HealthLog/src/app/api/admin/backups/route.ts

/**
 * GET /api/admin/backups — admin-only listing of `DataBackup` rows.
 *
 * Returns one row per (userId, type) pair. The `data` payload is NOT
 * shipped by THIS route — only metadata (id, userId, username, type, size in
 * bytes, createdAt), so the listing costs a count and not a decrypt. The
 * payload has its own route: `/api/admin/backups/<id>/download` decrypts one
 * snapshot for the admin who asks for it, and the ciphertext stays in the
 * database either way.
 *
 * Alongside the rows, the response says whether the SCHEDULE is still alive:
 * how old the newest scheduled copy is and how the last scheduled run ended.
 * Without those two the page cannot tell a working backup from one that
 * stopped six weeks ago — every row it lists has a perfectly ordinary
 * timestamp either way.
 *
 * And alongside both, the off-host leg: one row per account saying when this
 * host last put that account's encrypted copy in the operator's bucket and how
 * stale that is against the nightly schedule. It comes from the ledger the
 * worker writes, never from a bucket listing — the answer has to hold on a host
 * whose worker holds no listing grant on the bucket.
 */
import { prisma } from "@/lib/db";
import { apiHandler, requireAdmin } from "@/lib/api-handler";
import { apiSuccess } from "@/lib/api-response";
import { annotate } from "@/lib/logging/context";
import { DATA_BACKUP_QUEUE } from "@/lib/jobs/data-backup-policy";
import {
  readLastQueueRun,
  readQueueRunningSince,
} from "@/lib/jobs/job-failures";
import { readBackupPassActivity } from "@/lib/jobs/backup-pass-attempts";
import { summariseBackupSchedule } from "@/lib/jobs/backup-schedule-status";
import {
  offhostBackupConfigured,
  probeOffhostLifecycle,
} from "@/lib/jobs/offhost-backup";
import { OFFHOST_BACKUP_QUEUE } from "@/lib/jobs/offhost-backup-queue";
import {
  classifyOffhostBackup,
  OFFHOST_BACKUP_PERIOD_HOURS,
  type OffhostBackupFreshness,
} from "@/lib/jobs/offhost-backup-freshness";
// v1.4.41 — `BackupRow` / `BackupsList` moved to `src/types/backups.ts`
// so callers (in particular `components/admin/backups-section.tsx`) don't
// have to reach across the component → route-handler layer boundary.
import type {
  BackupRow,
  BackupsList,
  OffhostAccountRow,
} from "@/types/backups";

export const dynamic = "force-dynamic";

/** The type the weekly worker writes; anything else was uploaded by hand. */
const SCHEDULED_BACKUP_TYPE = "WEEKLY_AUTO";

interface BackupMetaRow {
  id: string;
  user_id: string;
  username: string;
  type: string;
  size_bytes: number;
  created_at: Date;
}

export const GET = apiHandler(async () => {
  await requireAdmin();
  annotate({ action: { name: "admin.backups.list" } });

  // Metadata only, and the size read as `octet_length` in the database rather
  // than by pulling every blob into this process to measure it. The listing
  // needs a number, not the ciphertext, and on an instance with large records
  // fetching them all was its own way of running out of memory. A copy is
  // either one value in `data` (written before v1.39.2) or a run of pieces in
  // `data_backup_chunks`, so the size is whichever of the two it has: what
  // the copy occupies in the database.
  const rows = await prisma.$queryRaw<BackupMetaRow[]>`
    SELECT b.id, b.user_id, u.username, b.type,
           COALESCE(octet_length(b.data), 0)
             + COALESCE(c.bytes, 0) AS size_bytes,
           b.created_at
    FROM data_backups b
    JOIN users u ON u.id = b.user_id
    LEFT JOIN (
      SELECT backup_id, SUM(octet_length(data)) AS bytes
      FROM data_backup_chunks
      GROUP BY backup_id
    ) c ON c.backup_id = b.id
    ORDER BY b.created_at DESC
  `;

  const list: BackupRow[] = rows.map((row) => ({
    id: row.id,
    userId: row.user_id,
    username: row.username,
    type: row.type,
    sizeBytes: Number(row.size_bytes),
    createdAt: row.created_at.toISOString(),
  }));

  // The off-host leg, per account, read from the ledger the nightly worker
  // writes rather than from the bucket. The page must be able to answer "which
  // account has no recent copy off-host" on a host whose worker holds the
  // three grants the runbook documents and no listing grant at all, and a
  // listing call from a page render would put the operator's credentials on
  // the request path of every page view.
  const now = new Date();
  const offhostAccounts = await prisma.user.findMany({
    select: {
      id: true,
      username: true,
      offhostBackupState: {
        select: {
          lastAttemptAt: true,
          lastSuccessAt: true,
          sizeBytes: true,
        },
      },
    },
    orderBy: { username: "asc" },
  });

