What Self-Hosters Are Ripping Out of Their Stacks in 2026

Every self-hosting blog runs the same piece: here's my glorious 40-container stack, look how independent I am. Nobody writes the other article. The one where a service gets ripped out after it stops being worth the maintenance tax. That piece is more useful, because it tells you where the actual pain lives instead of where the marketing lives.

I went looking for what self-hosters are dropping in 2026, and the picture split three ways. One category is getting abandoned for structural reasons that aren't going away. One is a genuine fight with people on both sides. And one thing everyone assumed was on the way out is doing the opposite.

Email: the one nobody's defending anymore

This is the clearest signal in the data, and it's not close. In November 2025, Gmail started rejecting non-compliant messages at the SMTP protocol level: hard bounces with 5.7.26 errors for authentication failures, not a quiet drop into spam. Microsoft followed with 550 5.7.515 rejections for senders missing proper SPF, DKIM, and DMARC. Yahoo aligned on the same timeline. If your mail server can't prove its identity, the message never reaches any folder at all.

Running a self-hosted mail server in 2026 means correctly maintaining SPF, DKIM, DMARC, MTA-STS, TLS, DNSSEC, and reverse DNS, at minimum, and then winning a reputation battle that has nothing to do with your configuration being correct. New IPs start at zero trust. VPS address ranges carry guilt by association from every spammer who's ever rented a box on the same provider. One blocklist entry can silently kill delivery to millions of inboxes, with no error message telling you it happened.

One self-hoster who documented giving up on it put it plainly: a decade ago, running your own mail server was a rite of passage. Now it's "absolutely asinine." He gave the specific reasons: DKIM, SPF, reverse DNS, and IP reputation turned a project that used to work by pointing MX records at a box into a permanent, adversarial maintenance job. Consumer and residential internet connections often can't even send port 25 traffic without a business-grade circuit.

The workaround gaining traction isn't abandoning self-hosted mail entirely. It's hybrid: self-host inbound (comparatively easy, and it's where the privacy win lives), and route outbound delivery through a service like Amazon SES, Postmark, or Mailgun to buy someone else's reputation instead of building your own from zero. If you're running Mailcow, that split is worth setting up before you hit your first mass rejection event, not after.

Password managers: a genuine fight, not a trend

This is the one where I expected a clean story and found a real argument instead.

On one side: a self-hoster who ran Vaultwarden on a NAS through multiple power outages and a RAID failure finally moved to a stateless, serverless Bitwarden-compatible setup running entirely on Cloudflare Workers, backed by D1 and daily S3 backups. His reasoning wasn't philosophical. It was that the effort spent on backup and disaster recovery for a password vault stopped being worth it compared to a subscription. Separately, a longtime self-hoster who wrote up three services he dropped in 2025 named password management first, describing the specific anxiety of self-hosting the one service where a missed security update is unacceptable, and where you are personally responsible for uptime on infrastructure you're not confident you can keep patched.

On the other side: a Reddit thread with over 400 comments and 400-plus upvotes on the same topic reached close to the opposite conclusion. The consistent theme was that password managers are the one self-hosted service people are least willing to give up, because the alternative (trusting your credentials to a cloud provider's breach history) feels worse than the maintenance burden. One commenter's framing: they'd rather spend a weekend setting up Vaultwarden than spend years wondering whether a cloud provider got breached without disclosing it.

Both groups are reasoning correctly from different risk models. If your failure mode is "I lose a weekend to a botched restore," that's an annoyance. If your failure mode is "a provider I don't control gets breached and I find out from a news article," that's a different kind of exposure entirely. Neither side is wrong. They're weighing different things.

If you're running Vaultwarden and haven't tested your backup restore process end to end in the last few months, that's the actual gap the "people are dropping it" crowd is reacting to. Fix the backup story before deciding whether to keep the service.

Photo backup and notes: quietly going hybrid, not going away

Two smaller but consistent patterns showed up. Immich users aren't abandoning self-hosted photo storage. They're demoting it to an archive tier and putting a cloud service in front as the daily-driver for sync, sharing, and mobile access, because syncing reliably from a phone over an unreliable home connection (even through Tailscale) turned out to be the actual weak point, not the storage itself. Notes apps followed the same shape: people who self-hosted Obsidian sync across a phone, tablet, and laptop hit enough sync conflicts and offline-access gaps that they moved to a simple hosted note app and kept self-hosting for everything with lower sync demands.

The pattern across both: it's not the self-hosted service failing at its core job. It's the always-available, always-in-sync requirement of mobile-first daily use running into a home network's actual reliability ceiling.

The one everyone assumes is dying, and isn't

Here's where the "people are giving up on self-hosting" narrative runs into contrary evidence. Self-hosted music is not being ripped out in 2026. It's one of the fastest-growing categories of new self-hosted deployments, and the direction is the opposite of what you'd expect: people are actively leaving Spotify and Apple Music for self-hosted Navidrome and Jellyfin setups, not the other way around.

The reasoning shows up over and over in the same shape: years of ripped CDs and purchased albums sitting untouched on a NAS, a subscription fee for a catalog you don't control, and a realization that Navidrome (a single static Go binary, runs comfortably on 128MB of RAM) turns that dead library into something people use daily. Subsonic-API compatibility means more than 40 third-party clients can connect to it, so nobody's stuck with one janky official app.

The reason this holds up where email and password managers don't is failure mode. If your Navidrome server is down for a day, you can't stream your own music library for a day. That's mildly annoying. It's nothing like getting locked out of every account tied to a dead password vault, or watching legitimate mail silently vanish because a shared VPS IP picked up someone else's spam reputation. Low-stakes services survive maintenance gaps. High-stakes, adversarial, or always-on services don't.

What's out of scope here

This piece covers what individual self-hosters are dropping and keeping based on documented experience and community sentiment, not enterprise self-hosting decisions, which run on a completely different cost and compliance calculus. It also doesn't cover home automation or Home Assistant, which has its own distinct set of reliability complaints that deserve separate treatment.

The actual signal

The services people are abandoning share a trait: they put you in an adversarial relationship with an entity that doesn't care about your setup (Gmail's spam filters, a determined attacker probing your vault). The services people are doubling down on are ones where a bad day means inconvenience, not exposure. If you're deciding what to self-host next in 2026, that's the filter worth applying before you spin up a container: what's the actual cost of this thing going down for a weekend while you fix it?