Skip to content
View in the app

A better way to browse. Learn more.

Unraid

A full-screen app on your home screen with push notifications, badges and more.

To install this app on iOS and iPadOS
  1. Tap the Share icon in Safari
  2. Scroll the menu and tap Add to Home Screen.
  3. Tap Add in the top-right corner.
To install this app on Android
  1. Tap the 3-dot menu (⋮) in the top-right corner of the browser.
  2. Tap Add to Home screen or Install app.
  3. Confirm by tapping Install.

kaedinger

Members
  1. I seem to not be able to post to the Plugin Support thread, so I'm posting it here. Please, Mods, move it to Pugin SUpport. Thanks! Unwake - Directory Caching for Unraid User SharesUnwake keeps every folder of your user shares in RAM, so browsing them doesn't wake sleeping disks. Not for a while after the last browse, not until a backup pushes the kernel's caches out - for good, kept current as files change. The main goal is fewer disk spin-ups; faster folder listings come for free. It replaces Dynamix Cache Directories, and it's the brother of Unspin, which has the same goal for file data: Unspin keeps reading files from waking your disks, Unwake keeps browsing from doing so. WhyTo list a folder of a user share, shfs asks every disk that may hold a part of it. If that disk's folder isn't in the kernel's caches any more, the disk spins up - for a folder listing, for a file name that doesn't exist (Windows asks for desktop.ini and Thumbs.db everywhere), or for the File Manager asking where a share lives. Cache Directories fights this by re-reading every folder over and over, so the kernel keeps them. A backup or a big copy still pushes them out, and the re-reading itself keeps disks busy. How It WorksUnwake runs unwaked, a FUSE daemon that sits on top of /mnt/user - shfs stays mounted underneath - and answers from its own tree in RAM: From RAM: folder listings (ls, SMB, NFS, File Manager), name lookups (including names that don't exist), file attributes, extended attributes (Samba's user.DOSATTRIB, system.LOCATION) From the disk, as before: reading and writing file data Through shfs, as before: creating, renaming, deleting Filling the tree. Unwake reads a disk's folders only while that disk spins anyway, at idle I/O priority, and never wakes a disk to do so. A disk that sleeps when Unwake starts is read the next time something else spins it up. Keeping it current. Unwake watches every disk and pool with fanotify. Every change - through Unwake, straight on /mnt/diskN, by the Mover, Unbalanced, a container or Unspin - updates the tree right away. Nothing is re-read on a timer. What stays out of RAM. Shares that live on an SSD pool only can't wake a disk, so they aren't kept in RAM; Unwake sends their requests straight to that pool, not through shfs, which would look every path up on every disk. If an array disk still holds folders of such a share, it's kept in RAM like the others. You can untick any share to pass it to shfs. Checked against the disks. A verifier compares the folders in RAM with the disks, slowly and only while a disk spins anyway and something else uses it, so it never keeps a disk awake. Whatever differs is repaired and written to /var/log/unwake-verify.log. VERIFY="no" in unwake.cfg turns it off. Safe by designshfs stays where it is. Stopping Unwake takes it off /mnt/user at once; whatever still uses it keeps working until it lets go. If unwaked dies, its supervisor takes the mount away immediately, so /mnt/user is plain shfs again, and restarts it. On array stop, Unwake gets out of the way within 10 seconds. Updates switch in place. The running version hands its FUSE connection, its tree and every open file and folder to the new one. Containers, VMs, SMB clients and shells carry on without a restart; they see a pause of about a second. If the new version can't take over, the old one simply keeps running. File data and kernel passthroughUnwake isn't in the way of file data: it hands every opened file to the kernel (FUSE passthrough, Linux 6.9+), which then reads and writes it directly - no copy through the daemon. A file that lives on exactly one known disk is opened on that disk, skipping shfs. One thing worth knowing: the kernel takes one backing file per file, shared by everything that has it open. So Unwake opens that backing file for reading and writing, even when the first program only reads - otherwise a program that wants to write a file while another one reads it (qBittorrent finishing a download that Plex is scanning, Word saving a document that Explorer's preview holds open) would be