K8S API for Pipeline

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 field in the application, and the pipeline is represented by the 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.

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:

Status

WorkflowRun has the following status:

WorkflowRun Step Status

WorkflowRun steps have the following status:

The Failed Reason of WorkflowRun Step

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.

Built-in Steps

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

You can refer to the to customize your steps.

caution

You cannot use application operations.

Core Features

Operate WorkflowRun

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.

Suspend

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.

Resume

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

vela workflow resume <name>

Terminate

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

Check the Logs

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

tip

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

vela workflow logs <name>

Suspend Manually

Please refer to Operate WorkflowRun.

Suspend Automatically（using suspend step）

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
          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.

Resume Manually

Please refer to .

Resume Automatically

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.

Sub Steps

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

Dependency

You can specify dependencies between steps with dependsOn.

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

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
      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.

Timeout

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
  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.

If Conditions

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

No If specified

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.

if: always

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

Custom If

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.

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"
      image: crccheck/hello-world

Built-in Context Data

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