely identifies the Pipeline object. For example, a name.
    • spec - Specifies the configuration information for this Pipeline object. This must include:
      • tasks - Specifies the Tasks that comprise the Pipeline and the details of their execution.
  • Optional:
    • resources - alpha only Specifies PipelineResources needed or created by the Tasks comprising the Pipeline.
    • tasks:
      • resources.inputs / resource.outputs
      • runAfter - Indicates that a Task should execute after one or more other Tasks without output linking.
      • retries - Specifies the number of times to retry the execution of a Task after a failure. Does not apply to execution cancellations.
      • conditions - Specifies Conditions that only allow a Task to execute if they successfully evaluate.
      • timeout - Specifies the timeout before a Task fails.
    • results - Specifies the location to which the Pipeline emits its execution results.
    • description - Holds an informative description of the Pipeline object.
    • finally - Specifies one or more Tasks to be executed in parallel after all other tasks have completed.

Specifying Resources

A Pipeline requires PipelineResources to provide inputs and store outputs for the Tasks that comprise it. You can declare those in the resources field in the spec section of the Pipeline definition. Each entry requires a unique name and a type. For example:

spec:
  resources:
    - name: my-repo
      type: git
    - name: my-image
      type: image

Specifying Workspaces

Workspaces allow you to specify one or more volumes that each Task in the Pipeline requires during execution. You specify one or more Workspaces in the workspaces field. For example:

spec:
  workspaces:
    - name: pipeline-ws1 # The name of the workspace in the Pipeline
  tasks:
    - name: use-ws-from-pipeline
      taskRef:
        name: gen-code # gen-code expects a workspace with name "output"
      workspaces:
        - name: output
          workspace: pipeline-ws1
    - name: use-ws-again
      taskRef:
        name: commit # commit expects a workspace with name "src"
      runAfter:
        - use-ws-from-pipeline # important: use-ws-from-pipeline writes to the workspace first
      workspaces:
        - name: src
          workspace: pipeline-ws1

For more information, see:

Specifying Parameters

You can specify global parameters, such as compilation flags or artifact names, that you want to supply to the Pipeline at execution time. Parameters are passed to the Pipeline from its corresponding PipelineRun and can replace template values specified within each Task in the Pipeline.

Parameter names:

  • Must only contain alphanumeric characters, hyphens (-), and underscores (_).
  • Must begin with a letter or an underscore (_).

For example, fooIs-Bar_ is a valid parameter name, but barIsBa$ or 0banana are not.

Each declared parameter has a type field, which can be set to either array or string. array is useful in cases where the number of compilation flags being supplied to the Pipeline varies throughout its execution. If no value is specified, the type field defaults to string. When the actual parameter value is supplied, its parsed type is validated against the type field. The description and default fields for a Parameter are optional.

The following example illustrates the use of Parameters in a Pipeline.

The following Pipeline declares an input parameter called context and passes its value to the Task to set the value of the pathToContext parameter within the Task. If you specify a value for the default field and invoke this Pipeline in a PipelineRun without specifying a value for context, that value will be used.

Note: Input parameter values can be used as variables throughout the Pipeline by using variable substitution.

apiVersion: tekton.dev/v1beta1
kind: Pipeline
metadata:
  name: pipeline-with-parameters
spec:
  params:
    - name: context
      type: string
      description: Path to context
      default: /some/where/or/other
  tasks:
    - name: build-skaffold-web
      taskRef:
        name: build-push
      params:
        - name: pathToDockerFile
          value: Dockerfile
        - name: pathToContext
          value: "$(params.context)"

The following PipelineRun supplies a value for context:

apiVersion: tekton.dev/v1beta1
kind: PipelineRun
metadata:
  name: pipelinerun-with-parameters
spec:
  pipelineRef:
    name: pipeline-with-parameters
  params:
    - name: "context"
      value: "/workspace/examples/microservices/leeroy-web"

Adding Tasks to the Pipeline

Your Pipeline definition must reference at least one Task. Each Task within a Pipeline must have a valid name and a taskRef. For example:

tasks:
  - name: build-the-image
    taskRef:
      name: build-push

You can use PipelineResources as inputs and outputs for Tasks in the Pipeline. For example:

spec:
  tasks:
    - name: build-the-image
      taskRef:
        name: build-push
      resources:
        inputs:
          - name: workspace
            resource: my-repo
        outputs:
          - name: image
            resource: my-image

You can also provide Parameters:

spec:
  tasks:
    - name: build-skaffold-web
      taskRef:
        name: build-push
      params:
        - name: pathToDockerFile
          value: Dockerfile
        - name: pathToContext
          value: /workspace/examples/microservices/leeroy-web

Using the from parameter

If a Task in your Pipeline needs to use the output of a previous Task as its input, use the optional from parameter to specify a list of Tasks that must execute before the Task that requires their outputs as its input. When your target Task executes, only the version of the desired PipelineResource produced by the last Task in this list is used. The name of this output PipelineReource output must match the name of the input PipelineResource specified in the Task that ingests it.

