Files
Kisani/Services/BackupEngine.swift
Robin Kutesa 0169486c72 fix: manifest out of sync with NAS causes gallery to show stale item count
Two root causes:
1. Backup interrupted before loop completion = manifest write never fires,
   leaving uploaded files as NAS orphans invisible to gallery and dedup.
2. Manifest only written once at loop end, so any crash loses all progress.

Fixes:
- On each backup run, list the NAS directory and merge any files not tracked
  in the manifest (orphaned from prior interrupted sessions). The healed
  manifest is written back to NAS and NASManifestCache before uploads start,
  so the gallery reflects actual NAS contents immediately on next open.
- Incremental checkpoint writes every 25 uploads using the already-open
  transfer connection — interrupted backups now preserve progress in blocks
  of 25 instead of losing everything.

Co-Authored-By: Kutesir <kutesir@provoc.ug>
Co-Authored-By: Sentry <sentry@provoc.ug>
2026-05-18 00:11:10 +03:00

427 lines
19 KiB
Swift
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

import Foundation
import Photos
import os.log
private let logger = Logger(subsystem: "com.albert.nasbackup", category: "BackupEngine")
private let signposter = OSSignposter(subsystem: "com.albert.nasbackup", category: "BackupEngine")
@MainActor
final class BackupEngine: ObservableObject {
static let shared = BackupEngine()
@Published private(set) var job: BackupJob = BackupJob()
@Published private(set) var queueItems: [BackupQueueItem] = []
private let photoService: PhotoLibraryProtocol
private let transferFactory: (NASProtocol) -> any NASTransferProtocol
private var activeTransfer: (any NASTransferProtocol)?
private var isCancelled = false
private static let queueCacheURL: URL = {
let caches = FileManager.default.urls(for: .cachesDirectory, in: .userDomainMask)[0]
return caches.appendingPathComponent("kisani_last_queue.json")
}()
private init(
photos: PhotoLibraryProtocol = PhotoLibraryService(),
transferFactory: @escaping (NASProtocol) -> any NASTransferProtocol = { proto in
switch proto {
case .smb: return SMBService()
case .sftp: return SFTPService()
}
}
) {
self.photoService = photos
self.transferFactory = transferFactory
// Load last session's queue so SyncView has data to show on next launch.
if let data = try? Data(contentsOf: Self.queueCacheURL),
let items = try? JSONDecoder.kisani.decode([BackupQueueItem].self, from: data) {
self.queueItems = items
}
}
static func testInstance(
transfer: any NASTransferProtocol,
photos: PhotoLibraryProtocol
) -> BackupEngine {
BackupEngine(photos: photos, transferFactory: { _ in transfer })
}
func run(
connection: NASConnection,
filter: BackupFilter,
triggeredByLAN: Bool = false
) async throws -> BackupResult {
isCancelled = false
let startDate = Date()
let store = ConnectionStore.shared
let lan = LANMonitor.shared
if !triggeredByLAN && lan.isOnCellular && !store.allowCellularBackup {
throw BackupError.networkUnavailable
}
let host = resolveHost(connection: connection, store: store, lan: lan)
// 1. Fetch assets off the main actor
// fetchAssets() enumerates every PHAsset and builds a [PhotoAsset] array.
// For a library of 10 k+ assets this is 100500 ms of synchronous work.
// Running it in a detached task keeps the main thread free for animations.
job.prepare()
let photoSvc = self.photoService
let spFetch = signposter.beginInterval("FetchAssets")
let assets: [PhotoAsset] = await Task.detached(priority: .userInitiated) {
photoSvc.fetchAssets(filter: filter)
}.value
signposter.endInterval("FetchAssets", spFetch, "\(assets.count) assets")
logger.info("Photo fetch: \(assets.count) assets in library")
try Task.checkCancellation()
// 2. Connect to NAS
// Network I/O suspends the main actor UI stays live the whole time.