  // Evidence beats this process's environment. `offhostBackupConfigured()`
  // reads the variables of whichever process serves this request, and a
  // hand-rolled split that gives `BACKUP_S3_*` only to the worker would make
  // the web process say "nothing leaves this host" over a ledger full of
  // fresh rows. A row in that table is proof an upload happened; no
  // environment read can outrank it.
  const configured =
    offhostBackupConfigured() ||
    offhostAccounts.some((account) => account.offhostBackupState !== null);

  const offhostRows: OffhostAccountRow[] = offhostAccounts.map((account) => {
    const state = account.offhostBackupState;
    const verdict = classifyOffhostBackup({
      lastAttemptAt: state?.lastAttemptAt ?? null,
      lastSuccessAt: state?.lastSuccessAt ?? null,
      now,
    });
    return {
      userId: account.id,
      username: account.username,
      lastAttemptAt: state?.lastAttemptAt.toISOString() ?? null,
      lastSuccessAt: state?.lastSuccessAt?.toISOString() ?? null,
      // BigInt column -> number on the wire. The value is bytes of one
      // object, capped by the uploader at 80 GB, so it is nowhere near
      // Number.MAX_SAFE_INTEGER and JSON has no BigInt.
      sizeBytes:
        state?.sizeBytes === null || state?.sizeBytes === undefined
          ? null
          : Number(state.sizeBytes),
      ageHours: verdict.ageHours,
      freshness: verdict.freshness,
    };
  });

  // Worst first. Sorting by username puts the one account that needs
  // attention wherever the alphabet happens to leave it, which on a
  // hundred-account cohort is the same "one bad row hides in the crowd"
  // problem this card was built to end, moved out of a count and into a list.
  const FRESHNESS_ORDER: Record<OffhostBackupFreshness, number> = {
    stale: 0,
    due: 1,
    never: 2,
    unknown: 3,
    fresh: 4,
  };
  offhostRows.sort(
    (a, b) =>
      FRESHNESS_ORDER[a.freshness] - FRESHNESS_ORDER[b.freshness] ||
      a.username.localeCompare(b.username),
  );

  // A wipe whose copies are already gone waits for the next nightly run
  // before its request is removed (see `offhost-purge.ts`); nothing of it is
  // left in the bucket, so it is not counted as a pending deletion.
  const stillInBucket = {
    NOT: {
      reason: "data_wiped",
      lastFailure: null,
      lastAttemptAt: { not: null },
    },
  };
  const [weeklyRunningSince, offhostRunningSince] = await Promise.all([
    readQueueRunningSince(DATA_BACKUP_QUEUE),
    readQueueRunningSince(OFFHOST_BACKUP_QUEUE),
  ]);
  const [scheduleActivity, offhostActivity] = await Promise.all([
    readBackupPassActivity(prisma, DATA_BACKUP_QUEUE, weeklyRunningSince),
    readBackupPassActivity(prisma, OFFHOST_BACKUP_QUEUE, offhostRunningSince),
  ]);
  const [purgeCount, oldestPurge, lifecycle] = await Promise.all([
    prisma.offhostPurgeRequest.count({ where: stillInBucket }),
    prisma.offhostPurgeRequest.findFirst({
      where: stillInBucket,
      orderBy: { requestedAt: "asc" },
      select: { requestedAt: true, lastFailure: true },
    }),
    configured
      ? probeOffhostLifecycle()
      : Promise.resolve({ state: "unknown" as const, expirationDays: null }),
  ]);

  const payload: BackupsList = {
    rows: list,
    // Matches the retention window the backup-prune job enforces so the
    // backups page states the same number the worker acts on.
    retentionDays: 90,
    schedule: summariseBackupSchedule({
      scheduledCreatedAt: rows
        .filter((row) => row.type === SCHEDULED_BACKUP_TYPE)
        .map((row) => row.created_at),
      lastRun: await readLastQueueRun(DATA_BACKUP_QUEUE),
      now,
    }),
    scheduleActivity,
    offhost: {
      configured,
      activity: offhostActivity,
      periodHours: OFFHOST_BACKUP_PERIOD_HOURS,
      // An unconfigured host has nothing to list, and the card says so rather
      // than painting an empty table that reads as "no problems".
      rows: configured ? offhostRows : [],
      pendingDeletions: {
        count: purgeCount,
        oldestRequestedAt: oldestPurge?.requestedAt.toISOString() ?? null,
        lastFailure: oldestPurge?.lastFailure ?? null,
      },
      lifecycle,
    },
  };

  return apiSuccess(payload);
});