TEST-ALM-001 / SOURCE CONTROLLED

Dataverse ALM drift evidence pack

Use when development, test and production do not behave alike even though the expected managed solution version appears to be deployed.

● CONTROLLEDOWNER / SYSTEM 8REVIEW / MONTHLY

Authority and zero gate

Name release, product, Dataverse, integration and acceptance owners.
Freeze direct edits and imports for the affected components.
Pause or queue unsafe write processes; reconcile before replay.
Export/capture affected definitions, layers, values and deployment records before repair.
Keep credentials, tokens and secret environment values outside this pack.
Do not bulk-remove active layers or force overwrite without component-level approval.

Environment and release inventory

FIELD	CAPTURE	ACCEPTANCE
Environment	Name, ID, type, region, managed status	Target identity confirmed
Solution	Unique name, publisher, managed state, version	Matches approved artifact
Artifact	Build/run ID, checksum, source commit, retained file	Same artifact passed test
Deployment	Pipeline host, stage/run, identity, time, notes	History retained across hosts
Journey	App, table, form, rule, flow, integration	All dependencies enumerated

Component-layer evidence

COMPONENT	TOP LAYER	EXPECTED SOURCE	DISPOSITION
Stable type + unique ID	Managed/unmanaged, solution, version, publisher	Artifact component/version	Keep, reproduce, remove or investigate

Capture each changed property with the component ID, source layer and timestamp. The presence of an active layer alone is incomplete.
Check later managed solutions, patches/upgrades and shared-component ownership.
Classify fallback-form and other platform-created layers separately from direct edits.
Preserve the active definition before approved removal.

Target-configuration inventory

TYPE	CAPTURE	NEGATIVE CHECK
Environment variable	Schema name, type, default/current value state, consumer, owner	Old endpoint/value is not resolved
Connection reference	Schema name, connector, bound connection, consumers	No stale or unintended binding
Connection	Owner/service principal, sharing, auth health	Former maker cannot be required
Configuration data	Table/key, source, count/hash, migration route	Missing/duplicate rows detected
Security	Roles, teams, app access, row ownership	Prohibited persona remains denied

Drift classification and decision record

Describe the observed behavior and operational consequence.
Link each observation to artifact, layer, target binding, data/security or unknown.
Record competing explanations and the evidence that excludes each one.
Classify changes as authorised hotfix, accidental edit, platform-generated layer, release defect, target-configuration mismatch, missing data/security or accepted difference.
Name the repair source, approver, rollback and expiry for every retained difference.

Controlled repair sequence

Reproduce the accepted intent in the development source.
Review dependencies, solution overlap and publisher ownership.
Run Solution Checker and component tests; record their limits.
Build, checksum, retain and promote one versioned managed artifact.
Supply approved environment values and connection bindings.
Migrate/reconcile required configuration rows through a separate controlled path.
After proving replacement, remove only the attributed active layer.
Publish as required and rerun the full acceptance matrix.

Positive and negative acceptance matrix

TEST	EXPECTED	EVIDENCE
Primary business journey	Expected rule and outcome	Persona, record, timestamp
Exception/rejection	Controlled failure; no hidden side effect	Run and target record
Old endpoint/value	Never selected	Resolved configuration/log
Former connection owner	Not required	Connection/reference inventory
Duplicate/retry	One authoritative outcome	Business key reconciliation
Prohibited persona	Read/update denied	Negative access result
Fresh session	Same behavior after publish/cache	New-session result
Layer reinspection	Expected component is effective	Post-release layer capture

Release evidence and operating cadence

Retain the source commit, exact artifact and checksum; checker results; component/dependency manifest; layer captures; environment-value and connection-binding manifest; controlled configuration-data result; deployment run; positive/negative tests; approval; rollback; residual differences and expiry. Redact personal data and secrets.

Every release: pre/post layer and binding comparison for changed high-risk components.
Monthly: connection health, owner/service-principal and failed deployment review.
Quarterly: broader target drift, solution overlap and exception review.
After emergency change: back-port, managed release, layer closure and evidence review.

Limits

This pack covers the sampled components, business rules, security paths and configuration records listed in the capture register. Untested components and Dataverse data rows remain outside the result. Solution Checker is static analysis and does not know all target state. A managed deployment does not include ordinary Dataverse table data. Blocking unmanaged customisations can reduce future drift only after current exceptions are reconciled. Third-party comparison or removal tools require separate security review, backup and component-level authority.