HoruxDespachosNuevo/apps/api/src/services/sat/sweep-stale-jobs.service.ts

import { prisma } from '../../config/database.js';

export interface SweepResult {
  pendingFound: number;
  runningFound: number;
  pendingMarked: number;
  runningMarked: number;
  entries: Array<{
    id: string;
    tenantId: string;
    kind: 'pending-stale' | 'running-stale';
    ageHours: number;
  }>;
}

const DEFAULT_RUNNING_HOURS_BY_TYPE: Record<string, number> = {
  initial: 24,
  daily: 4,
  incremental: 2,
  custom: 24,
};

/**
 * Watchdog para jobs `sat_sync_jobs` stale.
 *
 * Categorías:
 *   1. `pending` con `nextRetryAt` > pendingHours atrás. El cron horario
 *      `retryTimedOutJobs` normalmente los retoma, pero si no arranca
 *      (dev, caída, reinicio largo) el job queda colgado y bloquea el
 *      lock para nuevos syncs del mismo (tenant, contribuyente).
 *
 *   2. `running` con `startedAt` > runningHours atrás. Thresholds difieren
 *      por tipo: initial (8h) porque un bootstrap de 6 años puede tardar
 *      varias horas; daily (4h); incremental (2h) porque es ventana corta.
 *      Si lleva >threshold es casi seguro huérfano de un proceso que murió.
 *
 * Marca ambos como `failed` con `errorMessage` descriptivo. Idempotente
 * (volver a correrlo no reabre los ya-marcados-failed).
 *
 * - `apply=false` (default): dry-run, no toca BD.
 * - `pendingHours`: threshold pending (default 12h).
 * - `runningHours`: fallback threshold running si no se usa por-tipo (default 4h).
 * - `runningHoursByType`: override por tipo de sync.
 */
export async function sweepStaleSatJobs(params: {
  apply: boolean;
  pendingHours?: number;
  runningHours?: number;
  runningHoursByType?: Record<string, number>;
} = { apply: false }): Promise<SweepResult> {
  const pendingHours = params.pendingHours ?? 12;
  const runningHoursByType = { ...DEFAULT_RUNNING_HOURS_BY_TYPE, ...(params.runningHoursByType || {}) };
  const now = new Date();
  const pendingCutoff = new Date(now.getTime() - pendingHours * 3600 * 1000);

  const stalePending = await prisma.satSyncJob.findMany({
    where: { status: 'pending', nextRetryAt: { lt: pendingCutoff } },
    orderBy: { createdAt: 'asc' },
  });

  // running: evaluar por tipo usando thresholds distintos
  const allRunning = await prisma.satSyncJob.findMany({
    where: { status: 'running' },
    orderBy: { createdAt: 'asc' },
  });
  const staleRunning = allRunning.filter(j => {
    const thresholdHours = runningHoursByType[j.type] ?? params.runningHours ?? 4;
    const cutoff = new Date(now.getTime() - thresholdHours * 3600 * 1000);
    return (j.startedAt ?? j.createdAt) < cutoff;
  });

  const result: SweepResult = {
    pendingFound: stalePending.length,
    runningFound: staleRunning.length,
    pendingMarked: 0,
    runningMarked: 0,
    entries: [],
  };

  for (const j of stalePending) {
    const ageHours = Math.round((now.getTime() - (j.nextRetryAt ?? j.createdAt).getTime()) / 3_600_000);
    result.entries.push({ id: j.id, tenantId: j.tenantId, kind: 'pending-stale', ageHours });
  }
  for (const j of staleRunning) {
    const ageHours = Math.round((now.getTime() - (j.startedAt ?? j.createdAt).getTime()) / 3_600_000);
    result.entries.push({ id: j.id, tenantId: j.tenantId, kind: 'running-stale', ageHours });
  }

  if (!params.apply) return result;

  for (const j of stalePending) {
    await prisma.satSyncJob.update({
      where: { id: j.id },
      data: {
        status: 'failed',
        completedAt: now,
        errorMessage: `Abandoned by watchdog: pending with nextRetryAt ${j.nextRetryAt?.toISOString()} > ${pendingHours}h in the past. Retry cron didn't pick it up.`,
      },
    });
    result.pendingMarked++;
  }
  for (const j of staleRunning) {
    const thresholdHours = runningHoursByType[j.type] ?? params.runningHours ?? 4;
    await prisma.satSyncJob.update({
      where: { id: j.id },
      data: {
        status: 'failed',
        completedAt: now,
        errorMessage: `Abandoned by watchdog: running ${j.type} with startedAt ${j.startedAt?.toISOString()} > ${thresholdHours}h (process crash / orphan). SAT request is lost; re-launch manually.`,
      },
    });
    result.runningMarked++;
  }

  return result;
}