eptm_dashboard/scripts/cron_tick.py

#!/usr/bin/env python3
"""Cron tick — appelé toutes les minutes par cron OS via docker exec.

Logique :
1. Lire tous les CronJob.enabled = True
2. Pour chaque job, calculer s'il est dû maintenant (basé sur schedule + last_run_at)
3. Si dû :
   - Si déjà running (PID alive) → kill -9 (politique "kill")
   - Lancer subprocess (push, sync, ou push_then_sync)
   - Stream stdout+stderr dans un fichier log dédié
   - Mettre à jour last_status / last_message / last_run_at / last_pid
4. Envoyer notification Telegram selon notify_on

Le timezone effectif est celui du container (Europe/Zurich attendu).

Le script est idempotent et safe : si déjà passé pour un job aujourd'hui,
ne le relance pas. Si trop tard (>5 min après le slot), saute (un cron raté
n'est pas rattrapé).

Usage :
    python3 /app/scripts/cron_tick.py
    python3 /app/scripts/cron_tick.py --dry-run    # affiche ce qui serait lancé
    python3 /app/scripts/cron_tick.py --job <id>   # force un job précis
"""

from __future__ import annotations

import argparse
import json
import os
import signal
import subprocess
import sys
import time
from datetime import datetime, timedelta
from pathlib import Path

_ROOT = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(_ROOT))

from src.db import CronJob, get_session, init_db, get_engine  # noqa: E402
from src.notifier import notify_job_result  # noqa: E402

# Logs cron : par défaut /logs/cron (bind mount persistent), override via env var.
LOG_DIR = Path(os.getenv("CRON_LOG_DIR", "/logs/cron"))
try:
    LOG_DIR.mkdir(parents=True, exist_ok=True)
except Exception:
    # Fallback si /logs n'existe pas (ex: exécution hors container)
    LOG_DIR = _ROOT / "logs" / "cron"
    LOG_DIR.mkdir(parents=True, exist_ok=True)

SCRIPT_SYNC = _ROOT / "scripts" / "sync_esacada.py"
SCRIPT_PUSH = _ROOT / "scripts" / "push_to_escada.py"
SCRIPT_PUSH_NOTICES = _ROOT / "scripts" / "push_notices.py"
DATA_DIR    = _ROOT / "data"

# Marqueur écrit par run_imports.py à la fin des imports en DB
RUN_IMPORTS_RESULT = DATA_DIR / "sync_last_result.json"

# Timeout d'attente de run_imports après que sync_esacada.py exit
RUN_IMPORTS_TIMEOUT_SEC = 15 * 60   # 15 min, large mais raisonnable

# Slot de tolérance : si le cron OS rate une minute (charge, restart),
# on accepte d'exécuter dans la fenêtre [HH:MM, HH:MM+5min].
_SLOT_TOLERANCE_MIN = 5

_DAY_NAMES = ["MON", "TUE", "WED", "THU", "FRI", "SAT", "SUN"]


# ── Schedule logic ────────────────────────────────────────────────────────────

def _is_due(job: CronJob, now: datetime) -> bool:
    """Détermine si le job doit être lancé maintenant."""
    if not job.enabled:
        return False

    last = job.last_run_at

    if job.schedule_kind == "interval":
        # schedule_value = nb minutes
        try:
            minutes = int(job.schedule_value)
        except (TypeError, ValueError):
            return False
        if minutes < 1:
            return False
        if last is None:
            return True
        return (now - last).total_seconds() >= minutes * 60

    if job.schedule_kind == "daily":
        # schedule_value = "HH:MM"
        return _due_time_of_day(job.schedule_value, last, now)

    if job.schedule_kind == "weekly":
        # schedule_value = "MON,WED,FRI:HH:MM"
        try:
            days_part, time_part = job.schedule_value.split(":", 1)
        except ValueError:
            return False
        days = {d.strip().upper() for d in days_part.split(",") if d.strip()}
        today_name = _DAY_NAMES[now.weekday()]
        if today_name not in days:
            return False
        return _due_time_of_day(time_part, last, now)

    return False


def _due_time_of_day(hhmm: str, last: datetime | None, now: datetime) -> bool:
    """True si l'heure actuelle est dans la fenêtre [HH:MM, HH:MM+tolerance]
    et que le job n'a pas déjà tourné aujourd'hui."""
    try:
        hh, mm = hhmm.split(":")
        target = now.replace(hour=int(hh), minute=int(mm), second=0, microsecond=0)
    except (ValueError, AttributeError):
        return False
    delta = (now - target).total_seconds()
    if delta < 0 or delta > _SLOT_TOLERANCE_MIN * 60:
        return False
    if last is not None and last.date() == now.date() and last >= target:
        return False
    return True


