fix(sec): warn when deferred imports stay stale
This commit is contained in:
@@ -30,7 +30,7 @@ import hashlib
|
||||
import json
|
||||
import logging
|
||||
from dataclasses import dataclass, field
|
||||
from datetime import date, datetime, timezone
|
||||
from datetime import date, datetime, timedelta, timezone
|
||||
from typing import Any, Protocol, runtime_checkable
|
||||
|
||||
from sqlalchemy import select, text
|
||||
@@ -69,9 +69,11 @@ class ValidationResult:
|
||||
summary: dict[str, Any] = field(default_factory=dict)
|
||||
source_max_date: date | None = None
|
||||
messages: list[str] = field(default_factory=list)
|
||||
# Expected source-side lag: retry next run without an operational alert.
|
||||
# This is only meaningful when ok=False.
|
||||
# Expected source-side lag: retry without an immediate error alert. Sources
|
||||
# can bound the quiet period with deferred_alert_after_days. Only meaningful
|
||||
# when ok=False.
|
||||
retryable: bool = False
|
||||
deferred_alert_after_days: int | None = None
|
||||
|
||||
|
||||
@runtime_checkable
|
||||
@@ -127,19 +129,41 @@ async def _last_promoted_revision(db: AsyncSession, source: str) -> str | None:
|
||||
return row.scalar_one_or_none()
|
||||
|
||||
|
||||
async def _has_promoted_since(db: AsyncSession, source: str, cutoff: datetime) -> bool:
|
||||
row = await db.execute(
|
||||
select(DataImportRun.id)
|
||||
.where(
|
||||
DataImportRun.source == source,
|
||||
DataImportRun.status == STATUS_PROMOTED,
|
||||
DataImportRun.started_at >= cutoff,
|
||||
)
|
||||
.limit(1)
|
||||
)
|
||||
return row.scalar_one_or_none() is not None
|
||||
|
||||
|
||||
def _now() -> datetime:
|
||||
return datetime.now(timezone.utc)
|
||||
|
||||
|
||||
async def _alert(db: AsyncSession, source: str, code: str, messages: list[str]) -> None:
|
||||
async def _alert(
|
||||
db: AsyncSession,
|
||||
source: str,
|
||||
code: str,
|
||||
messages: list[str],
|
||||
*,
|
||||
severity: str = "error",
|
||||
dedup_hours: int = 24,
|
||||
) -> None:
|
||||
try:
|
||||
await system_event_service.log_event(
|
||||
db,
|
||||
severity="error",
|
||||
severity=severity,
|
||||
source="data_import",
|
||||
code=f"{source}_{code}",
|
||||
message=(("; ".join(messages)) or code)[:_MAX_ERROR_LEN],
|
||||
dedup_key=f"data_import:{source}:{code}",
|
||||
dedup_hours=dedup_hours,
|
||||
)
|
||||
except Exception: # noqa: BLE001 — alerting must never mask the real outcome
|
||||
logger.exception("Failed to emit data_import alert %s/%s", source, code)
|
||||
@@ -153,7 +177,8 @@ async def run_import(
|
||||
) -> DataImportRun | None:
|
||||
"""Run one import for ``importer``.
|
||||
|
||||
Returns the recorded ``DataImportRun`` (promoted / no_op / failed), or None
|
||||
Returns the recorded ``DataImportRun`` (promoted / no_op / deferred /
|
||||
failed), or None
|
||||
when the per-source advisory lock is already held (another run is active).
|
||||
|
||||
``force`` runs even when the revision is unchanged. The revision tracks the
|
||||
@@ -217,6 +242,24 @@ async def run_import(
|
||||
if result.retryable:
|
||||
run.status = STATUS_DEFERRED
|
||||
await session.commit()
|
||||
alert_days = result.deferred_alert_after_days
|
||||
if alert_days is not None:
|
||||
alert_days = max(1, alert_days)
|
||||
cutoff = run.started_at - timedelta(days=alert_days)
|
||||
if not await _has_promoted_since(session, source, cutoff):
|
||||
await _alert(
|
||||
session,
|
||||
source,
|
||||
"deferred_stale",
|
||||
[
|
||||
f"No successful promotion for at least "
|
||||
f"{alert_days} day(s); deferred source lag may "
|
||||
f"now hide aged-out unresolved items: "
|
||||
f"{run.error_details or 'validation deferred'}"
|
||||
],
|
||||
severity="warning",
|
||||
dedup_hours=alert_days * 24,
|
||||
)
|
||||
logger.info(
|
||||
"data_import %s: deferred for retry: %s",
|
||||
source,
|
||||
|
||||
@@ -413,14 +413,15 @@ class SecFundamentalsImporter:
|
||||
summary=summary,
|
||||
source_max_date=self._latest_index_date,
|
||||
messages=messages,
|
||||
# Pure Company-Facts publication lag is expected and self-healing:
|
||||
# keep the cursor in place and retry, but do not page operators. Any
|
||||
# other missing reason or validation message remains a real failure.
|
||||
# Company-Facts absence is usually publication lag, but can also be a
|
||||
# permanent co-registrant misfile that the daily index did not expose.
|
||||
# Defer quietly at first; the framework warns if promotions stay stale.
|
||||
retryable=(
|
||||
len(messages) == 1
|
||||
and bool(blocking)
|
||||
and all(m.get("reason") == "not_in_companyfacts" for m in blocking)
|
||||
),
|
||||
deferred_alert_after_days=MISSING_XBRL_RETRY_DAYS,
|
||||
)
|
||||
|
||||
async def promote(self, db, staged: StagedFundamentals, run_id: int) -> dict[str, int]:
|
||||
|
||||
Reference in New Issue
Block a user