SysAdmin Weekly #29: Back From the Graveyard
The tools you wrote off deserve a revisit; sometimes the reason you quit them got fixed
TL;DR
Thunderbird 145 now speaks native Exchange (EWS with modern OAuth2), which means an open source mail client that finally works with Microsoft 365 without legacy auth or IMAP duct tape.
New episode 051: Andy and returning guest Clay Tamam tour what actually runs in their labs, hardware to hypervisors to the Graveyard of gear they powered off.
The Take: “end of life” is often a status you assigned, not a permanent fact; the tools you wrote off deserve a revisit cadence, because the reason you quit them sometimes gets fixed.
Community Signal: Brandon Lee on the popular homelab advice he no longer follows, which is the same right-size-the-lab discipline the episode is built on.
Tool of the Week is Jellyfin, the open source answer to the latest round of Plex price and paywall changes.
From the Console
I did not expect to feel anything about an email client in 2026. Then Thunderbird went and surprised me.
For years, Thunderbird was effectively dead to me for anything touching Microsoft 365. If you wanted it to talk to Exchange Online, you were bolting on a paid add-on or falling back to IMAP and legacy authentication, which is to say the exact protocols Microsoft has spent years trying to kill off, and for good security reasons. So on my Debian box, the machine I actually enjoy working on, I had no real open source mail client for my M365 mail. I lived in a browser tab and told myself that was fine.
Version 145 changed that. Thunderbird now speaks Exchange Web Services natively, uses the standard Microsoft OAuth2 sign-in, and detects the account settings for you. No add-on. No legacy auth. No pretending IMAP is a plan. I set it up, it just worked… mostly (after I accounted for bug 1847846 due to my large mailbox size). After that, I sat there mildly stunned that a piece of software I had mentally filed under “lost cause” had quietly clawed its way back into my daily driver rotation.
We wrote a whole issue about this once. Issue #22 was about open source as a lifeline: the exit door you keep unlocked for when a vendor changes the terms on you. This is the other half of that same coin. Sometimes the lifeline is not a new tool at all; it is one you already knew, left for dead, that got picked back up and reworked to fit the modern world. More on that below.
The latest on the SysAdmin Weekly Podcast
Episode: What I Actually Run in My Homelab (and Why)
Topic: Andy is joined by returning guest Clay Tamam, a working SysAdmin in the Netherlands, for an honest tour of what is really sitting in their labs: the hardware, the hypervisors, the services they use every day, and the reasoning behind each choice.
Why it’s worth your time:
It is the antidote to homelab-flex culture: two practitioners talking about what earns its place and what got powered off, not what looks impressive in a rack photo.
Andy walks through tearing a four-node Kubernetes cluster down to five VMs on a single Debian box, and Clay builds a rack from scratch on a real budget, so you get both ends of the spectrum with the tradeoffs spelled out.
It closes with the Graveyard: the gear and services that got shut off and why pruning is part of the discipline, not a failure of it.
Watch on YouTube
Listen on Spotify
The Take
“End of life” is a status you assign as often as one a vendor declares. And the two are not the same thing.
The podcast episode (above) closes on the Graveyard, the pile of gear and services you power off because they stopped earning their place. That discipline is correct. Most labs, and most production stacks, are carrying dead weight that nobody has had the nerve to switch off. Don’t get me wrong, removing technical debt is a good thing (in most cases) but the graveyard is not a one-way door. Sometimes a tool you buried was not actually dead. It was just dead to you, because the specific thing that made it unusable had not been fixed yet.
Thunderbird is my case in point this week. I wrote it off for M365 work years ago, for a completely valid reason at the time, and then I stopped tracking it. That is the trap. “I wrote it off” quietly turns into “I stopped paying attention,” and the two feel identical right up until the day the project ships the exact fix you gave up waiting for. Thunderbird 145 added native Exchange support over EWS with modern authentication, and the coverage was blunt about how long people had wanted it. The reason I quit it got fixed. I just was not looking.
Let’s be specific about the discipline, though, because this is not nostalgia. A tool earns its way back on current merits, not sentiment. Is it actively maintained? Does it meet your real requirements today? What is the catch? Thunderbird bet on EWS, a protocol Microsoft is slowly phasing out in favor of Graph, so the honest read is that email works now, calendar and contacts are still in progress, and the team is already building toward Graph for when EWS finally sunsets. That is a tool worth adopting with eyes open, not a fairy tale. The point is the cadence: put the tools you wrote off on a revisit schedule instead of treating a one-time verdict as a life sentence. The graveyard occasionally coughs one back up, and you want to be the admin who noticed.
Community Signal
Brandon Lee - “I No Longer Follow This Popular Home Lab Advice” - Brandon has decades in the field, and this is the same right-size-your-lab thesis Episode 051 of the podcast (with a mention in 052!) is built on. He makes the case for using Kubernetes only when orchestration actually solves a problem, applying high availability selectively instead of everywhere, and treating a couple of services (email and password managers among them) as reasonable exceptions where managed reliability beats self-hosting. It is a rare homelab piece that argues for running less, and it pairs directly with my Kubernetes-to-VMs teardown in episode 051.
Tool of the Week
Jellyfin - a fully open source, self-hosted media server for your movies, shows, and music, with no accounts to create, no cloud dependency, and no feature you rely on sitting one pricing update away from a paywall.
The reason it earns the slot this week: Plex just tripled the price of a new lifetime Plex Pass to $749.99, effective July 1, on top of already gating remote streaming behind a paid pass, which is exactly the vendor-changes-the-terms moment that sends people looking for the exit. Jellyfin is that exit, and it is genuinely free with no premium tier gating the core experience. Honest scope, though. This is not a turnkey Plex clone. Hardware transcoding takes real configuration to get right, the client apps across smart TVs and streaming boxes are less polished than Plex’s, and Jellyfin ships no built-in remote-access magic, so getting to your library from outside the house means you bring your own reverse proxy or a mesh VPN like Tailscale (which came up in the episode for exactly this reason). For a homelab that already runs a hypervisor and a bit of networking, that is a comfortable weekend project. For a non-technical household that just wants to press play, Plex is still the smoother ride, and it is fair to say so.
Quick Win of the Week
Pick one tool you wrote off more than two years ago, one you filed under “dead” and stopped tracking, and spend thirty minutes with its current changelog and release notes. Not to adopt it. Just to find out whether the specific thing that made you quit it has been fixed. Thunderbird and Microsoft 365 was mine this week; you almost certainly have your own candidate sitting in your mental graveyard. The goal is to make “I wrote it off” a decision you revisit on a schedule, not a verdict you handed down once and never appealed.
Fun Retro SysAdmin Fact
Thunderbird 1.0 shipped on December 7, 2004, which makes Mozilla’s mail client older than most of the “modern” SaaS platforms built to replace it; it was very nearly left to rot when Mozilla stepped back from active development in 2012, and here it is twenty-one years later adding native Exchange support and quietly having a renaissance.
From My Other Corner of the Internet
AndyOnTech
The reconstruct-project-history piece I had been sitting on is finally live: Can Claude Code Reconstruct a Project History You Never Documented?. It walks through pointing Claude Code at nothing but a repo’s git history and having it rebuild the changelog and decision trail that no human ever sat down to write. If you have a months-old project you no longer fully understand and you’re asking yourself the question “how did I get here?”, this is the practical, boring-value use case for the tooling, not the hype one.
Until Next Week
Prune what stopped earning its place, but keep a list of the tools you buried; a few of them are going to knock.
Stay Frosty,
Andy
SysAdmin Weekly