refused. That changes nothing on the disk, but: a program that tries to execute that same file via /mnt/diskN or /mnt/cache while it's open through /mnt/user gets "text file busy" tools that count file opens on the disks (Unspin) see one extra open when a file is first opened where a read-write open isn't possible (read-only file system, a file being executed), the backing is read-only, and a program opening that file for writing while it's open elsewhere gets "device or resource busy" FeaturesFolders in RAM, for good - nothing pushes them out of memory Never wakes a disk to read it - folders are read while a disk spins anyway Event-driven - no polling, no re-scans All Unraid file systems - XFS, btrfs, ZFS (datasets are followed as they come and go), on array disks and pools In-place updates - same mount, same open files; containers and VMs don't notice Shares in RAM - every share listed with a tick box; SSD-pool-only shares are detected and go straight to their pool Spin-up diagnostics - when a disk spins up, the log lists what Unwake passed to shfs that could have reached that disk Verifier - the folders in RAM are checked against the disks while they spin, differences repaired and logged Diagnostics in one file - rc.unwake diag, and rc.unwake users names whatever still uses an older instance Live log viewer - repeated entries folded Stable and beta channel - switch between them on the settings page ScreenshotsStatus and settings: Cache details and requests served from RAM: Shares in RAM: Activity log, an in-place update, stable and beta channel: InstallationVia Community Applications (recommended): search for Unwake and click Install. Manual: plugin install https://github.com/kaedinger/unwake/releases/latest/download/unwake.plg Then turn off Dynamix Cache Directories under Settings → Folder Caching, and enable Unwake under Settings → Utilities → Unwake. If Cache Directories left vfs_cache_pressure at a low value, the settings page offers to set it back to 100. Stable and beta channel. New features come out as betas first. The switch at the bottom right of the settings page shows the newest version on each channel and moves you between them; settings stay when you switch. Every stable release reaches the beta channel too. On the beta channel, CA doesn't list Unwake under Installed Apps and Fix Common Problems says it's not known to Community Applications - both only know the stable version; nothing is wrong. To install the beta directly: plugin install https://raw.githubusercontent.com/kaedinger/unwake/channel/unwake.plg RequirementsUnraid 7.2.0 or newer (tested with 7.3.2) RAM: about 270 bytes per file and folder name for the tree (1.4 million names: about 380 MB), more during and after a backup - see below Linux 6.9+ for kernel passthrough (Unraid 7.2 has 6.12); without it, file data goes through the daemon No external dependencies - single C++ binary RAM after a backup. Besides the tree, Unwake keeps an entry for every name the kernel currently holds an inode for, and a backup or media scan through /mnt/user touches nearly all of them. When the kernel lets go of them, Return Freed RAM (on by default) hands the freed memory back to the system a few minutes later. With it off, Unwake keeps that memory for the next backup - a few milliseconds saved a night, but the night's peak held all day; the settings page then shows how much is kept. On my server (1.4 million names): 376 MB before the nightly backup, 864 MB the next morning with it off, 479 MB seconds after switching it on. LinksGitHub: https://github.com/kaedinger/unwake Issues / feature requests: https://github.com/kaedinger/unwake/issues License: GPLv2 AI disclaimer / HistoryUnwake came out of Unspin's disk wake-up tracking - once it showed who woke the disks, most of the wake-ups turned out to be folder listings and name lookups that Cache Directories couldn't keep in memory. Unwake was designed and written with Claude Code, and tested on my own server.
