rename() fails with EBUSY on Docker bind-mounted files. Add safeWriteFile()
helper that tries atomic rename first, falls back to direct write. Fixes
both init wizard and attach wizard fstab operations.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New Settings wizard to attach drives with existing filesystems without
formatting. Mounts partition at staging path, lets user browse and pick
a subfolder, then bind-mounts it at /mnt/<name> with fstab entries.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Bug 1 (HIGH): add --exclude _* to rsync --delete so _db/ and _config/
directories are never deleted between backup runs (crossdrive.go)
Bug 2 (MEDIUM): refactor RunDBDumps/RunBackup/RunFullBackup to use
acquireRunning/releaseRunning helpers; extract runDBDumpsInternal and
runBackupInternal so all three public entry points set m.running and
RunFullBackup no longer deadlocks calling the public methods (backup.go)
Bug 3 (MEDIUM): log [WARN] when GetDiskUsage returns nil in
ValidateDestination instead of silently skipping space checks (crossdrive.go)
Bug 4 (MEDIUM): add [WARN] on empty SystemDataPath in NewManager; add
[ERROR] in GetAppDrivePath; guard DumpStackDB against empty/relative paths
(backup.go)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Auto-enable daily rsync Tier 2 for apps without HDD mounts when ≥2
storage paths exist (AutoEnableSmallApps)
- Sync infrastructure config (stacks dir + controller.yaml) to all
secondary destinations via _infra/ directory (syncInfraConfig)
- Include infra paths in cross-drive restic snapshots
- Add SecondaryInfraPath() helper to paths.go
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Major refactor of backup and storage paths:
- Per-drive restic repos at <drive>/backups/primary/restic/
- Per-app DB dumps at <drive>/backups/primary/<app>/db-dumps/
- Remove global BackupDir, DBDumpDir, ResticRepo config fields
- Add SystemDataPath config (fallback for apps without HDD)
- New backup/paths.go with pure path computation helpers
- Add GetStackHDDPath to StackDataProvider interface
- Restic methods now accept repoPath as parameter
- Cross-drive backup uses new secondary path structure
- Rename storage/ to appdata/ in scripts and compose templates
- Update protected HDD paths (storage → appdata + backups)
- Simplify backup UI (remove global path displays)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Tier 2 cross-drive backup now configurable for all apps (not just HDD apps)
- Non-HDD apps (Mealie, Gokapi) can back up config + DB to secondary drive
- Status dot: removed "auto" gray — all apps start yellow, green = 2+ tiers OK
- Backup page: Tier 2 row always shown, Tier 3 placeholder added
- Deploy page: cross-drive config visible for all deployed apps
- Meta badges: non-HDD apps show "Konfig" or "Konfig + DB" instead of "Auto"
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Cross-drive now copies DB dumps (_db/) and config (_config/) alongside user data
- restic cross-drive includes config dir + full DB dump dir
- UI: per-tier rows (1. mentés / 2. mentés) instead of per-layer (DB/Konfig/Data)
- UI: BackupContents label shows what each tier protects (DB + Konfig + Adatok)
- UI: rsync backups show browsable indicator (📁)
- Cleanup: removed unused filterSnapshotsByPaths + pathCovers from router.go
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Fix 1: HDD data backup is now mandatory for all deployed apps.
resolveAppBackupPaths() iterates ListDeployedStacks() directly — no
longer reads GetAppBackupMap() or checks the Enabled flag. DiscoverAppData()
drops backupPrefs parameter; BackupEnabled is set from HasHDDData.
Five dead settings methods removed: IsAppBackupEnabled, SetAppBackup,
GetAppBackupMap, SetAppBackupBulk, GetAppBackupPrefs.
Fix 2: Cross-drive backup now triggers a fresh DB dump (DumpStackDB)
before running. DBDumper interface added to crossdrive.go; Manager
implements it; SetDBDumper wired in main.go. Non-fatal — proceeds with
user data backup even if DB dump fails.
Fix 3: Restore dropdown shows ALL deployed apps (not just HDD+enabled).
restore.go rewritten: always restores config+DB, adds user data if hasHDD.