let transfer = transferFactory(connection.nasProtocol)
let spConnect = signposter.beginInterval("NASConnect")
try await transfer.connect(
to: host, port: connection.port,
username: connection.username, password: connection.password
)
signposter.endInterval("NASConnect", spConnect)
activeTransfer = transfer
defer { transfer.disconnect(); activeTransfer = nil }
try Task.checkCancellation()
// 3. Load manifest + heal orphaned files
// After any interrupted backup, files may sit on NAS without a manifest
// entry. We list the directory and merge any untracked files so the
// gallery and dedup logic reflect actual NAS contents.
let manifestPath = "\(connection.remotePath)/\(BackupManifest.remoteFilename)"
let spManifest = signposter.beginInterval("ManifestLoad")
let manifestData = try? await transfer.downloadData(at: manifestPath)
let nasItems = (try? await transfer.listDirectory(at: connection.remotePath)) ?? []
// Decode + heal off main actor (CPU-bound work).
struct ManifestLoadResult { var index: ManifestIndex; var manifest: BackupManifest; var orphanCount: Int }
var loadResult: ManifestLoadResult = await Task.detached(priority: .userInitiated) {
var manifest: BackupManifest
if let data = manifestData,
let decoded = try? JSONDecoder.kisani.decode(BackupManifest.self, from: data) {
manifest = decoded
} else {
manifest = BackupManifest.buildFromNASListing(nasItems)
return ManifestLoadResult(index: ManifestIndex(manifest: manifest), manifest: manifest, orphanCount: 0)
}
// Heal: find NAS files not tracked in the manifest (orphans from
// backups that were interrupted before the manifest write step).
let knownFilenames = Set(manifest.entries.map { $0.filename.lowercased() })
let mediaExts: Set<String> = ["jpg","jpeg","png","heic","heif","gif","webp","bmp",
"tiff","tif","raw","dng","cr2","nef","arw",
"mp4","mov","m4v","avi","mkv"]
let orphans: [ManifestEntry] = nasItems.compactMap { item in
guard !item.isDirectory,
item.name != BackupManifest.remoteFilename,
!item.name.hasPrefix("."),
mediaExts.contains((item.name as NSString).pathExtension.lowercased()),
!knownFilenames.contains(item.name.lowercased()) else { return nil }
return ManifestEntry(
localIdentifier: "",
filename: item.name,
creationDate: item.modifiedDate,
fileSize: item.size,
remotePath: item.path,
uploadedAt: item.modifiedDate ?? Date()
)
}
if !orphans.isEmpty { manifest.merge(entries: orphans) }
return ManifestLoadResult(index: ManifestIndex(manifest: manifest), manifest: manifest, orphanCount: orphans.count)
}.value
signposter.endInterval("ManifestLoad", spManifest,
"\(loadResult.index.totalCount) entries, orphans=\(loadResult.orphanCount)")
logger.info("Manifest: \(loadResult.index.totalCount) entries (\(loadResult.orphanCount) orphans healed)")
// If the manifest gained entries from healing, write it back to NAS now
// so the gallery reflects actual NAS contents even before uploads start.
if loadResult.orphanCount > 0, let encoded = try? JSONEncoder.kisani.encode(loadResult.manifest) {
try? await transfer.writeData(encoded, to: manifestPath)
}
// Always seed local cache with the current (healed) state.