  2. Script: move finished torrents into dated folders so same-named downloads don't overwrite each other Hi all, If you use qBittorrent's "Set location" / move-on-complete with a shared final folder, you may have been bitten by this: a newer torrent with the same file or folder name as an older one silently overwrites it. I wrote a small bash script that runs inside binhex-qbittorrentvpn and avoids that. Each finished torrent whose save path is exactly one of your configured final folders gets moved through the WebUI API into a subfolder named after its completion date: /data/a/linux.iso -> /data/a/20261015/linux.isoIf something with the same name is already in that folder (on disk, or claimed by another torrent in qB), it uses 20261015_1, 20261015_2 and so on instead. Dated folders that end up empty (e.g. after you delete torrents with their files) are removed at the end of each run. Setup (binhex-qbittorrentvpn): 1. Put qb-dated-move.sh into /mnt/user/appdata/binhex-qbittorrentvpn/scripts/ (that's /config/scripts/ inside the container). 2. Copy qb-dated-move.conf.example to qb-dated-move.conf in the same folder and fill in your WebUI user/password and final folders. Use container paths, the way qB shows them. Then: chown nobody:users qb-dated-move.conf chmod 600 qb-dated-move.conf3. In qB: Tools → Options → Downloads → Run external program → Run on torrent finished: /bin/bash /config/scripts/qb-dated-move.shcurl, jq and flock are already in the binhex image, so nothing else needs to be installed. Try it first without moving anything: docker exec -it binhex-qbittorrentvpn bashto enter qB's container (assumes the container is still called binhex-qbittorrentvpn), or use qB's Console. Then: DRY_RUN=1 /config/scripts/qb-dated-move.shThe log is at /config/scripts/qb-dated-move.log. To sort an existing backlog, you can limit a manual run to older torrents with MIN_AGE_MONTHS=12. Tested with qBittorrent 5.2.4. Torrents with Automatic Torrent Management enabled are untested. Before the first big run, best back up /config/qBittorrent/BT_backup. Source, docs and license (GPL-3.0): https://github.com/kaedinger/qb-dated-move Feedback and bug reports are welcome, here or as GitHub issues.
  3. Unspin is back on CA! Thanks Unraid team!
  4. Thanks, done!
  5. Unspin is temporarily missing from Community Applications Hi all, If you've searched for Unspin in Community Applications (CA) in the last few days and couldn't find it: it has been removed from the CA listing for now. That was my mistake. What happened On 2026-09-18 I changed where the plugin checks for updates, from a file on GitHub's main branch to the GitHub release itself. I forgot to update the separate template that tells CA where Unspin lives, so the two URLs no longer matched. CA treats that as a fatal error and removed the plugin. Both sides were fixed on 2026-09-20 and match again, and CA no longer reports an error. It hasn't put Unspin back on its own, though, so I've asked the CA moderators to re-list it. If you already have Unspin installed Nothing changes for you. Your install, settings and update checks keep working, and the latest version, 2026.09.20.50926, is available through the normal plugin update check. If you want to install it now Until it's back in CA, you can install it by hand: go to Plugins → Install Plugin, paste this URL and click Install: https://github.com/kaedinger/unspin/releases/latest/download/unspin.plg You'll get the same plugin CA installs, and it keeps updating normally. I'll post here when Unspin is back in CA. Sorry for the trouble!
  6. Unspin plugin removed from CA after PluginURL mismatch, fixed but not re-listed Hi all, My plugin Unspin (Support page: https://forums.unraid.net/topic/198091-plugin-unspin-rudimentary-hot-file-tiering-for-unraid/, template repo: https://github.com/kaedinger/unraid-ca-xml-templates) was removed from CA on about 2026-09-18 with: That was my mistake: I moved the .plg's pluginURL to the GitHub release asset without updating the CA template. It was fixed on 2026-09-20 and both now match: Template PluginURL: https://github.com/kaedinger/unspin/releases/latest/download/unspin.plg The .plg served at that URL (release 2026.09.20.50926) declares the same pluginURL The error has gone from the template errors. But Unspin is still missing from the app feed after several days of regular rebuilds, so it looks like it has to be re-listed by hand. Could a moderator re-enable it, or tell me if anything else needs fixing? Thanks!
  7. Hi @mmm77, excellent diagnosis, your Gemini nailed it. Since I am not myself a ZFS user, I didn't immediately recognize the problem. I filed it as GH #8 with your comment as the source, credited. Fixed in 2026.09.14.59405 (make sure you're on 2026.09.18.38737 or later - there was an unrelated installer hiccup inbetween). The fix does basically what you suggested: at startup, unspind walks the immediate children of each disk mount and marks any that are themselves separate mounts with fanotify - that covers ZFS datasets, since Unraid only ever datasets the top-level shares, never nested subfolders. (It also rescans every 60s so a share you use for the first time (which is when Unraid actually creates its dataset) gets picked up without a daemon restart.) Would appreciate it if you could confirm on your box once you've updated - promotions on your ZFS shares should now behave the same as on XFS/Btrfs. Thanks again for the thorough writeup, that made this a quite simple fix instead of a guessing game. 😉 Best, kaedinger
