Notes, Memory, and Keeping Things Running

Homelab Highlights for 2026-08-14: self-hosting and homelab notes worth knowing.

[Audio embed placeholder]

News Roundup

August maintenance updates: Nextcloud Hub 25 Autumn, 26 Winter & 26 Spring

Nextcloud has released maintenance updates for three Hub versions: 32.0.14 for Hub 25 Autumn, 33.0.8 for Hub 26 Winter, and 34.0.3 for Hub 26 Spring. The updates include important bug fixes and security upgrades.

Why it matters: Keeping self-hosted Nextcloud instances updated is a standard chore for security and stability, and these consolidated releases simplify the process for users on these popular versions.

Source: Nextcloud community

RustDesk now supports true unattended remote access on Wayland

RustDesk, the open-source remote desktop software, has officially added support for true unattended remote access on the Wayland display server protocol.

Why it matters: This bridges a significant gap for remotely managing Linux desktops, making RustDesk a more viable option for homelab administrators whose systems use the modern Wayland display server.

Source: RustDesk Blog

PBS station fears losing 50TB of data after being ghosted by cloud storage provider

A PBS affiliate is suing to recover 50TB of archival data after its cloud storage provider, Open Source Storage, went defunct. The data is physically held in an Iron Mountain data center, but legal and contractual complexities are blocking access, requiring court intervention and a former employee’s help for retrieval.

Why it matters: This case is a stark, real-world reminder of the risks in third-party data custody chains, highlighting the importance of understanding physical media access, encryption keys, and exit strategies for critical backups.

Source: Ars Technica

Dell Poweredge T110 II ECC reporting and injection

A community toolkit has been created to enable proper ECC error reporting and testing for the older Dell PowerEdge T110 II server. It configures the rasdaemon stack to report hardware-detected memory errors and includes scripts for ECC injection testing.

Why it matters: For homelabs still relying on these capable, older servers, this project provides crucial monitoring support that extends their useful life and helps prevent silent data corruption from memory errors.

Source: github.com

Plex for Mobile (Android & iOS) BETA v2026.16.0 Beta 2 Released

Plex has released Beta 2 for its 2026.16.0 mobile apps on Android and iOS, focusing on bug fixes and performance improvements.

Why it matters: For users who test beta versions of their media server clients, this update represents incremental stability improvements for the mobile experience.

Source: Plex Forums

Tool & Software Highlight: NoteDiscovery 0.31.1, self-hosted notes app gets more community involvement

NoteDiscovery, a self-hosted, MIT-licensed notes app that reads and writes plain markdown files, has released version 0.31.1. The update introduces a new plugin system with a proper contract for community contributions, an opt-in ‘contrib’ folder for safety, and its first community plugin for open task search. It also includes fixes for search, wikilinks, statistics, and CORS, plus quality-of-life improvements like lazy-loading for Mermaid diagrams and immutable asset caching.

Why it matters: For homelab users seeking an Obsidian-like experience under their own control, this update marks a shift towards a stable core with extensible community features, all while keeping notes as simple, portable markdown files.

Source: github.com

One Thing to Try

Set up a basic cron job or script using docker inspect to check for containers that have restarted, helping you catch silent failures before they become a problem.

Sources

Transcript

Host A: Welcome to Homelab Highlights, the show that surfaces practical homelab wins and useful self-hosted tools.

Host A: Let’s start with today’s updates.

Host A: Nextcloud’s out with maintenance updates for three Hub versions. If you’re on Hub 25 Autumn, 26 Winter, or 26 Spring, you’ll want versions 32.0.14, 33.0.8, and 34.0.3. They’re calling these important bug fixes and security upgrades. Host B: [conversational] Good to see those moving. It’s one of those chores you sometimes put off, but having a consolidated set makes it easier. Standard advice applies—test first if it’s critical, but these maintenance releases are generally solid.

Host A: RustDesk now supports true unattended remote access on Wayland. Their blog says it’s officially available. Host B: [with a small lift] That’s welcome progress for managing Linux desktops remotely. The Wayland landscape’s been patchy for remote tools, so RustDesk bridging that gap is a solid step. It works with their self-hosted server option too.

