Files
Kisani/Services/BackupEngine.swift
Robin Kutesa b18a2aad14 fix: remove NAS directory healing (inflated archive count) and record skipped files
The directory healing step added every media file in the NAS folder as a manifest
orphan — including files from other devices — inflating NAS Archive to 7421 and
polluting the manifest with unrelated entries. Removed entirely.

Root cause of the loop: when BackupEngine calls fileExists and skips a file already
on NAS, it never added that file to manifestEntries. The file existed on NAS but
not in the manifest, so every reconcile still showed it as pending, triggering a
new backup on every app open.

Fix: skipped files are now added to manifestEntries with their localIdentifier so
they are written to the manifest at the end of the run (and at checkpoints). Next
reconcile sees them as safe via localIdentifier match and needBackup decrements.

Co-Authored-By: Kutesir <kutesir@provoc.ug>
Co-Authored-By: Sentry <sentry@provoc.ug>
2026-05-18 00:25:45 +03:00

395 lines
17 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
let manifestPath = "\(connection.remotePath)/\(BackupManifest.remoteFilename)"
let spManifest = signposter.beginInterval("ManifestLoad")
let manifestData = try? await transfer.downloadData(at: manifestPath)
let (baseIndex, baseManifest): (ManifestIndex, BackupManifest) = await Task.detached(priority: .userInitiated) {
guard let data = manifestData,
let manifest = try? JSONDecoder.kisani.decode(BackupManifest.self, from: data)
else { return (ManifestIndex(nasListing: []), BackupManifest()) }
return (ManifestIndex(manifest: manifest), manifest)
}.value
signposter.endInterval("ManifestLoad", spManifest, "\(baseIndex.totalCount) entries")
logger.info("Manifest: \(baseIndex.totalCount) entries")
// Seed local cache so BackupStatusService fast-path sees current NAS state.
await NASManifestCache.shared.update(baseManifest)
try Task.checkCancellation()
// 4. Build pending queue off the main actor
let spFilter = signposter.beginInterval("BuildPendingQueue")
let pendingAssets: [PhotoAsset] = await Task.detached(priority: .userInitiated) {
assets.filter { asset in
!baseIndex.matches(localIdentifier: asset.localIdentifier) &&
!(baseIndex.isFilenameOnly && baseIndex.matches(filename: asset.filename)) &&
!baseIndex.matchesUnclaimed(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()
}
// Record in manifest even though we didn't upload the file is on NAS.
// Without this, skipped files stay "pending" forever and trigger backup loops.
manifestEntries.append(ManifestEntry(
localIdentifier: asset.localIdentifier,
filename: asset.filename,
creationDate: asset.creationDate,
fileSize: 0,
remotePath: remotePath,
uploadedAt: 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 = baseManifest
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
}
}