await NASManifestCache.shared.update(loadResult.manifest)
try Task.checkCancellation()
// 4. Build pending queue off the main actor
let spFilter = signposter.beginInterval("BuildPendingQueue")
let currentIndex = loadResult.index
let pendingAssets: [PhotoAsset] = await Task.detached(priority: .userInitiated) {
assets.filter { asset in
!currentIndex.matches(localIdentifier: asset.localIdentifier) &&
!(currentIndex.isFilenameOnly && currentIndex.matches(filename: asset.filename))
}
}.value
signposter.endInterval("BuildPendingQueue", spFilter, "\(pendingAssets.count) pending")
logger.info("Pending queue: \(pendingAssets.count) of \(assets.count) total")
guard !pendingAssets.isEmpty else {
job.allAlreadySafe()
BackupStatusService.shared.refresh(force: false)
return BackupResult.empty(date: startDate)
}
// 5. Upload loop
job.start(totalFiles: pendingAssets.count, totalBytes: 0)
NotificationService.notifyBackupStarted(count: pendingAssets.count)
// Build initial queue (all queued) for SyncView
queueItems = pendingAssets.map { asset in
BackupQueueItem(
id: asset.localIdentifier,
filename: asset.filename,
nasPath: "\(connection.remotePath)/\(asset.filename)",
status: .queued, bytesUploaded: 0, totalBytes: 0, speedBytesPerSec: 0,
error: nil, retryCount: 0, queuedAt: Date(), completedAt: nil
)
}
var uploaded = 0
var skipped = 0
var failed = 0
var totalBytes: Int64 = 0
var speedTracker = SpeedTracker()
var throttle = ProgressThrottle(hz: 8)
var manifestEntries: [ManifestEntry] = []
manifestEntries.reserveCapacity(pendingAssets.count)
var lastCheckpointAt = 0 // uploaded count at last manifest checkpoint write
let checkpointInterval = 25
let spUpload = signposter.beginInterval("UploadLoop", "\(pendingAssets.count) files")
let maxRetries = 3
for (assetIdx, asset) in pendingAssets.enumerated() {
if isCancelled { break }
while case .paused = job.status {
try await Task.sleep(nanoseconds: 500_000_000)
}
let remotePath = "\(connection.remotePath)/\(asset.filename)"
if assetIdx < queueItems.count {
queueItems[assetIdx].status = .uploading
}
if (try? await transfer.fileExists(at: remotePath)) == true {
job.fileCompleted(skipped: true)
skipped += 1
if assetIdx < queueItems.count {
queueItems[assetIdx].status = .skipped
queueItems[assetIdx].completedAt = Date()
}
logger.debug("Skipped (NAS exists): \(asset.filename, privacy: .public)")
continue
}
var uploadSucceeded = false
var lastUploadError: Error?
for attempt in 0..<maxRetries {
if isCancelled { break }
if attempt > 0 {
// Exponential backoff: 2s, 4s before retries 2 and 3
let backoffNs = UInt64(pow(2.0, Double(attempt)) * 1_000_000_000)
try? await Task.sleep(nanoseconds: backoffNs)
if assetIdx < queueItems.count {
queueItems[assetIdx].retryCount = attempt
queueItems[assetIdx].status = .uploading
}
logger.info("Retry \(attempt)/\(maxRetries - 1) for \(asset.filename, privacy: .public)")
}
do {
let spFile = signposter.beginInterval("UploadFile",
"\(asset.filename, privacy: .public)")
let localURL = try await photoService.exportAsset(asset)
defer { try? FileManager.default.removeItem(at: localURL) }
let fileSize = (try? localURL.resourceValues(forKeys: [.fileSizeKey]).fileSize)
.flatMap { Int64($0) } ?? 0
let bytesSoFar = totalBytes
let capturedIdx = assetIdx
if capturedIdx < queueItems.count {
queueItems[capturedIdx].totalBytes = fileSize
queueItems[capturedIdx].bytesUploaded = 0
}
try await transfer.upload(localURL: localURL, remotePath: remotePath) { sent, _ in
let speed = speedTracker.update(bytesSent: sent)
guard throttle.shouldUpdate() else { return }
Task { @MainActor [weak self] in
guard let self else { return }
self.job.updateProgress(
fileName: asset.filename,
fileSize: fileSize,
bytesTransferred: bytesSoFar + sent,
speed: speed
)
if capturedIdx < self.queueItems.count {
self.queueItems[capturedIdx].bytesUploaded = sent
self.queueItems[capturedIdx].speedBytesPerSec = speed
}
}
}
signposter.endInterval("UploadFile", spFile, "\(fileSize) bytes")
totalBytes += fileSize
job.fileCompleted(skipped: false)
uploaded += 1
uploadSucceeded = true
if assetIdx < queueItems.count {
queueItems[assetIdx].status = .uploaded
queueItems[assetIdx].bytesUploaded = fileSize
queueItems[assetIdx].completedAt = Date()
}
manifestEntries.append(ManifestEntry(
localIdentifier: asset.localIdentifier,
filename: asset.filename,
creationDate: asset.creationDate,
fileSize: fileSize,
remotePath: remotePath,
uploadedAt: Date()
))
// Checkpoint write every N uploads protects against interruption.