  8. Hi Mike, sorry for the late answer. I work on Unspin in my free time. Thanks for the detailed report (no worries on the AI write-up, the log you attached is what I needed). Good news: Unspin should work on ZFS as well. fanotify (the mechanism Unspin uses to watch file access) works at a level below the filesystem, so it doesn't actually care whether a disk is XFS, Btrfs, or ZFS, encrypted or not. Your log confirms this is working correctly: all four of your ZFS disks got watched successfully (Watching mount at: /mnt/disk1... etc.), and events are being picked up fine. So this isn't a ZFS compatibility issue. (Caveat: I don't have a ZFS array to test against myself, so this is reasoned from the fanotify docs and your log rather than something I've verified hands-on) So I think it looks like this: every one of the events in your log is for a file sitting directly at the top level of the disk (e.g. /mnt/disk1/757176000.tmp...) rather than inside one of your shares (e.g. /mnt/disk1/Movies/...). Unspin only promotes files that belong to a real user share, so anything living outside a share folder gets skipped - which is exactly what the [skip share] lines mean. That's working as intended, just maybe not obvious from the log wording. The interesting question is what's actually creating those .tmp files directly at the disk root, at the same second, across four disks at once. The numeric format (757176000.tmp) indicates they are auto-generated atomic write fallbacks or temporary file tests. So maybe: Unraid's management engine itself or an active Docker container trying to test disk capabilities or sync states. A backup, sync, or maintenance tool that's mapped directly to /mnt/diskX instead of a share path A Docker container or VM with a bind-mount pointed straight at the disk root Some other periodic Unraid/ZFS housekeeping process Could you check what's writing those files? If you can tell me what created them, I can figure out whether Unspin should be handling that case differently, or whether it's something worth pointing you to reconfigure on your end. Thanks again for reporting this with logs, that's exactly what makes it possible to track down. Oh, and regarding share monitoring: a user share in Unraid is a virtual, cross-disk pooling of folders that exist independently on each physical disk - there's no single mount point for "a share" the way there is for a disk. fanotify marks a mount, so watching at the share level isn't a smaller/simpler version of what Unspin does now, it's structurally backwards - you'd need to watch every disk anyway and then filter, which is exactly the current design. Best, kaedinger
  9. Assuming I did not mess up terribly, it should not be possible for the plugin to keep the disks awake, as it's only accessing the disks directly after they actually have been accessed from somewhere else. As it does no live-switching (which is a thing best implemented by the Unraid team itself), it's suited best for frequently accessed files, like samples in an audio library, or pdfs holding data that you open every day, or the infamous desktop.ini files that windows creates, or drives holding original videos while editing, etc. If you want movies cached while they are playing, this is NOT what Unspin does. To me it's most helpful in audio and video editing projects: projects that I currently work on have their files moved to cache, despite having set the shares to "Cache -> Array". Kind of like the old "Prefer Cache" setting, but only for data that is actually being used. I think this plugin works best in conjunction with two other plugins: CA Mover Tuning and Dynamix Cache Directories. Since the goal is not spinning up disks unnecessarily, caching the directory content is the first thing you should do anyway. And Mover Tuning is used to move old/cold files back to the array where they belong - Unspin only moves to Cache. I have mover tuning set so that it only reacts to ATIME and moves files based on age, smart caching enabled (Auto setting in file age), fill up level 95%, free down to 50%, running every 4 hours. Disable the stock mover. This works perfect for me. Enable debug logging in Unspin to exactly see which files are being watched, moved, and why. Play around with the rules in Unspin to fine tune the behavior according to your needs. Use the File Activity plugin's filters to monitor and control Unspin if you are unsure. And pause Unspin automatically during backups, mover actions, and similar procedures - this is important, otherwise you have a lot of files moved back and forth unnecessarily. I hope this helps a bit. If you feel there's something wrong with the functionality, please open a Github issue. Thanks.
