GitLab.com had two outages on 2026-09-24, together stretching over 12 hours across the API, Git operations and CI
GitLab.com's public status page records two separate incidents on 2026-09-24 that together kept the API, website, Git operations and CI degraded or unavailable for more than twelve hours.
Why it mattersA team that pushes to GitLab.com as its single source of truth for code, CI and release artefacts lost more than half a working day, and the status page ran behind the reality on both incidents.
A team that pushes to GitLab.com had an unproductive Wednesday. The GitLab System Status history now records two separate incidents on 2026-09-24 that between them kept the API, website, Git operations and CI degraded or offline for more than twelve hours.
What broke, and when
Per the GitLab status history, the first incident, "500 errors while accessing projects", opened at 13:48 UTC on 2026-09-24 and resolved at 20:50 UTC the same day, seven hours and two minutes later. Website and Git operations were listed as affected. The status update on resolution reads: "We're no longer seeing elevated errors, so we are resolving this incident."
The second incident, "503s on GitLab.com", opened at 23:04 UTC on 2026-09-24 and resolved at 04:37 UTC the next morning, another five and a half hours. The affected services list on this one is longer: the website, API, Git operations, Package Registry, Container Registry, GitLab Pages, all CI runner types, SAML SSO, background processing and the support infrastructure. GitLab's final update states that "GitLab.com is operating normally, and we are continuing to monitor the service".
Neither incident writeup on the public history names a root cause. The status page also does not link to a longer postmortem yet.
The status page ran behind the reality
The second outage was the top story on Hacker News for several hours while it was still live, and one detail is worth pulling out because it changes what an operator should trust. On the thread, users reported being unable to push or pull with either SSH or HTTPS, and getting 503 errors from the web interface, while the status page still had "Git operations" green. One commenter put the lag at 10 to 15 minutes.
A second thread of comments describes what looked like a coordinated brute-force wave against MFA-protected admin accounts from thousands of IP addresses during the same window, and another user pointed at a separate concern about per-user GitLab-issued email addresses being usable as a credential. Neither of those is confirmed by GitLab. They are what other operators noticed and reported at the same time.
What a team can do this week
The two obvious lessons out of a Wednesday like this one. First, if your release path uses GitLab.com as its only source of truth for source code, CI and container images, a long outage on the provider is the whole plan gone for the day. A read replica of the repositories on a second remote, or a mirrored container registry, is what a team goes to during the next incident. Second, the status page is not authoritative in the first fifteen minutes: your own smoke test that clones a small repository and hits the API is faster to trust than the public dashboard.
The incidents are marked resolved as of 04:37 UTC on 2026-09-25 and GitLab.com has been operating normally since. GitLab has not yet posted a root cause explanation for either one.
Source
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
