Files
rf4-spotter/apps/api/app/community_scheduler.py

139 lines
6.7 KiB
Python

from __future__ import annotations
import logging
import time
from dataclasses import asdict
from datetime import datetime, timedelta, timezone
from sqlalchemy import select, text
from rf4_research.community_cli import SOURCES, fetch_html, fetch_site_key
from rf4_research.community_sources import parse_rf4map_point, parse_rf4posts_spot
from .community_importer import stage_observations
from .config import settings
from .database import SessionLocal
from .logging_config import configure_logging
from .models import CommunityImportRun, DataSource
from .source_lifecycle import classify_source_failure, record_scheduled_source_check
logger = logging.getLogger("rf4.community_scheduler")
MAX_BACKOFF_SECONDS = 24 * 60 * 60
def retry_delay(statuses: list[str]) -> int:
failures = 0
for status in statuses:
if status != "failed":
break
failures += 1
return min(settings.community_import_interval_seconds * (2 ** max(0, failures - 1)), MAX_BACKOFF_SECONDS)
def _static_registry() -> dict[str, tuple[str, callable]]:
"""Return the full static registry without DB access (for unit tests)."""
return {
"rf4db": SOURCES["rf4db"],
"rf4stat-fishing": SOURCES["rf4stat-fishing"],
"rf4stat-post": (SOURCES["rf4stat-posts"][0], SOURCES["rf4stat-posts"][1]),
"rf4map": (settings.rf4map_point_url, parse_rf4map_point),
"rf4posts-spot": (settings.rf4posts_spot_url, parse_rf4posts_spot),
}
def configured_sources(enabled_keys: set[str] | None = None) -> dict[str, tuple[str, callable]]:
"""Return enabled sources. When enabled_keys is None, query the DB."""
registry = _static_registry()
if enabled_keys is None:
with SessionLocal() as session:
enabled_keys = {
s.key for s in session.scalars(select(DataSource).where(DataSource.enabled.is_(True)))
}
return {k: v for k, v in registry.items() if k in enabled_keys}
def oldest_site_source(source_system: str, latest_by_source: dict[str, datetime], enabled_keys: set[str] | None = None) -> str:
"""Return the oldest candidate for the same site.
Uses the full registry (not just enabled) for cooldown history so that
disabling an endpoint does not reset the site-wide cooldown for its
neighbours. enabled_keys is used to filter candidates after the oldest
is found — if the oldest is disabled, the next oldest enabled is picked.
"""
registry = _static_registry()
site = fetch_site_key(registry[source_system][0])
candidates = [key for key, (url, _) in registry.items() if fetch_site_key(url) == site]
order = {key: index for index, key in enumerate(candidates)}
# Sort by (last_run, order) — pick oldest
sorted_candidates = sorted(candidates, key=lambda key: (latest_by_source.get(key, datetime.min.replace(tzinfo=timezone.utc)), order[key]))
# If enabled_keys is provided, prefer enabled; otherwise return oldest regardless
if enabled_keys:
enabled = [k for k in sorted_candidates if k in enabled_keys]
if enabled:
return enabled[0]
return sorted_candidates[0]
def run_source(source_system: str, *, now: datetime | None = None) -> bool:
current = now or datetime.now(timezone.utc)
with SessionLocal() as session:
enabled_keys = {s.key for s in session.scalars(select(DataSource).where(DataSource.enabled.is_(True)))}
registry = _static_registry()
url, parser = registry[source_system]
site_key = fetch_site_key(url)
site_sources = [key for key, (candidate_url, _) in registry.items() if fetch_site_key(candidate_url) == site_key]
with SessionLocal() as session:
source = session.get(DataSource, source_system)
if source is None or not source.enabled:
return False
# Lock before reading cooldown: committing the reservation makes it visible
# to the next contender before releasing this transaction lock.
if session.bind and session.bind.dialect.name == "postgresql" and not session.scalar(text("select pg_try_advisory_xact_lock(hashtext(:key))"), {"key": f"community-site:{site_key}"}):
return False
recent = list(session.scalars(select(CommunityImportRun).where(CommunityImportRun.source_system.in_(site_sources)).order_by(CommunityImportRun.started_at.desc()).limit(32)))
latest_by_source: dict[str, datetime] = {}
for previous in recent:
latest_by_source.setdefault(previous.source_system, previous.started_at if previous.started_at.tzinfo else previous.started_at.replace(tzinfo=timezone.utc))
if oldest_site_source(source_system, latest_by_source, enabled_keys) != source_system:
return False
latest = recent[0].started_at if recent else None
delay = retry_delay([run.status for run in recent])
if latest and (latest if latest.tzinfo else latest.replace(tzinfo=timezone.utc)) > current - timedelta(seconds=delay):
return False
run = CommunityImportRun(source_system=source_system, source_url=url, started_at=current, status="running")
session.add(run); session.commit()
try:
html = fetch_html(url)
records = parser(html, source_url=url) if source_system in {"rf4map", "rf4posts-spot"} else parser(html)
created, updated = stage_observations(session, [asdict(item) for item in records])
record_scheduled_source_check(
session, source_system=source_system, source_url=url,
status="available", checked_at=current,
)
run.status, run.rows_seen, run.rows_created, run.rows_updated = "success", len(records), created, updated
except Exception as exc:
session.rollback()
source_status = classify_source_failure(exc)
checked = datetime.now(timezone.utc)
affected = record_scheduled_source_check(
session,
source_system=source_system,
source_url=url,
status=source_status,
checked_at=checked,
)
run.status, run.error_summary = "failed", f"{type(exc).__name__}: {str(exc)[:500]}"
logger.exception("community import failed", extra={
"event":"community_import_failed", "source_system":source_system,
"source_check_status": source_status, "affected_observations": affected,
})
run.finished_at = datetime.now(timezone.utc); session.commit()
return True
def main() -> None:
configure_logging(settings.log_level)
while True:
for source_system in configured_sources():
run_source(source_system)
time.sleep(60)
if __name__ == "__main__":
main()