# ── Process management ───────────────────────────────────────────────────────

def _pid_alive(pid: int | None) -> bool:
    if not pid:
        return False
    try:
        os.kill(pid, 0)
        return True
    except (ProcessLookupError, PermissionError):
        return False
    except Exception:
        return False


def _kill_pid(pid: int) -> None:
    try:
        os.killpg(os.getpgid(pid), signal.SIGKILL)
    except Exception:
        try:
            os.kill(pid, signal.SIGKILL)
        except Exception:
            pass


# ── Build command lines ──────────────────────────────────────────────────────

def _classes_args(job: CronJob) -> list[str]:
    """Retourne la liste des classes à passer aux scripts. Vide = toutes."""
    raw = (job.classes_json or "").strip()
    if not raw or raw == "ALL":
        return []
    try:
        lst = json.loads(raw)
        if isinstance(lst, list):
            return [str(c) for c in lst]
    except Exception:
        pass
    return []


def _build_sync_cmd(job: CronJob) -> list[str]:
    classes = _classes_args(job)
    cmd = [sys.executable, str(SCRIPT_SYNC), "--sync-all", *classes]
    if not job.sync_abs:    cmd.append("--skip-abs")
    if not job.sync_bn:     cmd.append("--skip-bn")
    if not job.sync_notes:  cmd.append("--skip-notes")
    if not job.sync_fiches: cmd.append("--skip-fiches")
    if job.force_abs:       cmd.append("--force-abs")
    return cmd


def _build_push_cmd(job: CronJob) -> list[str]:
    return [sys.executable, str(SCRIPT_PUSH)]


def _wait_for_run_imports(log_fp, mtime_before: float) -> tuple[bool, str, dict]:
    """Après que sync_esacada.py a fini, run_imports.py tourne en sous-process
    détaché. Attend que sync_last_result.json soit mis à jour, puis log les
    résultats détaillés. Retourne (ok, summary_message, raw_result_dict)."""
    log_fp.write("\n━━━ Attente run_imports (subprocess détaché) ━━━\n")
    log_fp.flush()

    deadline = time.time() + RUN_IMPORTS_TIMEOUT_SEC
    poll_count = 0
    while time.time() < deadline:
        if RUN_IMPORTS_RESULT.exists() and RUN_IMPORTS_RESULT.stat().st_mtime > mtime_before:
            break
        poll_count += 1
        # Log un point tous les 30 polls (~1 min) pour montrer qu'on attend
        if poll_count % 30 == 0:
            elapsed = int(time.time() - (deadline - RUN_IMPORTS_TIMEOUT_SEC))
            log_fp.write(f"[poll] {elapsed}s écoulés, en attente…\n")
            log_fp.flush()
        time.sleep(2)
    else:
        log_fp.write("⚠ TIMEOUT — sync_last_result.json non mis à jour dans le délai\n")
        log_fp.flush()
        return False, "run_imports timeout (>15min sans résultat)", {}

    # Lire le résultat
    try:
        result = json.loads(RUN_IMPORTS_RESULT.read_text(encoding="utf-8"))
    except Exception as e:
        log_fp.write(f"⚠ Impossible de lire sync_last_result.json : {e}\n")
        return False, f"sync_last_result.json illisible : {e}", {}

    res_abs   = result.get("res_abs",   []) or []
    res_bn    = result.get("res_bn",    []) or []
    res_notes = result.get("res_notes", []) or []
    res_matu  = result.get("res_matu",  []) or []
    errors    = result.get("errors",    []) or []
    ts        = result.get("timestamp", "?")

    log_fp.write(f"run_imports terminé (timestamp {ts})\n")
    log_fp.write(f"  Absences PDF importés      : {len(res_abs)}\n")
    log_fp.write(f"  Bulletins de notes         : {len(res_bn)}\n")
    log_fp.write(f"  Notes d'examen             : {len(res_notes)}\n")
    log_fp.write(f"  Notes Matu                 : {len(res_matu)}\n")
    log_fp.write(f"  Erreurs                    : {len(errors)}\n")

