WIP: [SAP] Add volume history tracking

[hemna]

Summary

This PR adds a volume history tracking feature that records every mutation to a volume's DB row in a new volume_history table. Each history record captures JSON deltas of old/new values along with contextual metadata.

Status: Work In Progress

Changes

• New VolumeHistory model in models.py with FK to volumes.id
• Alembic migration 633b14d87cec to create the volume_history table
• DB API hooks in volume_create(), volume_update(), volume_destroy(), volume_attached(), volume_detached()
• Query function volume_history_get_all_by_volume() to retrieve history
• Config option volume_history_enabled to disable tracking in high-throughput environments
• Unit tests for all history tracking functionality
• SAP documentation in sap-doc/volume-history.md

Configuration

Volume history tracking is enabled by default but can be disabled:

[DEFAULT]
volume_history_enabled = False

Tracked Actions

Action	Description
create	Volume creation - all initial field values
update	Volume field update - only changed fields
destroy	Volume deletion - status transition
attach	Volume attachment - status changes
detach	Volume detachment - status changes

Performance Overhead Analysis

volume_update() — Heaviest overhead (most frequent operation)

Added work	Cost
Extra SELECT query (query.filter_by(id=volume_id).first())	1 DB round-trip
Python loop comparing old vs new values	Negligible
json.dumps() of changes dict	Negligible
session.add() for history INSERT	1 INSERT (batched in same transaction)

Each volume_update adds 1 extra SELECT by PK + 1 extra INSERT, both within the same DB transaction that was already open. The SELECT is a fast indexed lookup on the volumes.id primary key.

volume_destroy() — Moderate overhead

Added work	Cost
query.first() to get current status	1 DB round-trip (consumes existing query, so we re-create it)
Re-creating the query object	0 extra round-trips (Python object construction only)
session.add() for history INSERT	1 INSERT (batched)

Adds 1 extra SELECT + 1 extra INSERT.

volume_create() — Minimal overhead

Added work	Cost
Dict comprehension to build changes	Negligible (pure Python, no DB)
session.add() for history INSERT	1 INSERT (batched in same transaction)

Just 1 extra INSERT, no extra SELECT since we already have the values dict.

volume_attached() / volume_detached() — Minimal overhead

Same as create — just 1 extra INSERT each, no extra SELECT.

Mitigation

Set volume_history_enabled = False in cinder.conf to completely disable history tracking and eliminate all overhead.

Testing

All tests pass in Docker (Python 3.10):

• 8 new volume history unit tests (including config disable test)
• 67 existing volume tests (including updated test_volume_destroy_deletes_dependent_data)
• ORM relationship tests
• Purge tests

TODO

• Review and feedback
• Consider REST API endpoint (future iteration)