UI shows restore type banner (full / config+DB / config only) with
color-coded styling. Snapshot API clarified for non-HDD apps.
Fix 4: "Docker kötetek" → "Konfiguráció" — named volumes are not in
the restic backup paths; compose files + app.yaml are what's backed up.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two rsync bugs found during Immich cross-drive backup testing:
Fix 3: Simplified destination path structure in runRsyncBackup.
Old SplitN logic kept "storage/immich" as a subpath, creating redundant
nesting: backups/rsync/immich/storage/immich/<data>. New logic: single
mount → rsync directly into the stack folder; multiple mounts → use
each mount's leaf dir name as subfolder (disambiguated by _N suffix).
Fix 4: Exclude app-internal DB dump files from rsync. Apps like Immich
store periodic pg_dumps in <data>/backups/*.sql.gz. The controller
already handles DB backups via pg_dump — copying them again wastes space.
Added --exclude backups/*.sql.gz/sql/dump to the rsync command.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Fix cross-drive backup failing for system-drive destinations. The
/mnt/hdd_placeholder folder is on the internal SSD, so IsMountPoint()
returned false and the old code hard-blocked it.
- ValidateDestination() (crossdrive.go): onSystemDrive flag replaces
boolean-only check; system drives require ≥10 GB free and <90% usage
to protect OS stability; external drives just need ≥100 MB free.
- CheckBackupDestination() (mounts_linux.go): Tier 4 now branches on
h.SystemDrive; system drive blocks at <10 GB free or ≥90% usage
(matches runner enforcement); external drive unchanged (warn 90%,
block 95%). Removes the && h.Severity == "ok" guard that was
preventing system-drive critical messages from overriding warnings.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- StackDataProvider interface extended with StopStack/StartStack
- backup.Manager.GetStackHDDMounts() delegates to stackProvider
- RestoreApp() auto-stops app before restic restore, restarts after (even on failure)
- stackAdapter in main.go wires StopStack/StartStack through to stacks.Manager
- GET /api/backup/snapshots?stack={name} filters snapshots by app HDD paths via filterSnapshotsByPaths()
- Restore section simplified: no path list, per-app filtered snapshots, human-friendly timestamp format, single calm warning, empty-result inline message
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Bug 1: Remove hardcoded 'v' prefix from templates (layout.html, settings.html); version tag already contains 'v'
- Bug 2: primaryHDDPath() and metrics collector now use GetDefaultStoragePath() instead of paths[0].Path so the real HDD is used, not the first (SSD) path
- Bug 3: Apps without HDD data show green/yellow based on volumeLastStatus instead of always gray
- Bug 5: Add default background rgba(255,255,255,0.1) to .btn to fix white-on-transparent readability
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Bug fixes:
- GetFullStatus() returns deep copy; CrossDriveSummary/UnconfiguredApps/CrossDriveWarnings
are always nil in the copy so the handler builds them fresh (fixes duplicate-apps bug)
- Replace binary IsMountPoint check with tiered CheckBackupDestination() — path-not-exist,
not-writable, system-drive (warning), disk >90-95% full; shown as warning vs critical
- Remove dead settingsAppBackupHandler / POST /settings/app-backup route (toggle wrote
to settings.json but nothing consumed the flag)
Architecture:
- Unified per-app backup rows: new AppBackupRow struct + buildAppBackupRows() replaces
the two old sections with expandable rows showing all 3 layers per app
- Sequential backup chaining: cross-drive runs immediately after restic (removed
independent cross-drive-daily/cross-drive-weekly scheduler jobs)
- Deploy page: remove "Csak kézi indítás" schedule option; add weekly consistency note
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Replace the complex "Alkalmazás adatok" form (checkboxes, paths, volumes,
save button) with a clean read-only status list. Each app shows its name
(linked to its deploy page) and a simple status: Aktív / Inaktív / N/A.