    # Détailler chaque catégorie si non vide
    if res_abs:
        log_fp.write("\n  Détail absences :\n")
        for r in res_abs:
            log_fp.write(f"    - {r}\n")
    if res_bn:
        log_fp.write("\n  Détail BN :\n")
        for r in res_bn:
            log_fp.write(f"    - {r}\n")
    if res_notes:
        log_fp.write("\n  Détail notes d'examen :\n")
        for r in res_notes:
            log_fp.write(f"    - {r}\n")
    if res_matu:
        log_fp.write("\n  Détail Matu :\n")
        for r in res_matu:
            log_fp.write(f"    - {r}\n")
    if errors:
        log_fp.write("\n  ❌ ERREURS :\n")
        for err in errors:
            log_fp.write(f"    - {err}\n")
    log_fp.flush()

    if errors:
        summary = (
            f"Imports : abs={len(res_abs)}, bn={len(res_bn)}, "
            f"notes={len(res_notes)}, matu={len(res_matu)}, "
            f"⚠ {len(errors)} erreur(s)"
        )
        return False, summary, result

    summary = (
        f"Imports OK : abs={len(res_abs)}, bn={len(res_bn)}, "
        f"notes={len(res_notes)}, matu={len(res_matu)}"
    )
    return True, summary, result


# ── Run a single step ────────────────────────────────────────────────────────

def _run_step(cmd: list[str], log_fp, title: str) -> tuple[int, int]:
    """Lance une commande, stream stdout+stderr dans log_fp.
    Retourne (returncode, pid)."""
    log_fp.write(f"\n━━━ {title} ━━━\n")
    log_fp.write(f"$ {' '.join(cmd)}\n")
    log_fp.flush()

    proc = subprocess.Popen(
        cmd,
        stdout=log_fp,
        stderr=subprocess.STDOUT,
        env={**os.environ, "PYTHONUNBUFFERED": "1"},
        start_new_session=True,
    )
    pid = proc.pid
    rc = proc.wait()
    log_fp.write(f"\n[exit code = {rc}]\n")
    log_fp.flush()
    return rc, pid


# ── Run a job (full lifecycle) ───────────────────────────────────────────────

def run_job(job: CronJob, sess) -> None:
    """Exécute un job. Met à jour son état en DB et envoie notification."""
    started = datetime.now()
    ts = started.strftime("%Y%m%d-%H%M%S")
    log_path = LOG_DIR / f"job_{job.id}_{ts}.log"

    # Politique "kill" : si déjà running (selon DB) et PID alive, on kill avant.
    if job.last_status == "running" and _pid_alive(job.last_pid):
        _kill_pid(job.last_pid or 0)
        # Trace dans le log
        with log_path.open("w", encoding="utf-8") as fp:
            fp.write(f"[{started}] PID précédent {job.last_pid} kill -9 (politique kill)\n")

    # Marquer running
    job.last_run_at   = started
    job.last_status   = "running"
    job.last_message  = ""
    job.last_log_path = str(log_path)
    job.last_pid      = None
    sess.commit()

    overall_rc = 0
    final_msg = ""
    last_pid: int | None = None
    imports_result: dict = {}

    try:
        with log_path.open("a", encoding="utf-8") as fp:
            fp.write(f"\n=== Job #{job.id} '{job.name}' — démarré {started.isoformat(timespec='seconds')} ===\n")
            fp.write(f"task_kind={job.task_kind}  classes={job.classes_json}\n")

            steps: list[tuple[str, list[str]]] = []
            if job.task_kind == "push":
                steps = [("Push Escada", _build_push_cmd(job))]
            elif job.task_kind == "sync":
                steps = [("Sync Escada", _build_sync_cmd(job))]
            elif job.task_kind == "push_then_sync":
                steps = [
                    ("Push Escada", _build_push_cmd(job)),
                    ("Sync Escada", _build_sync_cmd(job)),
                ]
            elif job.task_kind == "push_notices":
                steps = [("Push notices", [sys.executable, str(SCRIPT_PUSH_NOTICES)])]
            else:
                fp.write(f"[error] task_kind inconnu : {job.task_kind}\n")
                overall_rc = 99
                final_msg = f"task_kind invalide : {job.task_kind}"

