K8S API for Pipeline

Compared with the Application Workflow, the standalone pipeline has the following characteristics:

1. It can manage multiple KubeVela Applications across multiple environments.
2. It is not bound to Applications and can be used independently. For example, it can expand or shrink a set of resources, perform process-oriented canary publishing for an Application, and perform a set of operation and maintenance operations in batches.
3. It is one-time and does not manage resources. Even if the pipeline is deleted, the created resources will not be deleted.
4. It uses the same execution engine as the Application Workflow, which completely inherits the features of KubeVela’s lightweight workflow. Compared with the traditional container-based CI pipeline, KubeVela’s pipeline does not depend on containers, No additional computing resources are required.

tip

In order to better reuse the existing capabilities and ensure technical consistency, we split the workflow engine part of the original application workflow. Both in-application workflow and pipeline use this workflow engine as the underlying technology implementation. The application workflow is represented by the Workflow field in the application, and the pipeline is represented by the WorkflowRun resource.

This means that most of the workflow steps are common between the two, such as: suspend, notification, send HTTP request, read configuration, etc.

However, in WorkflowRun, there is only the configuration of steps, and no configuration of components, traits, and policies. Therefore, steps related to components/traits/policy can only be used in in-app workflows, such as: deploying/updating components, traits, etc.

Please make sure that you have enabled workflow addon:

vela addon enable vela-workflow

WorkflowRun is the K8S API for pipeline. You can choose to execute an external Workflow template in the WorkflowRun or execute the steps in the WorkflowRun spec (if you declare both, the step in the WorkflowRun spec will override the content in the template). A WorkflowRun consists of the following:

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: <name>
  namespace: <namespace>
spec:
  mode: <optional execute mode for the workflowRun, default execute mode is StepByStep for steps, DAG for subSteps>
    steps: <DAG or StepByStep>
    subSteps: <DAG or StepByStep>
  context:
    <optional custom contest values>
  workflowRef: <optional external workflow template to run>
  workflowSpec: <optional workflow spec to run>
    steps:
    - name: <name>
      type: <type>
      dependsOn:
        <optional array of step names, specify the dependency for the step>
      meta: <optional meta data for the step>
        alias: <optional alias of the step>  
      properties:
        <parameter values>
      if: <optional if condition to decide whether this step should be executed>
      timeout: <optional timeout for the step>
      outputs: <optional outputs value>
        - name: <name>
          valueFrom: <value source of the output>
      inputs: <optional inputs value>
        - name: <name>
          parameterKey: <optional set the inputs data to the steps'parameter>
      subSteps:
        <optional sub steps if the type of this step is step-group>

WorkflowRun has the following status:

WorkflowRun steps have the following status:

For steps that fail to execute, the message of the step status will display the failed message, and the reason will display the failed reason, which is divided into the following types:

You can define execution mode in WorkflowRun or Workflow templates:

mode:
  steps: <DAG or StepByStep>
  subSteps: <DAG or StepByStep>

If not explicitly specified, the WorkflowRun will execute the steps sequentially (StepByStep) and execute sub-steps in parallel (DAG) by default.

caution

If you specify the execution mode in both WorkflowRun and Workflow, the mode in WorkflowRun will override the mode in the Workflow template.

You can use KubeVela built-in steps that without label: custom.definition.oam.dev/scope: Application in WorkflowRun.

You can refer to the custom steps documentation to customize your steps.

caution

You cannot use application operations.

tip

The vela workflow command can operate both Application Workflow and WorkflowRun. By default, it will look for the application with the same name first, and if it is not found, it will look for WorkflowRun. You can also use --type=workflow to indicate that the operation object is WorkflowRun.

If you have an executing WorkflowRun, you can use vela workflow suspend to suspend the workflow.

vela workflow suspend <name>

tip

If the workflow has executed completely, using the vela workflow suspend command has no effect.

When the WorkflowRun is suspended, you can use vela workflow resume command to manually resume the workflow.

vela workflow resume <name>

If you have an executing WorkflowRun, you can use vela workflow terminate to terminate the workflow.

vela workflow terminate <name>

If you want to view the WorkflowRun logs, you can use vela workflow logs command to view the logs.

tip

Only steps configured with op.#Log in its definition will have log output.

vela workflow logs <name>

Please refer to Operate WorkflowRun.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: suspend
  namespace: default
spec:
  workflowSpec:
    steps:
      - name: step1
        type: apply-deployment
        properties:
          image: nginx
      - name: step2-suspend
        type: suspend
      - name: step2
        type: apply-deployment
        properties:
          image: nginx

The WorkflowRun will automatically suspend when the first step is completed, and the third step will not be executed until you continue the WorkflowRun.

Please refer to Operate WorkflowRun.