Also include ALL deployed stacks in the list (not just HDD-capable ones),
so apps without user data appear with N/A status.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Detect stale data on non-active storage paths after migration; show on
deploy/settings page with size info and two-step delete confirmation
- Add POST /api/storage/stale-cleanup handler with safety checks (active
path protection, registered-path validation, ProtectedHDDPaths guard)
- Export ProtectedHDDPaths() from stacks package for reuse in web handlers
- Sync FileBrowser mounts after successful app data migration
- Deploy page title/h2 now shows "Beállítások" for already-deployed apps
instead of always showing "Telepítés"
- Also add delete-old-data button on migration-done card in migrate.html
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Bug 1 (sfdisk): Add wipefs before sfdisk; change partition type from
,,,L (unsupported GPT shorthand) to ,, (default Linux GUID); add
--force --wipe always flags to handle existing partition tables.
Bug 2 (mount): Replace fstab-lookup mount with explicit device path:
mount -t ext4 -o defaults,noatime /host-dev/sdb1 /mnt/hdd_1
Container's /etc/fstab is Docker's auto-generated one, not the host's.
Bug 3 (mount propagation): Change /mnt volume to long-form bind with
propagation: rshared so mounts created inside container propagate to
the host. Requires mount --make-rshared /mnt on host before restart.
Safety: Use req.MountName (ASCII) for ext4 -L label (16-byte limit;
UTF-8 display label stays in settings.json). Add findmnt verification
after mount. Improve progress messages with command details.
Smart partition: In storageInitAPIHandler, if disk already has exactly
1 empty partition (no filesystem), skip wipefs+sfdisk and format the
existing partition directly.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
sfdisk is in the fdisk package on Debian bookworm, not util-linux.
util-linux provides lsblk/blkid/mount; fdisk provides sfdisk/fdisk/cfdisk.
Without this, FormatAndMount fails at the partitioning step.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Docker always creates a fresh tmpfs at /dev, silently dropping any
/dev:/dev bind mount. Block devices must be accessed via /host-dev
where the host /dev is actually mounted.
Changes:
- docker-compose.yml: /dev:/dev → /dev:/host-dev:rw
- safety.go: add HostDevPath constant + HostDevicePath() helper
- format_linux.go: all device ops (stat, sfdisk, partprobe, mkfs.ext4,
blkid) use HostDevicePath() to resolve /dev/sdb → /host-dev/sdb
- safety_linux.go: IsSystemDisk() stats device via /host-dev
- scan_linux.go: enrichWithBlkid() probes each partition individually
via /host-dev/sdXN instead of batch blkid -o export (which can't
find devices when /dev is Docker's minimal tmpfs)
Fixes "stat /dev/sdb: no such file or directory" in FormatAndMount.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Bug 1: Inside container, lsblk reports container mount points not host's.
Parse /host-fstab (or /etc/fstab fallback) to find system partitions
(/, /boot, /boot/efi, swap), resolve UUIDs via blkid, and mark parent
disks as system — replaces unreliable mount-point string matching.
Bug 2: lsblk returns null fstype in containers (no udev cache). Run
blkid -o export after lsblk scan and enrich partitions with FSType,
UUID, and Label from direct device probing.