            for title, cmd in steps:
                # Capturer mtime du marqueur run_imports AVANT le sync
                # (utilisé après pour détecter la fin de run_imports.py)
                is_sync = title.startswith("Sync")
                mtime_before = (
                    RUN_IMPORTS_RESULT.stat().st_mtime
                    if is_sync and RUN_IMPORTS_RESULT.exists() else 0.0
                )

                rc, pid = _run_step(cmd, fp, title)
                last_pid = pid
                if rc != 0:
                    overall_rc = rc
                    final_msg = f"{title} a échoué (code {rc})"
                    break

                # Si c'était une étape sync, attendre que run_imports termine
                if is_sync:
                    imports_ok, imports_msg, imports_result = _wait_for_run_imports(fp, mtime_before)
                    if not imports_ok:
                        overall_rc = 2
                        final_msg = imports_msg
                        break
                    # On garde le message du sub pour la notif finale
                    final_msg = imports_msg

            if overall_rc == 0 and not final_msg:
                final_msg = f"{len(steps)} étape(s) OK"

    except Exception as e:
        overall_rc = 1
        final_msg = f"Exception : {e}"
        try:
            with log_path.open("a", encoding="utf-8") as fp:
                import traceback
                fp.write("\n[fatal exception]\n")
                fp.write(traceback.format_exc())
        except Exception:
            pass

    # État final en DB
    finished = datetime.now()
    duration = (finished - started).total_seconds()
    job.last_status   = "ok" if overall_rc == 0 else "fail"
    job.last_message  = final_msg
    job.last_pid      = last_pid
    sess.commit()

    # Notification
    try:
        notify_job_result(
            job_name=job.name,
            status=job.last_status,
            message=final_msg,
            log_path=log_path,
            chat_id=job.notify_chat_id or None,
            notify_on=job.notify_on,
            notify_level=getattr(job, "notify_level", "normal"),
            duration_s=duration,
            details=imports_result,
            job_options={
                "task_kind":   job.task_kind,
                "sync_abs":    job.sync_abs,
                "sync_bn":     job.sync_bn,
                "sync_notes":  job.sync_notes,
                "sync_fiches": job.sync_fiches,
            },
        )
    except Exception as e:
        # Ne fait pas échouer le job pour une notif KO
        try:
            with log_path.open("a", encoding="utf-8") as fp:
                fp.write(f"\n[notify] échec envoi : {e}\n")
        except Exception:
            pass


# ── Main loop ────────────────────────────────────────────────────────────────

def main() -> int:
    parser = argparse.ArgumentParser(description="Cron tick — exécute les CronJob dûs.")
    parser.add_argument("--dry-run", action="store_true",
                        help="Liste les jobs dûs sans les exécuter.")
    parser.add_argument("--job", type=int, default=None,
                        help="Force l'exécution d'un job par son id (ignore schedule).")
    args = parser.parse_args()

    # Garantir que la table existe
    try:
        init_db()
    except Exception as e:
        print(f"[cron_tick] init_db error : {e}", file=sys.stderr)
        return 1

    sess = get_session()
    now = datetime.now()

    try:
        if args.job is not None:
            job = sess.get(CronJob, args.job)
            if job is None:
                print(f"[cron_tick] job id={args.job} introuvable")
                return 1
            print(f"[cron_tick] forçage job #{job.id} '{job.name}'")
            if args.dry_run:
                return 0
            run_job(job, sess)
            return 0

        from sqlalchemy import select as _sel
        jobs = sess.execute(_sel(CronJob).where(CronJob.enabled == True)).scalars().all()  # noqa: E712

        due_jobs = [j for j in jobs if _is_due(j, now)]
        if not due_jobs:
            return 0  # rien à faire

        print(f"[cron_tick] {now.isoformat(timespec='seconds')} — {len(due_jobs)} job(s) dûs")
        for job in due_jobs:
            print(f"  - #{job.id} '{job.name}' kind={job.task_kind} schedule={job.schedule_kind}:{job.schedule_value}")
            if args.dry_run:
                continue
            run_job(job, sess)

        return 0
    finally:
        sess.close()


if __name__ == "__main__":
    sys.exit(main())