Pipeline maintenance
What it is
Pipeline maintenance is the operational discipline of keeping CI definitions healthy over time — beyond authoring. It covers the run lifecycle (create/update/delete/[rerun]), retention (purging old runs and artifacts), credentials/rotating secret stores, fleet/agent upkeep, and watching for failures and slowdowns so a pipeline stays trustworthy.
Why it exists
Pipelines that are authored once and never touched drift: dependencies fall out of support, agent hardware ages, noisy secret stores expire, and retention grows unbounded. Maintenance exists to keep the CI signal clear (few false failures, no stale runs) so that a red run always means something.
Key ideas
- Retention: purge old runs and their artifacts on a schedule — this is storage hygiene, distinct from authoring.
- Lifecycle ops: create/update/delete runs; rerun failed jobs; version your pipeline definitions.
- Secrets & credentials: a central, maintained store (Key Vault / org secrets) with rotation — never code keys.
- Observation: monitor status/duration/queue lag; alert on failed-and-slow so flaky is caught early.
- Agent upkeep: for self-hosted fleets, patch and capacity-track agents (see agent-pool).
How it fits (diagram)
Exam notes
- Maintenance is operational and ongoing — it is commonly confused with authoring but is lifecycle-driven.
- Retention defaults exist; understand run purging vs keeping history.
- Credential/rotation management is a recurring AZ-400 theme (often paired with Key Vault).
Related
pipeline-run · azure-pipelines · agent-pool · pipeline-run-diagnostics-and-maintenance
📘 Source: What is Azure Pipelines? (lifecycle)
⚠ verify: retention-purging / run-cleanup doc — pending reference.