// Uses the same open connection so no extra connect/disconnect overhead.
if uploaded - lastCheckpointAt >= checkpointInterval {
lastCheckpointAt = uploaded
var checkpoint = loadResult.manifest
checkpoint.merge(entries: manifestEntries)
if let encoded = try? JSONEncoder.kisani.encode(checkpoint) {
try? await transfer.writeData(encoded, to: manifestPath)
await NASManifestCache.shared.update(checkpoint)
logger.info("Manifest checkpoint: \(checkpoint.entries.count) total entries")
}
}
} catch {
lastUploadError = error
logger.warning("Upload attempt \(attempt + 1)/\(maxRetries) failed for \(asset.filename, privacy: .public): \(error.localizedDescription, privacy: .public)")
}
if uploadSucceeded { break }
}
if !uploadSucceeded, let error = lastUploadError {
logger.error("Upload permanently failed for \(asset.filename, privacy: .public): \(error.localizedDescription, privacy: .public)")
job.fileFailed()
failed += 1
if assetIdx < queueItems.count {
queueItems[assetIdx].status = .failed
queueItems[assetIdx].completedAt = Date()
queueItems[assetIdx].error = BackupQueueItem.UploadError(
code: (error as NSError).domain,
message: error.localizedDescription,
timestamp: Date()
)
}
}
}
signposter.endInterval("UploadLoop", spUpload, "\(uploaded) uploaded, \(failed) failed")
let duration = Date().timeIntervalSince(startDate)
let result = BackupResult(
uploadedCount: uploaded,
skippedCount: skipped,
failedCount: failed,
duration: duration,
totalBytes: totalBytes,
date: startDate
)
job.finish(result: result)
BackupStatusService.shared.refreshAfterBackup(entries: manifestEntries, connection: connection)
let entry = BackupHistoryEntry(result: result, nasHost: host, triggeredByLAN: triggeredByLAN)
store.appendHistoryEntry(entry)
// Notifications
if result.failedCount > 0 {
NotificationService.notifyBackupFailed(count: result.failedCount)
} else {
NotificationService.notifyBackupCompleted(
uploaded: result.uploadedCount,
total: BackupStatusService.shared.snapshot.phoneTotal
)
}
// Persist queue so SyncView shows last session on next launch
let snapshot = queueItems
Task.detached(priority: .utility) { [snapshot, url = Self.queueCacheURL] in
guard let data = try? JSONEncoder.kisani.encode(snapshot) else { return }
try? data.write(to: url, options: .atomic)
}
return result
}
// Returns Tailscale host when off trusted LAN and tunnel is active.
private func resolveHost(connection: NASConnection, store: ConnectionStore, lan: LANMonitor) -> String {
let onTrustedLAN = store.trustedSSIDs.contains(lan.currentSSID ?? "")
guard !onTrustedLAN,
store.useTailscaleWhenRemote,
!store.tailscaleHost.isEmpty,
lan.isTailscaleActive else { return connection.host }
return store.tailscaleHost
}
func cancel() { isCancelled = true; job.cancel() }
func pause() { job.pause() }
func resume() { job.resume() }
func resolveSuccess() { job.resolveSuccess() }
}
// MARK: SpeedTracker
private struct SpeedTracker {
private var lastBytes: Int64 = 0
private var lastTime: Date = Date()
mutating func update(bytesSent: Int64) -> Double {
let now = Date()
let elapsed = now.timeIntervalSince(lastTime)
guard elapsed > 0.1 else { return 0 }
let speed = Double(bytesSent - lastBytes) / elapsed
lastBytes = bytesSent
lastTime = now
return max(0, speed)
}
}
// MARK: ProgressThrottle
/// Limits the rate at which progress callbacks trigger @MainActor UI updates.
/// Without this, SMB/SFTP stacks fire hundreds of callbacks per second, causing
/// SwiftUI to redraw the full view tree on every tick.
private struct ProgressThrottle {
private var lastUpdate: Date = .distantPast
private let minInterval: TimeInterval
init(hz: Double = 8) { minInterval = 1.0 / max(1, hz) }
mutating func shouldUpdate() -> Bool {
let now = Date()
guard now.timeIntervalSince(lastUpdate) >= minInterval else { return false }
lastUpdate = now
return true
}
}