  10. Unspin - Rudimentary Hot-File Tiering for UnraidUnspin automatically promotes your most-accessed files to faster storage (cache/NVMe) using a lightweight C++ daemon that watches file access in real time. The main goal is to have fewer disk spin-ups (hence the name), but we'll take the faster access, too ;-) It also tells you which program, container, VM or SMB client woke each of your sleeping disks, so the eternal "what spun up my disk?" question finally gets an answer. There's no explicit cold demotion - Unraid's mover (or Mover Tuning, see below) takes care of that. Make sure this does what you want! There's a dry-run mode; use it! How It WorksUnspin runs unspind, a daemon that uses Linux's fanotify interface to receive an event every time a file is opened or read on your array disks. It applies three configurable promotion rules: Small file rule - files below a size threshold get promoted after N total reads (e.g. promote a config file on first access) Large file - short window - N reads within M minutes triggers promotion (e.g. 100 reads in 5 min = streaming video) Large file - long window - N opens within H hours triggers promotion (e.g. 3 opens in 24 h = frequently consulted PDF) Why these rules? Under ideal circumstances we would cache everything - but we don't have the space. We can't have a huge Linux ISO using up precious cache space if it's just held open for a few bytes of torrent upload. But of course that changes if it's being constantly read, or accessed on a regular basis. You'll have to play around with the rule settings to best match your own use cases. Safe by designPromotion waits until the file is closed - everywhere, not just by the program Unspin saw. A movie you're watching keeps playing from the array and is promoted after you stop; the next open picks up the cache copy. (No live switching during playback.) A file is never lost, doubled, left half-copied or missing from the share, not even for a moment. The copy stays invisible until it's complete and is only swapped in if nobody moved, changed, deleted or opened the file in the meantime. Whatever you do with the file during a promotion wins, and Unspin simply tries again later. Even a crash mid-promotion is cleaned up on the next start. Unspin never wakes a sleeping disk. A file can be read from RAM while its disk sleeps; it isn't promoted then (removing the array copy would wake the disk and parity), but on a later access with the disk awake. FeaturesEvent-driven - reacts to each file read in real time via fanotify, no polling or cron Three-rule promotion - separate thresholds for small files and large files (streaming burst / periodic access), each rule can be switched on or off Per-share cache pool routing - destinations follow each share's own cache pool setting in Unraid, so different shares can promote to different pools. Shares with "Use Cache: No" or "Only" are skipped automatically Detected Shares - the settings page lists every share with its routing and a Watch toggle to opt individual shares out Per-pool fill guard - each pool gets its own fill % limit; promotion to that pool stops once it's exceeded (ZFS pools with child datasets included) Disk Wake-ups - which program, container, VM or SMB client (user and IP) woke which disk, with counts per day and the latest wake-ups. Access through user shares is traced back to the program that asked for the file. Directory listings, name lookups and metadata writes are named as such Promotion statistics - files and size moved to each pool today, per day on average and in total Mover / rsync aware - pauses while Unraid's mover (including Mover Tuning) or, optionally, any rsync runs Pause locks for backups - stackable rc.unspin pause NAME, optionally with a time to live so a crashed backup job can't leave Unspin paused forever Pool Access Times - turns on access times for the pools Unspin promotes into, so Mover Tuning's "Use ATIME" filter can demote files that haven't been read for a while ZFS array disks - dataset-per-share mounts are detected and watched automatically Waits for slow mounts at boot, with an Unraid notification if it has to Dry-run mode - logs every