feat: migrate backend persistence from TinyDB to MongoDB Atlas

- Implement lazy MongoDB client initialization in backend/db/mongo_client.py.
- Create Gunicorn configuration to ensure MongoDB client is initialized per worker.
- Refactor database access layer to support MongoDB with a new MongoLockedTable adapter.
- Add migration script to transfer existing TinyDB data to MongoDB, preserving idempotency.
- Update tracking event handling to ensure deterministic ordering with a monotonic sequence.
- Modify tests to use mongomock for MongoDB integration and ensure existing tests pass.
- Add integration test script to run tests against a local MongoDB Docker container.
- Document environment variables and migration process in specs/feat-database-migration.md.
This commit is contained in:
2026-07-28 01:10:52 -04:00
parent c9c1fa18a2
commit da7ed41938
17 changed files with 1364 additions and 62 deletions
+113
View File
@@ -0,0 +1,113 @@
# python
"""Lazy MongoDB client factory and database-name helpers.
The client is intentionally **not** created at module import. Use
``get_mongo_client()`` to obtain a cached singleton. For Gunicorn multi-worker
deployments call ``init_mongo_client()`` from a ``post_fork`` hook so each
worker process owns its own connection pool rather than inheriting the parent
process's client.
"""
import os
import threading
from urllib.parse import urlparse
from pymongo import MongoClient
from pymongo.uri_parser import parse_uri
_mongo_client_lock = threading.Lock()
_mongo_client = None
def _create_mongo_client():
"""Build a fail-fast MongoClient from environment variables."""
uri = os.environ.get('MONGO_URI')
if not uri:
raise RuntimeError(
'MONGO_URI environment variable is required when USE_MONGODB=true.'
)
# mongomock is used for unit/integration tests without a real server.
if uri.lower().startswith('mongomock') or uri.lower() == 'mongomock':
try:
import mongomock
except ImportError as exc: # pragma: no cover - test dependency
raise RuntimeError(
'mongomock is required for test MongoDB mode. '
'Install it with: pip install mongomock'
) from exc
return mongomock.MongoClient()
return MongoClient(
uri,
serverSelectionTimeoutMS=5000,
connectTimeoutMS=5000,
maxPoolSize=20,
)
def init_mongo_client():
"""Create a fresh MongoClient and store it as the process singleton.
Call this from a Gunicorn ``post_fork`` hook so each worker process gets
its own client after forking. It can also be called in tests to reset the
shared client to a known state.
"""
global _mongo_client
with _mongo_client_lock:
_mongo_client = _create_mongo_client()
return _mongo_client
def get_mongo_client():
"""Return the cached process-level MongoClient, creating it lazily once."""
global _mongo_client
if _mongo_client is None:
with _mongo_client_lock:
if _mongo_client is None:
_mongo_client = _create_mongo_client()
return _mongo_client
def _db_name_from_uri(uri: str) -> str | None:
"""Extract the database name from a MongoDB connection URI, if present."""
if not uri or uri.lower().startswith('mongomock'):
return None
try:
parsed = parse_uri(uri)
return parsed.get('database') or None
except Exception:
# Fallback to a simple path-based parse for non-standard URIs.
try:
path = urlparse(uri).path
return path.lstrip('/') or None
except Exception:
return None
def get_mongo_db_name() -> str:
"""Resolve the MongoDB database name from env vars or the connection URI.
Precedence:
1. ``MONGO_DB_NAME`` environment variable.
2. Database name parsed from ``MONGO_URI``.
3. Default based on ``DATA_ENV`` / ``DB_ENV``:
* prod -> ``chore_db``
* e2e -> ``chore_db_e2e``
* test -> ``chore_db_test``
"""
env_name = os.environ.get('MONGO_DB_NAME')
if env_name:
return env_name
uri = os.environ.get('MONGO_URI', '')
db_name = _db_name_from_uri(uri)
if db_name:
return db_name
env = (os.environ.get('DATA_ENV') or os.environ.get('DB_ENV', 'prod')).lower()
if env == 'prod':
return 'chore_db'
if env == 'e2e':
return 'chore_db_e2e'
return 'chore_db_test'