Configure duration: <duration> in the suspend type of step, when the duration time expires, WorkflowRun will automatically continue to execute.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: suspend
  namespace: default
spec:
  workflowSpec:
    steps:
      - name: step1
        type: apply-deployment
        properties:
          image: nginx
      - name: step2-suspend
        type: suspend
        properties:
          duration: 10s
      - name: step2
        type: apply-deployment
        properties:
          image: nginx

When the first step is completed, the WorkflowRun will suspend, and after ten seconds, the WorkflowRun will automatically continue to execute the third step.

There is a special step type called step-group. When using a step-group type of step, you can declare sub steps in it.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: group
  namespace: default
spec:
  workflowSpec:
    steps:
      - name: my-group
        type: step-group
        subSteps:
          - name: sub1
            type: apply-deployment
            properties:
              image: nginx
          - name: sub2
            type: apply-deployment
            properties:
              image: nginx

You can specify dependencies between steps with dependsOn.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: dependency
  namespace: default
spec:
  mode:
    steps: DAG
  workflowSpec:
    steps:
      - name: step1
        type: apply-deployment
        dependsOn:
          - step2
          - step3
        properties:
          image: nginx
      - name: step2
        type: apply-deployment
        properties:
          image: nginx
      - name: step3
        type: apply-deployment
        properties:
          image: nginx

step1 will be executed after step2 and step3 are completed.

Data passing between steps can be done through inputs and outputs. For details, please refer to Input and output between steps.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: request-http
  namespace: default
spec:
  workflowSpec:
    steps:
    - name: request
      type: request
      properties:
        url: https://api.github.com/repos/kubevela/workflow
      outputs:
        - name: stars
          valueFrom: |
            import "strconv"
            "Current star count: " + strconv.FormatInt(response["stargazers_count"], 10)
    - name: notification
      type: notification
      inputs:
        - from: stars
          parameterKey: slack.message.text
      properties:
        slack:
          url:
            value: <your slack url>

In this WorkflowRun, the first step will request the GitHub API to get the number of stars in the workflow repository as Output, and then use this Output as Input in the next step to send the star number as the message to Slack.

You can specify timeout for a step to indicate the timeout for that step.

timeout follows the duration format, e.g. 30s, 1m, etc. You can refer to Golang’s parseDuration.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: timeout
  namespace: default
spec:
  workflowSpec:
    steps:
      - name: suspend
        type: suspend
        timeout: 3s

If the above WorkflowRun is not resumed within three seconds, the suspend step will fail with timeout.

You can use if in a step to determine whether to execute the step.

If the step does not specify if, if the step before the step fails to execute, then the step will be skipped and will not be executed.

With if: always specified in a step, the step will be executed no matter what.

You can also write your own judgment logic to determine whether the step should be executed. Note: The value in if will be executed as CUE code. WorkflowRun provides some built-in variables in if, these are:

• status：status contains status information for all workflow steps. You can use status.<step-name>.phase == "succeeded" to determine the status of a step, or you can use the simplified status.<step-name>.succeeded to determine.
• inputs：inputs contains all the inputs parameters of the step. You can use inputs.<input-name> == "value" to get input for the step.
• context: context contains all the context data of WorkflowRun. You can use context.<context-name> == "value" to get the context of the WorkflowRun.

tip

Note that if your step name or inputs name is not a valid CUE variable name (eg: contains -, or starts with a number, etc.), you can refer to it as follows: status["invalid-name"].failed.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: if-condition
  namespace: default
spec:
  workflowSpec:
    steps:
      - name: suspend
        type: suspend
        timeout: 3s
      - name: my-step
        type: apply-deployment
        if: status.suspend.failed
        properties:
          image: nginx
      - name: my-step2
        type: apply-deployment
        if: status.suspend.succecceed
        properties:
          image: busybox

In the above WorkflowRun, if the suspend step fails due to a timeout, then the my-step step will be executed, otherwise the my-step2 step will be executed.

Steps in WorkflowRun have some built-in context data, and you can also declare your custom context parameters in context.

tip

If your custom context data has the same name as a built-in context data, the built-in context parameter will be overridden by the custom parameter.

You can control the execution of WorkflowRun in different situations through the combination of conditional if and custom data.

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: deploy-run
  namespace: default
spec:
  context:
    env: test
  workflowRef: deploy-template
---
apiVersion: core.oam.dev/v1alpha1
kind: Workflow
metadata:
  name: deploy-template
  namespace: default
steps:
  - name: apply
    type: apply-deployment
    if: context.env == "dev"
    properties:
      image: nginx
  - name: apply-test
    type: apply-deployment
    if: context.env == "test"
    properties:
      image: crccheck/hello-world

The above WorkflowRun will refer to the deploy-template Workflow as the execution template. If the env in the context is dev, then the apply step will be executed, otherwise the apply-test step will be executed.

The built-in context data in WorkflowRun are as follows:

Last updated on Aug 4, 2023 by Daniel Higuero