The Always-On Tax: Why 24/7 Arrays Make No Sense for Media
A typical self-hosted media server is a paradox. You might maintain an 80TB array across eight enterprise drives, running continuous parity calculations in a 24/7 RAID-Z2 or unRAID pool. Yet, at any given moment, you are streaming at most one or two files.
The rest of those platters are spinning idle, consuming between 60W and 150W around the clock, emitting thermal exhaust, and steadily marching toward mechanical wear out. If your library runs on high-density Shingled Magnetic Recording (SMR) drives, continuous random reads and writes punish them even harder.
The 24/7 Always-On Pool
All hard drives spin constantly. Power consumption is static and high. Drives accumulate continuous running hours regardless of access. Every file is instantly readable, but at significant acoustic, thermal, and electrical cost.
On-Demand Tiering
Disks sleep spun down until requested. The filesystem stays 100% visible via stubs. When a user presses Play, only the target disk wakes, hydrates the stream to an SSD hot tier, and goes back to sleep.
With HuskHoard, your entire library is presented to the OS as a seamless directory structure. Movies and TV shows appear complete with their real names, sizes, and file attributes. Behind the scenes, the vast majority of the payload lives on cold drives that are completely spun down. When someone selects a film in Plex, HuskHoard catches the open() system call via Linux fanotify, powers up the drive, extracts the file to the SSD hot tier, and streams it. The physical drive can then immediately spin back down.
The Hydration Trap: Why Plex Scanners Kill Cold Tiers
This sounds like an ideal architecture—until you introduce Plex Media Server. By default, Plex does not behave like a normal human user. Plex is equipped with aggressive, automated background daemons designed to inspect your library.
Plex regularly deploys workers to inspect media containers:
Plex Media Scannerlooks for newly modified paths.ffprobeand deep media analysis check audio loudness, color bit depths, and stream markers.- Bif-generators scan every few frames of a movie to create scrubbing thumbnail images.
In a standard storage pool, this background activity is invisible. But in a tiered cold archive, it is lethal. If Plex detects new paths and attempts to open 400 movie files in the background, HuskHoard interprets every single file open as a playback request. It will wake your storage drives and attempt to hydrate hundreds of gigabytes—or tens of terabytes—into your local SSD hot tier all at once.
If your SSD fills up faster than the HuskHoard janitor can evacuate files, the hot tier saturates. The filesystem errors out, processes freeze waiting on unfulfilled fanotify permits, and your system encounters a catastrophic cache-thrashing loop.
To safely run Plex against an on-demand archive, you must construct two defensive perimeters: fanotify process exclusion and a sandboxed loopback hot tier.
Defense Layer 1: fanotify Process Filtering
Because HuskHoard uses the Linux fanotify API in permission mode, it can inspect not only what file is being accessed, but who is accessing it. We can instruct HuskHoard to categorically ignore calls made by Plex’s indexing and probing processes.
When Plex triggers an automatic scan, its worker processes inspect the stub. Rather than waking the sleeping disk array and recalling 60GB of video payload, HuskHoard checks the caller PID against its process whitelist and immediately returns permission without triggering a hydration event.
In your husk_config.toml, specify the exclusion list:
Defense Layer 2: Sandboxing with a Loopback Hot Tier
Configuration files prevent routine accidents, but safe system design requires a physical blast wall. If a user triggers an unplanned mass rescan or a new utility bypasses the process filter, you must ensure that your host SSD’s primary operating system drive cannot be overwhelmed.
The solution is to host your hot_tier inside a loopback filesystem. A loopback filesystem is a fixed-size file on your SSD that is formatted as an independent ext4 volume and mounted to a directory.
/) remains completely unaffected.Implementation Walkthrough
1. Neutralize Plex Server Settings
Open Plex Web, head to Settings > Server > Library, and configure the following:
- Uncheck: Scan my library automatically
- Uncheck: Run a partial scan when changes are detected
- Uncheck: Empty trash automatically after every scan (essential to prevent Plex from purging records if a disk takes a moment to spin up)
- Set Generate video preview thumbnails to Never
- Set Generate chapter thumbnails to Never
- Set Analyze audio tracks for loudness to Never
2. Build the Loopback Hot Tier
Execute the following steps on your host machine to create a dedicated, sandboxed hot_tier filesystem.
Step 1: Allocate space (Create the virtual disk image)
Use fallocate to create a fixed-size image file. For this example, we will allocate a 10 Gigabyte boundary (scale this based on your peak concurrent playback needs):
Step 2: Format the image file
Format the image with a native ext4 filesystem:
Step 3: Create the mount destination
Step 4: Mount the loopback image
Step 5: Fix ownership and permissions
Grant full write permissions to your server user account:
Step 6: Make the mount persistent across reboots
To ensure your mount automatically restores after a system restart, append it to your system’s filesystem table:
Verify that your loop mount is active and constrained to its allocated size:
3. Link the Hot Tier in HuskHoard
Update your husk_config.toml to point directly to the new loopback directory:
The Result
With this configuration active:
- Your disk drives sit completely idle, spun down into low-power states.
- Plex continues to display your full library, complete with original filenames and paths.
- When Plex executes a routine inventory or script cycle,
ignore_processesprevents spurious drive wakes. - When a user hits Play on a film, the calling binary is allowed through. The physical drive wakes, populates the loopback hot tier, and begins streaming in seconds.
- Even under abnormal conditions, your system partition cannot be corrupted or saturated by an unexpected hydration flood.