TenantAtlas/specs/056-remove-legacy-bulkops/data-model.md

Phase 1 — Data Model: Remove Legacy BulkOperationRun & Canonicalize Operations (v1.0)

This document describes the domain entities and data contracts impacted by Feature 056.

Entities

OperationRun (canonical)

Represents a single observable operation execution. Tenant-scoped.

Core fields (existing):

• tenant_id: Tenant scope for isolation.
• user_id: Initiator user (nullable for system).
• initiator_name: Human name for display.
• type: Operation type (stable identifier).
• status: queued | running | completed.
• outcome: pending | succeeded | partial | failed (exact naming per existing app semantics).
• run_identity_hash: Deterministic identity used for dedupe of active runs.
• context: JSON object containing operation inputs and metadata.
• summary_counts: JSON object with canonical metrics keys.
• failure_summary: JSON array/object with stable reason codes + sanitized messages.
• started_at, completed_at: Timestamps.

Invariants:

• Active-run dedupe is tenant-wide.
• Monitoring renders from DB only.
• summary_counts keys are constrained to the canonical registry.
• Failure messages are sanitized and do not include secrets.

Operation Type (logical)

A stable identifier used to categorize runs, label them for admins, and enforce consistent UX.

Attributes:

• type (string): e.g., policy_version.prune, backup_set.delete, etc.
• label (string): human readable name.
• expected_duration_seconds (optional): typical duration.
• allowed_summary_keys: canonical registry.

Target Scope (logical)

A directory/remote tenant scope that an operation targets.

Context fields (when applicable):

• entra_tenant_id: Azure AD / Entra tenant GUID.
• directory_context_id: Internal directory context identifier.

At least one of the above must be present for directory-targeted operations.

Bulk Selection Identity (logical)

Deterministic definition of “what the bulk action applies to”, required for idempotency.

Decision: Hybrid identity.

• Explicit selection: selection.ids_hash
• Query/filter selection (“select all”): selection.query_hash

Properties:

• selection.kind: ids | query
• selection.ids_hash: sha256 of stable, sorted IDs.
• selection.query_hash: sha256 of normalized filter/query payload.

Bulk Idempotency Fingerprint (logical)

Deterministic fingerprint used to dedupe active runs and prevent duplicated work.

Components:

• operation type
• target scope (entra_tenant_id or directory_context_id)
• selection identity (hybrid)

Removed Entity

BulkOperationRun (legacy)

This entity is removed by Feature 056.

• Legacy model/service/table are deleted.
• No new writes to legacy tables after cutover.
• Historical import into OperationRun is not performed.