decision without moving anything Exclude patterns - skip any path substring Live log viewer and, on debug logging, the recently accessed files per disk Using Unspin with Mover TuningUnspin promotes files because they're read, so the natural way to demote them is Mover Tuning's age filter with Use ATIME: files go back to the array once they haven't been accessed for a while. Unraid mounts every pool with noatime, so access times never change on a read - switch on Pool Access Times in Unspin's settings to fix that for the pools it promotes into. Promoted files get the promotion time as their access time, so they aren't demoted on Mover Tuning's next run. Screenshots(new screenshots coming) InstallationVia Community Applications (recommended): search for Unspin and click Install. Manual: plugin install https://github.com/kaedinger/unspin/releases/latest/download/unspin.plg Pause for BackupsWrap your backup jobs with pause/unpause over SSH so Unspin doesn't count (or promote) files while your backup tool reads them: ssh root@nas /etc/rc.d/rc.unspin pause borgbackup borg create ssh://root@nas/mnt/disk1/backups::daily /data ssh root@nas /etc/rc.d/rc.unspin unpause borgbackup Each lock has a name, and locks stack - multiple backup tools can pause independently, and Unspin only resumes once all of them are released. rc.unspin status lists the locks currently held. Locks that expireA normal lock stays until it's released, so if the job that set it dies (killed, crashed, its VM stopped mid-run), Unspin stays paused until you unpause it or reboot. To avoid that, give the lock a time to live and refresh it while the job runs: ssh root@nas /etc/rc.d/rc.unspin pause mybackup --ttl 15m # ... during the job, every 5 minutes: ssh root@nas /etc/rc.d/rc.unspin pause mybackup --ttl 15m # ... at the end: ssh root@nas /etc/rc.d/rc.unspin unpause mybackup Calling pause again with the same name refreshes the lock; pause mybackup without --ttl makes it permanent again. Times are a number with s, m, h or d (e.g. 90s, 15m, 2h, 1d). A lock that isn't refreshed in time is removed and logged as Pause lock expired - mybackup. rc.unspin status shows how long each of these locks has left. RequirementsUnraid 7.0+ (probably even earlier, but tested with 7.2.3 onwards, currently 7.2.4) No external dependencies - single C++ binary LinksGitHub: https://github.com/kaedinger/unspin Issues / feature requests: https://github.com/kaedinger/unspin/issues License: GPLv2 AI disclaimer / HistoryClaude has been mainly used to simplify refactoring, restructuring, commenting & documenting, cleaning up etc. It also helped immensely setting up the general structure of an Unraid plugin and with getting the settings page right. BUT: Every functional code line has been human written at some point by meticulously reordering bits on an old hard drive using a VERY small magnet. Jokes aside, this one started as a batch script a few years ago on my old Ubuntu machine Gregory (RIP), then turned into a C++ daemon for a proprietary project I did for a German car company (where it was called Hotfile and already using fanotify). As I started using Unraid this year and wanted to give something back to its fantastic community, I turned it into this plugin.
  11. ...but I can't because I am not allowed to start a topic in "Plugins". I wanted to publish this plugin: https://github.com/kaedinger/unspin I am relatively new to Unraid (so no other posts in the forum except one reply) but have been developing software for 30 years, and I want to give back something useful. If I understand correctly, someone has to just approve this one, and then I can post a new plugin? Keep up the good work, Florian
  12. kaedinger changed their profile photo
  13. Just a heads up in 2026: The current Schneider Electric Firmware Update Wizard (v4.3.8v9a) did recognize my trusty old APC Smart-UPS 750I as upgradeable but did NOT contain a firmware file for it... In addition my UPS did not show any Firmware ID as described by Schneider, just Firmware Versions 1-4. I managed to track down a v4.2.1 which could update the UPS to Firmware ID 18. After that I could enable Modbus on the UPS, and - using nut-2.8.4 stable - I had to switch the UPS Driver to apc_modbus and voila:

Account

Navigation

Search

Search

Configure browser push notifications

Chrome (Android)
  1. Tap the lock icon next to the address bar.
  2. Tap Permissions → Notifications.
  3. Adjust your preference.
Chrome (Desktop)
  1. Click the padlock icon in the address bar.
  2. Select Site settings.
  3. Find Notifications and adjust your preference.