In the example below, the deploy-app Task ingests the output of the build-app Task named my-image as its input. Therefore, the build-app Task will execute before the deploy-app Task regardless of the order in which those Tasks are declared in the Pipeline.

- name: build-app
  taskRef:
    name: build-push
  resources:
    outputs:
      - name: image
        resource: my-image
- name: deploy-app
  taskRef:
    name: deploy-kubectl
  resources:
    inputs:
      - name: image
        resource: my-image
        from:
          - build-app

Using the runAfter parameter

If you need your Tasks to execute in a specific order within the Pipeline but they don’t have resource dependencies that require the from parameter, use the runAfter parameter to indicate that a Task must execute after one or more other Tasks.

In the example below, we want to test the code before we build it. Since there is no output from the test-app Task, the build-app Task uses runAfter to indicate that test-app must run before it, regardless of the order in which they are referenced in the Pipeline definition.

- name: test-app
  taskRef:
    name: make-test
  resources:
    inputs:
      - name: workspace
        resource: my-repo
- name: build-app
  taskRef:
    name: kaniko-build
  runAfter:
    - test-app
  resources:
    inputs:
      - name: workspace
        resource: my-repo

Using the retries parameter

For each Task in the Pipeline, you can specify the number of times Tekton should retry its execution when it fails. When a Task fails, the corresponding TaskRun sets its Succeeded Condition to False. The retries parameter instructs Tekton to retry executing the Task when this happens.

If you expect a Task to encounter problems during execution (for example, you know that there will be issues with network connectivitity or missing dependencies), set its retries parameter to a suitable value greater than 0. If you don’t explicitly specify a value, Tekton does not attempt to execute the failed Task again.

In the example below, the execution of the build-the-image Task will be retried once after a failure; if the retried execution fails, too, the Task execution fails as a whole.

tasks:
  - name: build-the-image
    retries: 1
    taskRef:
      name: build-push

Guard Task execution using Conditions

To run a Task only when certain conditions are met, it is possible to guard task execution using the conditions field. The conditions field allows you to list a series of references to Condition resources. The declared Conditions are run before the Task is run. If all of the conditions successfully evaluate, the Task is run. If any of the conditions fails, the Task is not run and the TaskRun status field ConditionSucceeded is set to False with the reason set to ConditionCheckFailed.

In this example, is-master-branch refers to a Condition resource. The deploy task will only be executed if the condition successfully evaluates.

tasks:
  - name: deploy-if-branch-is-master
    conditions:
      - conditionRef: is-master-branch
        params:
          - name: branch-name
            value: my-value
    taskRef:
      name: deploy

Unlike regular task failures, condition failures do not automatically fail the entire PipelineRun – other tasks that are not dependent on the Task (via from or runAfter) are still run.

In this example, (task C) has a condition set to guard its execution. If the condition is not successfully evaluated, task (task D) will not be run, but all other tasks in the pipeline that not depend on (task C) will be executed and the PipelineRun will successfully complete.

       (task B) — (task E)
     / 
 (task A) 
     \
       (guarded task C) — (task D)

Resources in conditions can also use the from field to indicate that they expect the output of a previous task as input. As with regular Pipeline Tasks, using from implies ordering – if task has a condition that takes in an output resource from another task, the task producing the output resource will run first:

tasks:
  - name: first-create-file
    taskRef:
      name: create-file
    resources:
      outputs:
        - name: workspace
          resource: source-repo
  - name: then-check
    conditions:
      - conditionRef: "file-exists"
        resources:
          - name: workspace
            resource: source-repo
            from: [first-create-file]
    taskRef:
      name: echo-hello

Configuring the failure timeout

You can use the Timeout field in the Task spec within the Pipeline to set the timeout of the TaskRun that executes that Task within the PipelineRun that executes your Pipeline. The Timeout value is a duration conforming to Go’s ParseDuration format. For example, valid values are 1h30m, 1h, 1m, and 60s.

Note: If you do not specify a Timeout value, Tekton instead honors the timeout for the PipelineRun.