Fixes sda appearing as available disk and "nincs fájlrendszer" showing
for all partitions.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Startup ping: fire heartbeat + health + hub report immediately on boot
(5s delay after scheduler start, instead of waiting 5-15 min for first tick)
- Storage init wizard: new internal/storage/ package with disk scanning
(lsblk -J), format+mount pipeline (sfdisk → mkfs.ext4 → blkid → fstab →
mount → chown), safety guards (system disk detection, confirmation "FORMÁZÁS"),
progress channel, auto-register in settings.json
- Data migration: MigrateAppData() with rsync --info=progress2 progress parsing,
stop/rsync/update-config/start flow, rollback on failure, old data preserved
- New pages: /settings/storage/init (wizard), /stacks/{name}/migrate (migration)
- New API routes: /api/storage/{scan,init,init/status,migrate,migrate/status}
- Deploy page: storage info section for deployed apps (path, size, free, migrate link)
- Settings page: "Mozgatás" button per app in storage path details
- Container: privileged: true, /dev:/dev, /etc/fstab:/host-fstab, /run/udev:/run/udev:ro
- Dockerfile: add util-linux, e2fsprogs, rsync, parted for disk ops
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Health severity fix: mount-point check downgraded from issue (FAIL) to warning (WARN)
- All storage health messages translated to Hungarian
- Success flash messages for all storage operations
- Edit storage path labels (inline edit UI + backend)
- App details per storage path on settings page (expandable list with names + sizes)
- Storage badge on stacks page showing which storage each app uses
- Deploy dropdown with free space display and low-space warning (<20%)
- Filesystem & disk info on settings page (ext4/btrfs, device, model via findmnt)
- Backup page storage context with per-app storage label badges
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix backup toggles not appearing (read each app's own HDD_PATH from app.yaml)
- Storage paths registry in settings.json with auto-discovery from deployed apps
- Settings page "Adattárolók" section with disk usage, add/remove/default/schedulable
- Deploy page path field as dropdown of registered storage paths
- Health check storage monitoring (mount point, disk usage alerts)
- Mount-point validation utilities (Linux syscall + cross-platform stubs)
- Controller docker-compose mount changed to /mnt:/mnt:rw for multi-storage
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- SyncPreferences() method on Notifier: POST to hub /api/v1/preferences
- IsEnabled() getter for hub connectivity check
- settingsNotificationsHandler: sync to hub after local save (3 flash message variants)
- Startup sync: non-blocking goroutine pushes prefs to hub on boot (DB rebuild recovery)
- Updated CONTEXT.md, README.md with v0.7.2 changes
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Monitoring page: "Távoli monitoring" section showing healthcheck ping UUID
configuration status (configured/not configured) for each of the 5 pings
- Alert manager: persistent dashboard banners on all pages generated from
health check results, missing pings, and backup status
- Notification system: controller-side notifier sends events to hub relay,
with cooldown tracking and event-type filtering
- Notification preferences UI: email, event checkboxes, cooldown settings
on the settings page with test email functionality
- Settings refactored: shared settingsData() helper, NotificationPrefs
struct with getter/setter and defaults
New files:
- controller/internal/web/alerts.go (AlertManager)
- controller/internal/notify/notifier.go (hub notification client)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- New settings.json persistence layer (internal/settings/settings.go)
- Atomic write (tmp + rename), thread-safe with sync.RWMutex
- Stores password hash overrides and DB validation cache
- Auto-creates on first save, graceful handling if missing
- Auth improvements
- Password resolution priority: settings.json > controller.yaml > none
- Session duration extended to 7 days (was 24h)
- ?next= redirect after session expiry (returns to original page)
- Flash messages on login page (used after password change)
- Conditional logout link (hidden when auth disabled)
- Session invalidation on password change
- New Settings page (/settings)
- Read-only system config display (customer, domain, git, backup, monitoring)
- Password change form with validation (min 8 chars, match check)
- Sidebar "Beállítások" item pinned to bottom above version
- DB validation persistence
- Validation results saved to settings.json after each dump
- Cached data survives container restarts
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add 4-branch template guard to handle zero-value DumpValidation
(shows "–" instead of misleading "Hiba" when validation never ran)
- Add debug logging to ValidateDump for all code paths
- Add cross-check re-validation in RefreshCache to heal stale
lastDBDump validation state from disk
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix tooltip "Invalid Date" by using raw.x with parsed.x fallback
- Fix chart x-axis bounds by including initial min/max in chartOpts
- Fix sysinfo alignment with fixed 2-column grid and consistent styling
- Fix chart overflow on mobile with min-width:0 and overflow:hidden
- Remove redundant <style> block from monitoring.html (moved to style.css)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Bug 1: Read hostname from /host/etc/hostname instead of os.Hostname()
which returns the container ID inside Docker. Added volume mount.
- Bug 2: Tooltip callback used parsed.x (category index) instead of
label (actual timestamp), showing 1970 dates.
- Bug 3+4: Default range changed from 24h to 1h so charts show data
immediately on new deployments with limited history.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
admin