Azure DevOps Engineer Flashcards & Memory Notes

A single main branch plus short-lived topic branches merged by pull request.

Azure Boards work-item lead time runs from creation to completion. DORA lead time for changes runs from commit to production. Name the metric before comparing teams.

Time waiting before work becomes active. Cycle time measures active work through completion; lead time also includes the earlier wait.

How long it takes to restore service after a failure.

Links from the work item to the PR, commit, build artifact, and deployment record. A closed work item alone does not prove deployment.

Markdown, including Mermaid diagrams.

It sends an HTTP callback when a repository or pipeline event occurs.

Microsoft Teams.

A work item linked to the commit and the test result.

Git history, generated with the release rather than a hand-edited mail.

One main line with short-lived branches.

Required reviews and checks plus effective branch permissions and restricted bypass rights. Test the permissions of the identities that can push or merge.

Provider file-size limits, storage and bandwidth quotas, and whether existing large-file history needs migration. Enabling LFS does not automatically shrink old commits.

Managing large files outside normal Git blobs.

git revert.

No. It changes local branch, index, and working-tree state and can discard tracked edits. Rewriting the remote requires a separate push; prefer revert for an ordinary shared-history undo.

Restore it from the commit that still contains it.

Revoke or rotate it, then investigate exposure and coordinate any necessary history cleanup. Rewriting central history cannot guarantee erasure from external clones.

A point in history, usually the commit a release built.

Git Scalar or another repository scaling strategy.

An incompatible change.

A calendar date rather than API compatibility.

Publish to the base feed, validate the version, then promote it to the view. The view exposes selected versions; it does not run quality checks automatically.

Not by itself. Preserve test failures, configure publisher failure settings, and enforce required coverage thresholds. Missing-coverage checks do not enforce a percentage threshold.

Configure a Build validation policy on the target branch. Azure Repos does not use the YAML pr trigger for this; GitHub trigger configuration is different.

GitHub. You patch a self-hosted runner yourself.

On the protected environment or other resource, outside pipeline YAML, by authorized resource administrators. The deployment must actually use that protected resource.

Database schema changes and external side effects. Keep releases compatible during transition and test the recovery plan; a traffic switch is not a database restore.

Compare error rates and latency with a suitable baseline, identify the version or cohort, and check traffic volume. A small raw error count or empty data is insufficient evidence.

Azure App Configuration Feature Manager.

It is tied to one Azure resource and deleted with that resource.

It is a standalone identity you can attach to multiple resources.

GitHub creates a repository-scoped installation token for each job. Set minimal job permissions; it is not a permanent organization-wide personal access token.

They can reduce dependence on a person and on reusable stored credentials. If a PAT is required, minimize scope and lifetime and manage revocation explicitly.

Secrets, keys, and certificates.

No. It permits requesting an OIDC token. Azure must trust its issuer, audience, and actual subject claims; the authenticated identity also needs appropriate Azure RBAC.

Known vulnerabilities in dependencies.

Credentials committed to the repository.

No. CodeQL analyzes supported source languages; image scans examine image packages. For a containerized compiled build, CodeQL must observe compilation in a supported container setup.

No. Posture visibility must be connected to remediation owners and explicit required checks or release policy. Define and audit any exceptions.

Application requests, dependencies, and traces.

Virtual machine performance and dependencies.

Container workloads such as Azure Kubernetes Service.

Repository and Actions activity, not in-process application traces.

CPU, memory, disk, and network.

One request across the services that handled it.

Workspace queries use AppRequests, TimeGenerated, and Success; application-context queries can use requests, timestamp, and success. Check the schema where the query runs.

Use a comparable time window and release cohort. With sampling weights, divide failed ItemCount by total ItemCount. No requests means insufficient evidence, not measured health.

Dependency duration in the distributed trace.

Application telemetry and alerts, not only the pipeline result.