Exercise Answers¶
Page Maps¶
graph LR
family["Reproducible Research"]
program["Deep Dive DVC"]
section["Migration Governance Dvc Boundaries"]
page["Exercise Answers"]
capstone["Capstone evidence"]
family --> program --> section --> page
page -.applies in.-> capstone
flowchart LR
orient["Orient on the page map"] --> read["Read the main claim and examples"]
read --> inspect["Inspect the related code, proof, or capstone surface"]
inspect --> verify["Run or review the verification path"]
verify --> apply["Apply the idea back to the module and capstone"]
These answers are model explanations, not the only acceptable wording.
What matters is whether the reasoning makes evidence, risk, repair, and ownership clear.
Answer 1: Review a repository by evidence¶
Strong outline:
- identity and state layers: inspect
.dvcfiles,dvc.lock, andmake -C capstone state-summary - pipeline and experiment truth: inspect
dvc.yaml,params.yaml, and experiment review notes - metrics, promotion, and release surfaces: inspect
metrics/,publish/v1/, and release audit output - remote, retention, and recovery: inspect remote config, recovery guide, and recovery review output
- tool-boundary recommendations: inspect deployment or registry handoff points and name what DVC should own
The main lesson is that final review starts from evidence surfaces, not impressions.
Answer 2: Plan one safe migration¶
Strong plan:
- Inventory the existing
publish/v1/manifest, params, metrics, review note, and artifacts. - Copy the supported bundle to
registry/incident-escalation/v1/. - Keep
publish/v1/temporarily as a rollback boundary. - Run release audit and recovery review against the new boundary.
- Update consumer documentation to name the registry entry.
- Remove or deprecate the old boundary only after verification and consumer handoff.
The main lesson is to move one boundary and verify before cutting over.
Answer 3: Write a governance rule¶
Strong rule:
Any change to
params.yaml,dvc.lock, orpublish/must include a review note naming the state contract being changed and must pass the relevant verification route before merge.
Contract protected:
- parameter and metric comparability
- lock evidence consistency
- promoted release auditability
The main lesson is that governance should protect a named contract.
Answer 4: Intervene on an anti-pattern¶
Strong intervention:
outputs/latest/gives consumers a moving internal target. Please publish a versioned bundle such aspublish/v1/or a registry entry with a manifest, promoted params, promoted metrics, and review note. Consumers should depend on that versioned boundary, not the internallatestpath.
The main lesson is to replace the shortcut with a repair path.
Answer 5: Decide tool ownership¶
Strong ownership split:
- DVC should own artifact lineage, data identity, pipeline evidence, params, metrics, and recovery for tracked outputs.
- A registry should own model lifecycle and consumer-facing version contracts.
- A deployment platform should own rollout, runtime health, and serving policy.
- Alerting should belong to operational monitoring, not DVC.
DVC can provide evidence to those systems, but it should not become responsible for every production concern.
Self-check¶
If your answers consistently explain:
- which evidence proves the current state
- how migration preserves or changes a boundary
- which governance rule protects which contract
- how to repair shortcuts before they become normal
- where DVC authority ends
then you are using Module 10 correctly.