GitLab had two separate incidents on 20 October 2026: a brief run of intermittent 500s on GitLab.com's project pages, and a longer, 3 hour 28 minute spell of 404 errors on the Customers Portal. Both were GitLab's own infrastructure problems, confirmed and resolved on their status page.
Started
21:57 UTC
Duration
Lasted 3h 28m
Source
IsDown
Next time
GitLab is degraded. Is your product working?
An incident on their side does not always mean an outage on yours, and the only way to know is to already be watching your own endpoints. Set that up in under a minute and you will be notified the next time your own site is affected. And there will be a next time.
50 monitors for your own websites. No card.
Affected components: GitLab Customers Portal, Website
What happened?
The first incident started at 07:11 UTC and ran for 28 minutes. GitLab's own update, posted at 07:08, described intermittent HTTP 503 errors when opening project pages, and noted the issue was limited to specific projects rather than the whole site. No further updates were published for this one beyond the initial notice and its resolution, so there's no detail on what caused it or how many projects were affected. The second incident began at 21:57 UTC and lasted 3 hours and 28 minutes, affecting the GitLab Customers Portal. GitLab opened with a short note at 21:56 saying they were investigating 404s on the customers panel, followed by a holding update at 22:47 that repeated the same wording without new information. The fix landed at 23:51, when GitLab reported the portal was loading normally again, that the intermittent 404s on sign-in and page loading had stopped, and that a redeployment had completed successfully. They said they were monitoring for recurrence before formally closing the incident. Two components were listed as affected across the day: Website and GitLab Customers Portal. No user reports were filed during either window, so the only account of impact is GitLab's own. Over the past 90 days GitLab has logged 39 incidents with a median duration of 3 hours 32 minutes, and the Customers Portal incident here lands almost exactly on that median. Across all time, GitLab averages 7.6 incidents a month, so two incidents in a single day sits above a typical day but not outside its usual pattern.
Learning
UptimeRobot's own 30 day figures show 20 incidents for GitLab, all resolved, with a median duration of 3h 30m, a number that lines up closely with GitLab's self-reported median. That consistency matters because it means an external monitor watching GitLab's public endpoints over time will tend to see the same shape of outages that GitLab itself reports, even though no single check caught either incident the moment it started. Both of today's problems were backend failures, intermittent 500s tied to specific projects and a portal serving 404s during sign-in, the kind of partial, project-specific or session-specific fault that a simple uptime ping against the homepage can easily miss while real users hit errors. That's the general shape to watch for with GitLab: errors that come and go depending on which project or which session you land on, rather than a clean full outage. If your own status checks only hit one or two pages, you can stay green through exactly this kind of incident. Your monitoring covers your half, UptimeRobot now watches the provider's half too, so next time you know which side broke without guessing.