What this section is for
Connect CI/CD and your incident tool, then specify which events belong to each service. Patterns match repository, workflow and environment names. Precise settings keep production deployments separate from builds and test environments.
When to use
- Before opening the DORA section for the first time.
- When a new service starts deploying independently.
- When deployment counts do not match the release history in CI/CD.
Quick start
- 1Connect a CI/CD provider โ GitHub Actions, GitLab CI, Bitbucket Pipelines, Jenkins, CircleCI or Azure Pipelines.
- 2Connect an incident tool: PagerDuty, Opsgenie, VictorOps, or a custom webhook.
- 3Create a service and set repository, workflow and environment patterns; wildcards are supported.
- 4Sync, then check on the DORA page that deployment counts are in the range you would expect.
Key settings
| Setting | What it does | How to use it | Common mistake |
|---|---|---|---|
| Repository patterns | Match deployments to a service by repository name. | Use a repository name or a pattern unique to the service, such as org/payments-*. | Using a pattern that includes repositories from another service. |
| Workflow patterns | Restrict which CI workflows count as deployments. | Select the workflow that deploys the application. Use deploy-* only if it matches your actual names. | Matching all workflows, which counts every build and test run as a deployment. |
| Environment patterns | Restrict which environments count. | Select production for production metrics and check the exact environment name in CI/CD. | Leaving it open, so staging deploys inflate frequency and shorten lead time. |
Usage example
Scenario: In this example, DORA shows 40 daily deployments while there were far fewer production releases.
- 1Open the service definition and read the workflow and environment patterns.
- 2Match the production environment and deployment workflow, excluding builds and tests.
- 3Sync again and compare the timeline with production deployment history in CI/CD.
What this shows: The calculation includes the intended service and environment, and you can check its count against CI/CD.
FAQ
Do I have to define services?
Define services to count their deployments and related incidents separately. Combined metrics may mix systems with different release schedules and requirements.
What if the incident tool has no service concept?
You can use the supported custom webhook. Check that the data links an incident to its service and deployment; failure and recovery metrics will be incomplete without those links.
Why is change failure rate zero?
There may be no deployments with linked failures in the selected period. Also check the incident connection, dates, and service and deployment links: missing data does not prove that no failures occurred.
Tips
- After changing patterns, recalculate comparison periods consistently: the set of included events may change.
- Check one service against CI/CD and incident data first. Then configure the others using their own names and environments.