In the example below, the build-the-image Task is configured to time out after 90 seconds:

spec:
  tasks:
    - name: build-the-image
      taskRef:
        name: build-push
      Timeout: "0h1m30s"

Using Results

Tasks can emit Results when they execute. A Pipeline can use these Results for two different purposes:

  1. A Pipeline can pass the Result of a Task in to the Parameters of another.
  2. A Pipeline can itself emit Results and include data from the Results of its Tasks.

Passing one Task’s Results into the Parameters of another

Sharing Results between Tasks in a Pipeline happens via variable substitution - one Task emits a Result and another receives it as a Parameter with a variable such as $(tasks.<task-name>.results.<result-name>).

When one Task receives the Results of another, there is a dependency created between those two Tasks. In order for the receiving Task to get data from another Task's Result, the Task producing the Result must run first. Tekton enforces this Task ordering by ensuring that the Task emitting the Result executes before any Task that uses it.

In the snippet below, a param is provided its value from the commit Result emitted by the checkout-source Task. Tekton will make sure that the checkout-source Task runs before this one.

params:
  - name: foo
    value: "$(tasks.checkout-source.results.commit)"

For an end-to-end example, see Task Results in a PipelineRun.

Emitting Results from a Pipeline

A Pipeline can emit Results of its own for a variety of reasons - an external system may need to read them when the Pipeline is complete, they might summarise the most important Results from the Pipeline's Tasks, or they might simply be used to expose non-critical messages generated during the execution of the Pipeline.

A Pipeline's Results can be composed of one or many Task Results emitted during the course of the Pipeline's execution. A Pipeline Result can refer to its Tasks' Results using a variable of the form $(tasks.<task-name>.results.<result-name>).

After a Pipeline has executed the PipelineRun will be populated with the Results emitted by the Pipeline. These will be written to the PipelineRun's status.pipelineResults field.

In the example below, the Pipeline specifies a results entry with the name sum that references the outputValue Result emitted by the calculate-sum Task.

  results:
    - name: sum
      description: the sum of all three operands
      value: $(tasks.calculate-sum.results.outputValue)

For an end-to-end example, see Results in a PipelineRun.

Configuring the Task execution order

You can connect Tasks in a Pipeline so that they execute in a Directed Acyclic Graph (DAG). Each Task in the Pipeline becomes a node on the graph that can be connected with an edge so that one will run before another and the execution of the Pipeline progresses to completion without getting stuck in an infinite loop.

This is done using:

For example, the Pipeline defined as follows

- name: lint-repo
  taskRef:
    name: pylint
  resources:
    inputs:
      - name: workspace
        resource: my-repo
- name: test-app
  taskRef:
    name: make-test
  resources:
    inputs:
      - name: workspace
        resource: my-repo
- name: build-app
  taskRef:
    name: kaniko-build-app
  runAfter:
    - test-app
  resources:
    inputs:
      - name: workspace
        resource: my-repo
    outputs:
      - name: image
        resource: my-app-image
- name: build-frontend
  taskRef:
    name: kaniko-build-frontend
  runAfter:
    - test-app
  resources:
    inputs:
      - name: workspace
        resource: my-repo
    outputs:
      - name: image
        resource: my-frontend-image
- name: deploy-all
  taskRef:
    name: deploy-kubectl
  resources:
    inputs:
      - name: my-app-image
        resource: my-app-image
        from:
          - build-app
      - name: my-frontend-image
        resource: my-frontend-image
        from:
          - build-frontend

executes according to the following graph:

        |            |
        v            v
     test-app    lint-repo
    /        \
   v          v
build-app  build-frontend
   \          /
    v        v
    deploy-all

In particular:

  1. The lint-repo and test-app Tasks have no from or runAfter clauses and start executing simultaneously.
  2. Once test-app completes, both build-app and build-frontend start executing simultaneously since they both runAfter the test-app Task.
  3. The deploy-all Task executes once both build-app and build-frontend complete, since it ingests PipelineResources from both.
  4. The entire Pipeline completes execution once both lint-repo and deploy-all complete execution.

Adding a description

The description field is an optional field and can be used to provide description of the Pipeline.

