Codeberg.org Is Offline After Its Registrar Put the Domain on Hold, and a Forgejo Security Fix Is Stuck
Codeberg.org, the non-profit code-hosting site many open-source projects use instead of GitHub, has been unreachable since early Wednesday morning.
Its servers are running. Its name isn’t. The public registry record for codeberg.org shows a “client hold”, a status only the domain’s registrar can set, added at 05:52 UTC on 18 November. The registrar is NameSilo. Once DNS caches ran out, codeberg.org stopped resolving for everyone, and the first reports of that came in at about 07:20 UTC.
Codeberg says the domain “entirely disappeared” from its NameSilo account. It didn’t forget to renew, it sees no suspicious logins, and the account’s domain history shows nothing. It also got no email warning it about a problem. A support ticket is open with NameSilo.
The outage has a second victim. Forgejo, the software Codeberg runs and many others host on their own servers, had to stop a security release that was due today, after its patches had already been publicly visible for a while. Forgejo’s interim advice for self-hosters is in chapter 3.
When we published, the outage was being discussed on Mastodon and Hacker News. We found no news coverage.
What Happened, Hour by Hour
All times are UTC on Wednesday, 18 November.
- 05:52: the registry record for codeberg.org changes to “client hold” (next to “client transfer prohibited”, a lock against moving the domain). The record is public and anyone can look it up.
- About 07:20: users on Mastodon report that codeberg.org returns no address at all from Google’s and their own DNS resolvers.
- 07:46: Forgejo announces that it can’t go ahead with today’s security release “due to urgent issues with Codeberg”.
- 08:07: Codeberg opens an incident on its status page, which runs on its second domain, codeberg.eu.
- 08:13: Codeberg’s status account says the domain has disappeared from its NameSilo account and that it is trying to set up a workaround on codeberg.eu.
Everything under codeberg.org is down: the website, cloning and pushing over HTTPS and SSH, the CI services, the container and package registry, the documentation, the forum and the translation service. Codeberg Pages sites on codeberg.page still load, because that’s a separate domain.
One thing that doesn’t look at risk is the name itself. The registration runs until July 2028, and the transfer lock means nobody can move the domain to another registrar while this is sorted out.
Why a Registrar Can Switch Off a Website
A domain name is rented through a registrar. The registrar can tell the registry to stop publishing the name in DNS, and that is what “client hold” does: the domain stays registered, but browsers, git and every other program can no longer find the server behind it.
NameSilo’s own help page lists two common reasons for a hold. Either the owner didn’t confirm their contact details within 15 days, which ICANN rules require, or someone reported the domain for spam, phishing or malware and NameSilo’s abuse team suspended it “pending investigation”. Abuse holds can’t be fixed through live chat, the page says.
Neither NameSilo nor Codeberg has said what caused this hold, and a domain disappearing from the owner’s account entirely is not something that page describes. We’re not going to guess.
What the morning shows is plain enough. People move their projects to Codeberg to depend less on one big company. The Zig programming language did it in November 2025. But every site still hangs on the company that rents it its name, and today that one link was enough to take the whole service offline.
The Forgejo Security Release That Can't Ship
Forgejo is the free software behind Codeberg, and it runs on many self-hosted servers too, from companies to hobbyists. Its releases depend on codeberg.org: the container images, the release notes and the security announcements live there.
A security release was planned for today. In its announcement at 07:46 UTC, Forgejo said the patches “were temporarily publicly accessible as part of the release process”, and that the Codeberg outage keeps it from finishing the release.
That’s an awkward gap. Once fixes are public, anyone can read them and work out the flaw, while admins can’t yet install the update. Forgejo hasn’t said what the flaw is or which versions it affects. Its advice for now:
- if you use Forgejo’s built-in SSH server, turn it off for the time being or block access to it
- in the configuration that’s
START_SSH_SERVER = truein the[server]section (Forgejo’s settings reference); in Docker it’s usually set through the environment - Forgejo’s rootless Docker images use the built-in server, so if you run one of those, this applies to you
Servers that hand SSH to the system’s own OpenSSH aren’t covered by the advice. Turning the built-in server off means your users can’t push over SSH until you turn it back on, so tell them to use HTTPS in the meantime.
When the release does ship, Forgejo’s documentation names a mirror for its container images, data.forgejo.org, for times when codeberg.org can’t be reached.
What to Do Now
data.forgejo.orgSTART_SSH_SERVER = false in the [server] section) or block its port until the security release is outIf you host your own projects on Codeberg, one thing worth doing once it’s back: add a second remote, on another forge or your own server, so the next outage costs you an afternoon and not a release.
We're Contacting NameSilo and Codeberg
We’re asking NameSilo why codeberg.org was put on hold and how it disappeared from Codeberg’s account. We’re also asking Codeberg whether it has an answer yet.
We’ll update this article when they reply or when the domain comes back.
Our Mission
We believe security online security matters and its our mission to make it a safer place.