Files
kite/backend/tasks/cloud_tasks.py
T
curo1305 bfa4dc502a feat(12.1-02): centralize listing freshness gate — apply_listing_and_finalize
- Add ListingResult dataclass to cloud_items.py
- Add apply_listing_and_finalize() — single shared gate for complete/incomplete semantics
- complete=True: upserts, soft-deletes unseen children, marks fresh, advances last_refreshed_at
- complete=False: upserts seen items only, never deletes unseen, preserves prior timestamp, warns
- browse.py: replace unconditional update_folder_state('fresh') with apply_listing_and_finalize
- cloud_tasks.py: replace unconditional fresh transition with apply_listing_and_finalize
- Celery task returns structured warning status on incomplete (no credentials/provider item names)
- Fix timezone naive/aware comparison in test (SQLite strips tzinfo)
2026-06-22 08:29:41 +02:00

217 lines
8.9 KiB
Python

"""
Celery tasks for cloud metadata refresh in DocuVault — Phase 12.
refresh_cloud_folder — called via .delay(user_id, connection_id, parent_ref)
by the browse endpoint when durable cached rows are stale.
The task is a plain sync def (Celery workers have no asyncio event loop); it
bridges into the async service layer via asyncio.run().
Retry harness (D-13/D-14 — bounded transient retries only):
CRITICAL: self.retry() MUST be raised from the outer sync task, NOT inside
asyncio.run(). _TransientProviderError is a sentinel raised by _run() to signal
a retryable provider/network failure. The outer task catches it and calls
self.retry(countdown=...) in the sync layer.
Terminal errors (auth failure, invalid_grant, scope error):
- Written as CloudFolderState warning state with controlled reason/remedy
- NOT retried — retrying auth failures wastes network calls and may trigger
provider rate limits or OAuth token revocation
"""
from __future__ import annotations
import asyncio
import random
from celery.exceptions import MaxRetriesExceededError
from celery_app import celery_app
# ── Retry sentinels ───────────────────────────────────────────────────────────
class _TransientProviderError(Exception):
"""Raised by _run() to signal a retryable network/provider failure.
Escapes asyncio.run() and is caught by refresh_cloud_folder's sync layer,
which calls self.retry() in the Celery context (not inside asyncio.run).
"""
class _TerminalProviderError(Exception):
"""Raised by _run() for non-retryable auth/scope errors.
The task writes a warning state and does not retry.
"""
# ── Async worker ─────────────────────────────────────────────────────────────
async def _run(user_id: str, connection_id: str, parent_ref: str | None) -> dict:
"""Open a fresh DB session, revalidate owner, refresh and reconcile.
D-18/CACHE-01: never downloads file bytes, never alters quota.
D-13/D-14: on success → fresh state; on transient failure → raise sentinel
for Celery retry; on auth/scope failure → warning state + return.
"""
import uuid as _uuid
from db.session import AsyncSessionLocal
from services.cloud_items import (
ConnectionNotFound,
apply_listing_and_finalize,
get_or_create_folder_state,
reconcile_cloud_listing,
resolve_owned_connection,
update_folder_state,
)
from storage.cloud_backend_factory import build_cloud_resource_adapter
from storage.cloud_utils import decrypt_credentials
from config import settings
master_key = settings.cloud_creds_key.encode()
async with AsyncSessionLocal() as session:
# Revalidate ownership — worker never trusts broker payload alone
try:
conn = await resolve_owned_connection(
session,
connection_id=connection_id,
user_id=user_id,
)
except ConnectionNotFound:
# Connection deleted or user changed — silently drop, do not retry
return {"status": "skipped", "reason": "connection_not_found"}
# Mark refreshing so the browse endpoint can show spinner state
await update_folder_state(
session,
user_id=user_id,
connection_id=connection_id,
parent_ref=parent_ref or "",
refresh_state="refreshing",
)
await session.commit()
# Decrypt credentials inside the worker — never put them in broker payload
try:
credentials = decrypt_credentials(master_key, user_id, conn.credentials_enc)
except Exception as exc:
