GitHub Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

GitHub Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Mahmoud (@thisismahmoud) reported 3 seconds ago

  @cderm @Railway hmm, I'm unable to reproduce the issue. Login with GitHub works for me in both Chrome and Firefox Are you consistently running into this?

• 
  Kenny (@kenny_aaarrrhhh) reported 40 seconds ago

  Every company thinks its data goldmine is Slack, GitHub, or the email server. Those capture what people decided. Not how they did it. The merged PR doesn't show the four approaches that got rejected. The sent email doesn't show the three drafts before it. Games are different. Games capture the decision itself; every move, under time pressure, with a win or loss attached to it. That's labeled human reasoning at scale. It's the rarest data there is.

• 
  Коллинс💻☁️ (@omwenga_c) reported 52 seconds ago

  @VusiThembekwayo ooh also, a nifty trick, set up strong authentication for X, Meta platforms, Google account, and those like me, your Microsoft account and Github. Then use those ones to sign in to whatever platform. Most of them support these logins. I avoid the ones that don't if I can.

• 
  PADRE🏝️ (@0xpadre) reported 2 minutes ago

  Cursor is coming for GitHub and it's not subtle anymore. GitHub went down for 7 and a half hours. Copilot, Actions, merges, all of it. 15,000 devs hit the outage. Every minute of downtime is a migration notice. AI-native tools take the code layer next.

• 
  Felix Dolderer (@felix_dolderer) reported 4 minutes ago

  I need some help here. Microsoft Defender for Cloud Apps (MCAS) and GitHub are both in the Microsoft product portfolio. But somehow nobody deemed it worth to test interoperability. At this point I'm looking for alternatives not out of curiosity but out of necessity. GitHub is completely unusable through MCAS and getting worse. Specific breakages: - Merge button on PRs never renders (new since last week) - I can't mark a file as reviewed (new since last week) - GitHub Action progress only appears after after hard refresh Please fix this and add at least one e2e interop test @MSFTSecSuppTeam @msftsecurity @github

• 
  Ruthrasankar (@SRuthrasan71495) reported 5 minutes ago

  Lost 45 mins to Github and Supabase issues 😭😭 Unfortunately I connected the wrong github account to Supabase and when I tried to change, Github asked me to verify my account. And resert my password and email. Then it took some time for the previous repositories to appear. I was afraid to create a new repo when I can't even see my old ones - Result, I lost 45 mins in this process 🤯 I don't think I can pull it within the next 50 mins. What should I do?

• 
  Stats Wire (@StatsWire) reported 6 minutes ago

  We’ve reached the point where AI coding agents can solve problems companies never bothered to support. A developer had an HP Laser 1008a that never had a macOS driver. So he gave the problem to Claude Code. 4+ hours later, after trying drivers, debugging errors, digging through open-source code and using HP’s Linux codec inside a container… The printer worked on macOS. Not some hacky “run this script every time” setup. Just Cmd+P → Print. The code is now on GitHub. This is what makes coding agents interesting to me. They’re not just generating code anymore. They’re becoming the missing engineering layer between “this doesn’t work” and “I made it work.”

• 
  Arham (@arhamsai) reported 7 minutes ago

  @yoimkonrad This was on a PR that contained an entire build which was about 15,000 lines. From the initial commits CodeRabbit gave the most findings, almost 2x more than Greptile and Codex, however after 2-3 rounds of fixes it didn't find anything while Codex kept giving 2-3 findings per round. Codex kept doing this for about ~20 further rounds and CodeRabbit did give findings on a handful of those commits as well, but what I have majorly noticed is that after a fix Codex tends to find other things that came up elsewhere because of the fix. I'm not sure if this is because CodeRabbit has an incremental review system and Codex maybe does it differently, but it does feel like it rereviews everything again. On the experience side however CodeRabbit is much much better, Codex does not give a prompt and doesn't even use GitHub checks, also been loving the autofix feature in CodeRabbit.