Adding Finally to the Pipeline

You can specify a list of one or more final tasks under finally section. Final tasks are guaranteed to be executed in parallel after all PipelineTasks under tasks have completed regardless of success or error. Final tasks are very similar to PipelineTasks under tasks section and follow the same syntax. Each final task must have a valid name and a taskRef or taskSpec. For example:

spec:
  tasks:
    - name: tests
      taskRef:
        Name: integration-test
  finally:
    - name: cleanup-test
      taskRef:
        Name: cleanup

Specifying Workspaces in Final Tasks

Finally tasks can specify workspaces which PipelineTasks might have utilized e.g. a mount point for credentials held in Secrets. To support that requirement, you can specify one or more Workspaces in the workspaces field for the final tasks similar to tasks.

spec:
  resources:
    - name: app-git
      type: git
  workspaces:
    - name: shared-workspace
  tasks:
    - name: clone-app-source
      taskRef:
        name: clone-app-repo-to-workspace
      workspaces:
        - name: shared-workspace
          workspace: shared-workspace
      resources:
        inputs:
          - name: app-git
            resource: app-git
  finally:
    - name: cleanup-workspace
      taskRef:
        name: cleanup-workspace
      workspaces:
        - name: shared-workspace
          workspace: shared-workspace

Specifying Parameters in Final Tasks

Similar to tasks, you can specify Parameters in final tasks:

spec:
  tasks:
    - name: tests
      taskRef:
        Name: integration-test
  finally:
    - name: report-results
      taskRef:
        Name: report-results
      params:
        - name: url
          value: "someURL"

PipelineRun Status with finally

With finally, PipelineRun status is calculated based on PipelineTasks under tasks section and final tasks.

Without finally:

PipelineTasks under tasks PipelineRun status Reason
all PipelineTasks successful true Succeeded
one or more PipelineTasks skipped and rest successful true Completed
single failure of PipelineTask false failed

With finally:

PipelineTasks under tasks Final Tasks PipelineRun status Reason
all PipelineTask successful all final tasks successful true Succeeded
all PipelineTask successful one or more failure of final tasks false Failed
one or more PipelineTask skipped and rest successful all final tasks successful true Completed
one or more PipelineTask skipped and rest successful one or more failure of final tasks false Failed
single failure of PipelineTask all final tasks successful false failed
single failure of PipelineTask one or more failure of final tasks false failed

Overall, PipelineRun state transitioning is explained below for respective scenarios:

  • All PipelineTask and final tasks are successful: Started -> Running -> Succeeded
  • At least one PipelineTask skipped and rest successful: Started -> Running -> Completed
  • One PipelineTask failed / one or more final tasks failed: Started -> Running -> Failed

Please refer to the table under Monitoring Execution Status to learn about what kind of events are triggered based on the Pipelinerun status.

Known Limitations

Specifying Resources in Final Tasks

Similar to tasks, you can use PipelineResources as inputs and outputs for final tasks in the Pipeline. The only difference here is, final tasks with an input resource can not have a from clause like a PipelineTask from tasks section. For example:

spec:
  tasks:
    - name: tests
      taskRef:
        Name: integration-test
      resources:
        inputs:
          - name: source
            resource: tektoncd-pipeline-repo
        outputs:
          - name: workspace
            resource: my-repo
  finally:
    - name: clear-workspace
      taskRef:
        Name: clear-workspace
      resources:
        inputs:
          - name: workspace
            resource: my-repo
            from: #invalid
              - tests

Cannot configure the Final Task execution order

It’s not possible to configure or modify the execution order of the final tasks. Unlike Tasks in a Pipeline, all final tasks run simultaneously and start executing once all PipelineTasks under tasks have settled which means no runAfter can be specified in final tasks.

Cannot specify execution Conditions in Final Tasks

Tasks in a Pipeline can be configured to run only if some conditions are satisfied using conditions. But the final tasks are guaranteed to be executed after all PipelineTasks therefore no conditions can be specified in final tasks.

Cannot configure Task execution results with finally

Final tasks can not be configured to consume Results of PipelineTask from tasks section i.e. the following example is not supported right now but we are working on adding support for the same (tracked in issue #2557).

spec:
  tasks:
    - name: count-comments-before
      taskRef:
        Name: count-comments
    - name: add-comment
      taskRef:
        Name: add-comment
    - name: count-comments-after
      taskRef:
        Name: count-comments
  finally:
    - name: check-count
      taskRef:
        Name: check-count
      params:
        - name: before-count
          value: $(tasks.count-comments-before.results.count) #invalid
        - name: after-count
          value: $(tasks.count-comments-after.results.count) #invalid

Cannot configure Pipeline result with finally

Final tasks can emit Results but results emitted from the final tasks can not be configured in the Pipeline Results. We are working on adding support for this (tracked in issue #2710).

  results:
    - name: comment-count-validate
      value: $(finally.check-count.results.comment-count-validate)

In this example, PipelineResults is set to:

"pipelineResults": [
  {
    "name": "comment-count-validate",
    "value": "$(finally.check-count.results.comment-count-validate)"
  }
],

Code examples

For a better understanding of Pipelines, study our code examples.

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License.