# Decryption failure is terminal — key mismatch, corruption, etc.
await update_folder_state(
session,
user_id=user_id,
connection_id=connection_id,
parent_ref=parent_ref or "",
refresh_state="warning",
error_code="credential_error",
error_message="Credential decryption failed. Re-connect the account.",
)
await session.commit()
raise _TerminalProviderError("Credential decryption failed. Re-connect the account.") from exc
# Build adapter and fetch listing
try:
adapter = build_cloud_resource_adapter(conn.provider, credentials)
conn_uuid = conn.id if isinstance(conn.id, _uuid.UUID) else _uuid.UUID(str(conn.id))
user_uuid = _uuid.UUID(str(user_id))
listing = await adapter.list_folder(
conn_uuid,
user_uuid,
parent_ref=parent_ref,
)
except Exception as exc:
err_str = str(exc).lower()
# Detect auth/scope errors — these are terminal
if any(kw in err_str for kw in ("invalid_grant", "unauthorized", "401", "403", "scope", "revoked")):
await update_folder_state(
session,
user_id=user_id,
connection_id=connection_id,
parent_ref=parent_ref or "",
refresh_state="warning",
error_code="auth_error",
error_message="Authentication failed. Re-connect the account.",
)
await session.commit()
raise _TerminalProviderError("Authentication failed. Re-connect the account.") from exc
# Transient failure — let Celery retry
raise _TransientProviderError("Provider error. Will retry.") from exc
# apply_listing_and_finalize is the single shared gate — it handles
# complete vs. incomplete semantics; only a complete authoritative
# listing transitions the folder to fresh state.
listing_result = await apply_listing_and_finalize(
session,
user_id=user_id,
connection_id=connection_id,
parent_ref=parent_ref,
listing=listing,
)
await session.commit()
if listing_result.is_fresh:
return {"status": "ok", "items_fetched": len(listing.items)}
else:
# Incomplete listing: task succeeded but folder is not fresh.
# Return a controlled status — no credentials, no provider item names.
return {
"status": "warning",
"warning_code": listing_result.warning_code,
"items_fetched": len(listing.items),
}
# ── Celery task ───────────────────────────────────────────────────────────────
@celery_app.task(
bind=True,
max_retries=3,
name="tasks.cloud_tasks.refresh_cloud_folder",
serializer="json",
acks_late=True,
reject_on_worker_lost=True,
)
def refresh_cloud_folder(self, user_id: str, connection_id: str, parent_ref: str | None) -> dict:
"""Refresh cloud metadata for (user_id, connection_id, parent_ref) idempotently.
D-13/D-14: Bounded transient retries with increasing countdown + jitter.
Terminal auth/scope failures mark warning state without retry.
D-18/CACHE-01: Never downloads file bytes; never alters quota.
"""
try:
return asyncio.run(_run(user_id, connection_id, parent_ref))
except _TerminalProviderError:
# Already wrote warning state in _run — do not retry
return {"status": "terminal_error"}
except _TransientProviderError as exc:
try:
# Bounded exponential backoff with jitter: 30s, 90s, 270s (± 10s)
jitter = random.randint(-10, 10)
countdown = (30 * (3 ** self.request.retries)) + jitter
raise self.retry(exc=exc, countdown=countdown)
except MaxRetriesExceededError:
# Write permanent warning state after all retries exhausted
asyncio.run(_write_final_warning(user_id, connection_id, parent_ref))
return {"status": "max_retries_exceeded"}
async def _write_final_warning(
user_id: str, connection_id: str, parent_ref: str | None
) -> None:
"""Write a terminal warning state after all Celery retries are exhausted."""
from db.session import AsyncSessionLocal
from services.cloud_items import update_folder_state
async with AsyncSessionLocal() as session:
await update_folder_state(
session,
user_id=user_id,
connection_id=connection_id,
parent_ref=parent_ref or "",
refresh_state="warning",
error_code="refresh_failed",
error_message="Background refresh failed after 3 attempts. Will retry on next browse.",
)
await session.commit()