• 
  Guy Swann (@TheGuySwann) reported 8 minutes ago

  @axexang @littlejoeward @MeiYiXing1 As I said, “it has no bearing whatever on the discussion of whether a small or large number of people can change the consensus rules or not.” A “gentleman’s agreement” has nothing to do with Bitcoin. The most it can reveal is social problems with how the github project is handled.

• 
  Julien Zammit (@JulienZammit1) reported 15 minutes ago

  5. can your agent edit .github/workflows? if a stranger can open an issue, a stranger can reach your CI. 1 min

• 
  Dr Milan Milanović (@milan_milanovic) reported 15 minutes ago

  We had a big GitHub 8h outage, again On August 17, GitHub services degraded significantly for a few hours, which affected many developers on the platform. At the peak of the outage, error rates were around 20%, while content download errors were 50%. GitHub managed to restore most of the service within three hours, yet the GitHub Actions and Copilot Token Service didn't work for prolonged time. All in all, it took almost eight hours to fix it properly. The impact was, on GitHub .com, PRs, Actions, Copilot and more. Now, GitHub published a detailed root cause analysis (RCA) for the outage, and here is what we learned: - The immediate cause of failure was network saturation on load balancers in Central US, due to increased traffic. - This was caused by an Istio sidecar (a Kubernetes pod with multiple containers) which reached its concurrency limits and failed to auto scale correctly. - Then failure spread to four HAProxy (reverse proxy which route TCP and HTTP traffic) nodes and degraded the gateway auth path - Then clients made it even worse, as failed Copilot token calls entered retry loops and a latent retry bug in VS Code multiplied traffic by 10X. A service that usually has 7-9K requests per second, all of a sudden had 70-100K. GitHub stopped it by returning 403 at the load balancer. And this was the second GitHub outage in a month. The main issue is that monthly commits grew from 1.4 to 2.9 billion since April, but the capacity didn't support this. What I would check in any system after reading this postmortem: 🔹 Retry budgets. Cap total retries per request path, not per client. 🔹 Exponential backoff with jitter, everywhere a retry exists. 🔹 Circuit breakers that fail fast instead of queueing pressure. 🔹 Autoscaling limits on the proxy layer, not only on the service behind it. 🔹 A written plan for the day GitHub, or your GitHub, is degraded for a full working day. Source: GitHub

• 
  Florian🪶 (httpz) (@zktrivo) reported 16 minutes ago

  @zksecurityXYZ no wonder github went down the other day

• 
  Pratik Sharda (@pratiksharda2) reported 23 minutes ago

  Thousands of passengers injured every year by turbulence no one sees coming. Clear air. No clouds. Radar misses it completely. A problem that can't be sensed, only felt. One of our buildathon engineers put an ML model on an ESP32 reading an IMU at 400 kHz. Classifies severity on device. No cloud. Streams live telemetry to a 3D flight viz in the browser. In just 4 hours... Github repo in comments.

• 
  Ramtin (@ramtinrhmti) reported 27 minutes ago

  Day 2/180 UPDATE yesterday I asked if I should build or validate. did some digging today but not the kind that actually counts yet killed the clinic booking idea, a funded competitor already does exactly what I wanted to build. glad I checked before sinking a month into it pivoted to something I deal with myself, AI agents saying a task is done when it isnt. before touching code I checked 5 real PRs an AI coding agent made against the actual github issues they were supposed to fix, just to see if the problem is even real 2 were legit. 3 got closed without merging even though they looked fine on paper. so the problem seems real at least on paper but I havent talked to a single person about this. no dms, no "would you use this," nothing. that's still the actual next step, not more building scaffolded it anyway bc I wanted to know if it was technically doable. a github app + a small action, both build and run. broke something mid way, fixed it properly not attached to this yet. could easily be wrong, just havent found out yet

• 
  Dr. Jaime Alnassim, M.S. (@JaimeAlnassim) reported 29 minutes ago

  @austinginder @AnchorHost On github WP Rocket had the bug reported a month before. So they choice not to fix it till sites crashed