Host A: And a cautionary tale from Ars Technica: a PBS affiliate is fighting to recover 50 terabytes of archival data after their cloud storage provider, Open Source Storage, went defunct and ghosted them. The data’s in an Iron Mountain data center, but OSS owns the hardware. Host B: [skeptical] Fifty terabytes of local history, just stuck. A judge ruled Iron Mountain has to hand over the physical devices, but the station needs a former OSS employee to help retrieve it within 30 days. Stark reminder about custody chains when outsourcing storage. Host A: The timeline’s rough. The station paid OSS for years, then the company stopped responding. They discovered OSS had dissolved. The data includes decades of local footage—irreplaceable stuff. Iron Mountain’s position is contractual; they’re holding drives as collateral for unpaid bills. Host B: [thoughtful] It’s a nightmare that makes you double-check your own backup and exit strategies. Do you have a direct line to the physical media? Know the encryption keys? It’s the extreme end of “cloud is just someone else’s computer,” playing out in real time.

Host B: Let’s talk about a tool update. NoteDiscovery just hit version 0.31.1. Host A: Right, the self-hosted notes app that works on a folder of plain markdown files. MIT licensed, single Docker container for the base. Host B: [curious] This release leans into community. Big change is a new plugins system with a proper contract—hooks declare what they can replace, plugins get context and can serve their own endpoints. Host A: They set up a plugins/contrib folder, so community submissions live there but won’t load until you explicitly copy one in. Nice opt-in safety net. Host B: First community plugin is already in, from lubeda. It’s an open task search that lists every note with an unchecked checkbox—handy for a personal knowledge base. Taps the existing search index, so it’s fast. Host A: Beyond new features, solid fixes: search was broken for partial words, wikilinks had a 404 bug, note stats inconsistency, and a CORS issue with wildcard origins. Host B: [with emphasis] Quality-of-life tweaks: search needs at least two characters, Mermaid diagrams load only on first view, static assets cached immutably per version. The Mermaid change lazy-loads that hefty library only if you use diagrams. Host A: Project’s hitting that phase where the core is stable and community can build on top. For a homelab, having notes as files you own, with a clean interface that plugs into dashboards like Homepage, it’s compelling. Host B: If you want an Obsidian-like experience but in a Docker container you control, this is one to watch. File-based means you can use any other markdown editor on the same folder.

Host A: Let’s talk about a community project from a specific pain point. Host B: Someone encountered a failed DIMM on their Dell PowerEdge T110 II server running Proxmox. Host A: The T110 II—an old workhorse many have in a closet. When a DIMM fails, the server might log ECC errors, but getting those reported through rasdaemon wasn’t straightforward. Host B: [quick aside] So they put together a toolkit on GitHub called ‘Dell_Poweredge_T110ii-edac’. It sets up kernel modules and config so rasdaemon can see and report hardware-detected ECC errors. Host A: It goes further—enables ECC injection for testing. That’s deliberately introducing memory errors to verify monitoring and alerting. The README walks through loading the specific i5400_edac kernel module for the T110 II’s Intel 5400 chipset. Host B: It’s the kind of deep, system-level fix you get when someone hits the problem and documents the solution. If you still run one of these boxes, this is the difference between a cryptic light and a proper log entry. Host A: Homelab ethos: you run into a problem with older gear, figure out the fix, and share it. The repo includes a script for an ECC injection test—flip a single bit in memory, see a corrected error pop up in logs. Great way to validate your monitoring pipeline. Host B: [lighter] These older servers are still capable for homelab use—low power draw, ECC memory, built like tanks. The main drawback is often lack of modern driver and monitoring support. Projects like this bridge that gap, extending useful life.

Host B: Quick tip: a simple way to get notified when Docker containers silently restart. Host A: Good one. Easy to miss when a container crashes and the orchestrator just brings it back up. Host B: You can set up a script checking Docker events. Command docker events --filter 'event=die' --format '{{.Time}} {{.Actor.Attributes.name}} exited with code {{.Actor.Attributes.exitCode}}' streams exits. For a periodic check, a cron job using docker ps -a looks for containers with a recent restart. Host A: More direct: use docker inspect output. Something like docker inspect --format='{{.Name}} restarted {{.RestartCount}} times' $(docker ps -aq) | grep -v ' 0 times' lists only containers that restarted at least once. Pair with a date check on start time for a basic health monitor. Host B: If you use Prometheus and cAdvisor, this is built in. But for a lighter setup, a five-line bash script in a daily cron can catch silent failures before they become a problem. Key is looking at restart count and running status—if something’s stuck in a loop, you want to know.

Host A: That’s Homelab Highlights for Friday. Until next time, happy hosting!