Azure child pipelines across repositories with responses

Question

I have multiple repos thanks to a micro-frontend architecture and would like to run the test suite found in another repo on a specified environment/url.

My parent pipeline creates and environment for a PR preview. Rather than checkout the other repos, I was wondering if it possible to call the child repos pipeline?

I have looked at Trigger one pipeline after another but it doesn't appear to fail a parent pipeline, should a child pipeline fail.

For example:

• parent repo pipelines build pr env
• 1a. child repo 1 pipeline tests env x and passes
• 1b. [child repo 2 pipeline tests env x and fails
• 2 parent repo pipeline fails

The hope is that, if we can do this, the test suite and tooling can be managed by the child app teams, without the parent having to support specific frameworks, allowing independence.

Answer 1

Only way I can see this working is using custom script task in parent pipeline to call rest API to queue the child pipeline and wait for the result.

Example python script

import requests
import time
import sys

# Azure DevOps organization URL
org_url = 'https://dev.azure.com/your_org'

# Azure DevOps project name
project_name = 'your_project'

# Azure DevOps pipeline ID
pipeline_id = 'your_pipeline_id'

# Azure DevOps personal access token with access to pipelines
pat = 'your_personal_access_token' # System_AccessToken can be passed from pipeline to use the PAT from the build service account 

# Queue the pipeline
queue_url = f'{org_url}/{project_name}/_apis/pipelines/{pipeline_id}/runs?api-version=6.0-preview.1'
headers = {
    'Content-Type': 'application/json',
    'Authorization': f'Bearer {pat}'
}
response = requests.post(queue_url, headers=headers)
if response.status_code != 200:
    print(f'Error queuing pipeline: {response.text}')
    sys.exit(1)

# Get the pipeline run ID
run_id = response.json()['id']
print(f'Queued pipeline run {run_id}')

# Wait for the pipeline to complete
status_url = f'{org_url}/{project_name}/_apis/pipelines/runs/{run_id}?api-version=6.0-preview.1'
while True:
    response = requests.get(status_url, headers=headers)
    if response.status_code != 200:
        print(f'Error getting pipeline status: {response.text}')
        sys.exit(1)

    result = response.json()['result']
    if result is not None:
        break

    print('Pipeline still running, checking again in 10 seconds')
    time.sleep(10)

# Check the result of the pipeline
if result == 'succeeded':
    print('Pipeline run succeeded')
    sys.exit(0)
else:
    print('Pipeline run failed')
    sys.exit(1)

Answer 2

We have a parent-child pipeline structure implemented. Using the REST API to queue up a child pipeline is straightforward, but waiting for it to finish is more elaborate. We didn't want to have the parent pipeline's agent sitting idle just waiting for the child pipeline, because that would be expensive; so we have a stage with four jobs:

1. the first job fetches a preconfigured PAT Token from Key Vault, and uses it in a Powershell task to call the REST API to queue the child pipeline
2. an agentless server job waits for the child pipeline to finish by making an API call to trigger a cheap Logic App, which repeatedly queries the child pipeline's status until it is finished:
3. then an Evaluation job which uses a different REST API endpoint to fetch the final state of the child pipeline, then does a number of things e.g. copying test results, reporting failures, etc.
4. a Cancel job, run only if the parent pipeline is cancelled, which cancels the child pipeline through the REST API.