I don't think this portents anything great for software in general.
We're a good year+ into the use LLMs for all major bits of software that we all rely upon and GitHub here is down to one 9 of uptime. I've been using GitHub for a _long_ time, my first commits there go back to August 2009!, and I honestly don't recall GitHub going down as much as it has in the last year.
I'm sure there's other things happening in the background, but I can not help but believe that this is directly correlated with the increase of LLM usage.
Though I would love to hear someone else's pet theory how a rock of the internet went from four+ nines of uptime to maybe one.
Are these outages caused by introduced bugs, though, or by load issues?
As someone who has spent many years working in high load environments, this is not an uncommon pattern.
You design a system and it works great. It can handle failures, load spikes, it is horizontally scalable, things are great. You think you figured it out.
And then load keeps increasing and you suddenly hit a tipping point where everything keeps failing, and you cant keep up. The things that you thought were perfectly horizontally scalable turn out to have a bottleneck you didn’t even think about until you got to a truly massive scale. Your systems suddenly don’t have the excess capacity to handle load spikes or catchup work, so suddenly any failure cascades and recovery is more and more difficult. You can’t solve the problem with additional hardware, and your perfect scalable design actually can’t scale any more.
This doesn’t have to be about GitHub using LLMs in their code to still be related to LLMs. GitHub gets a lot more commits now because of LLMs and probably get a lot more reads because of LLMs as well.
It could be that the extra usage just pushed them past one of those capacity thresholds.
Which is somewhat humorous because it was arguably more stable when they ran on their own hosted colo infra before moving to Azure. This was a choice versus keeping the infra compartmentalized and using Azure for elastic overflow compute needs. I'm sure marketing and bonuses rest on throwing it all on the Azure quicksand though.
But that's just a perfectly fitting excuse for GitHub degradation.
Web search, steaming, high frequency trading, and many other systems are resource demanding, and keep scaling. Whether to handle the surge due to bots or wider adoption. But GitHub can't scale git? It isn't even git failing.
Incompetence in leadership is what makes a tech business technically unable to meet growing demand.
Which wasn't a problem before MS acquisition. Git is the write intensive process. It scaled to millions of contributors. Then couldn't keep scaling. But sure if you believe the narrative and blame ai bots.
> Incompetence in leadership is what makes a tech business technically unable to meet growing demand.
I don't think this is true. Take LLMs for example, you need GPUs to serve them and these are in short supply at the moment making it hard to meet demand. I don't think this is necessarily the fault of incompetent leadership.
Actions is one thing, that seems like a difficult, dynamic and bursty thing to host, even before the Vibe Cambrian Explosion. And _nothing_ at scale is easy... But Pages? The static sites? Down for so many hours? Oof.
To me this correlates more to them being bought by Microsoft, a company known for being seemingly incapable of creating quality software to the point that it's not even funny anymore, and also known for sloppifying all the products they touch.
GitHub Actions was always a mess. Because it is a Microsoft product.
If you want to have a horrible time, try reading some of the runner code. It's early 2010s-style Windows-First, MSFT C# crud that has trouble not racing several threads to inconclusive status codes.
I don't think so. GitHub was bought by Microsoft 8 years ago and people have only started complaining about its uptime in the last year or so - exactly correlating with the surge in LLM use.
It's not the whole story. The biggest change is actually the internal rules for how downtime was reported, it wasn't actually such a large change in the actual reliability then.
I don't think so. I've seen those complaints for more than a year.
I have an Ops background and I strongly suspect they were given a stupid timeline for the Azure migration.
I've got to believe Microsoft have decent Ops people but the management wanted to move faster than was reasonable and screwed it up. Move one thing at a time and double check it all works and you can do a migration like this.
> people have only started complaining about its uptime in the last year or so
I'm sorry but this made me laugh out loud. That isn't true at all, this has been going on for years. This conversation[0] from six years ago has discussion about the outages starting to become much more frequent in December 2019. It has never gotten better in that time, it's just continually degraded.
Ok I should have said people have only started constantly complaining in the past year or so. Of course there was the odd complaint in the past, but now I see a thread about it at least ever month on HN, and loads of articles about migration off GitHub.
It's true, Microsoft made github worse, but the more recent issues seem to have to do a lot more with Microsoft selling its soul to AI. Microsoft really appears to have gotten dumber as they became dependent on AI. Most recent example: they used to promote Win11 and 32GB RAM. Now they are down to 8GB silently ... this is quite hilarious. There are now so many side effects that you see degradation in so many other areas. Or the gaming industry: it is not quite dying but it is taking a huge hit with skyrocketing RAM prices. Consoles selling less is an example here. It's quite fascinating how deadly disruptive AI is now.
This really needs to have a lot more evidence that it’s because GitHub’s code is being written by AI vs them being bombarded by activity from all kinds of AI agents all over the world vs they were already on a trajectory of quality loss.
It's caused by human laziness and corner cutting. LLMs can write buggy code or incomplete architectures as much as humans can, but standards have lowered. It's not the LLMs are not capable of also fixing these same issues, but that's additional work.
They mentioned earlier this year that they were beginning the migration to Azure and that it would take a couple years. I would assume it has more to do with that migration then anything else.
The funny thing is that if companies wanted they could probably use AI to instead increase uptime. Keep existing QA teams (instead of replacing them) and then use AI for better and more timely monitoring (and messaging even) and as an additional Always-Testing™ layer of QA
I wonder if GitHub actions was a bad idea? Like maybe it's being abused for other kinds of compute besides just builds? And even builds themselves can require a lot of compute. I've only recently had a repo there where I wanted to do builds to make a release (both linux binaries and WASM) and whenever I do that tag and wait a few minutes for those builds to finish I think about all the other projects/repos out there on GitHub doing the same.
I'm really kind of surprised they let us do that - like, why didn't they just have you upload the binaries after building on your local machine?
> why didn't they just have you upload the binaries after building on your local machine?
You can do that already with GH Releases. Actions is if you want CI/CD managed by GitHub. And you can also use your own machines via self-hosted runners.
It is definitely and absolutely caused by LLMs. I must do 20x more GitHub operations now, and since the agents know GitHub far better than me, I'm using more advanced features. Multiply this times all of us.
someone from github posted a usage graph from the last year on twitter a while ago and they were serving like 14x more requests in a matter of months. it's frankly impressive they've kept up.
It's probably not attributable to AI in the way that you're thinking - Github has been absorbing an exponential increase in usage, and that increase is mostly due to new AI-related projects being created and worked on.
Though I'm sure some of the blame can go to internal slop code.
For the longest time, I didn't appreciate the "AI-induced traffic" excuse. But seriously, I checked the rough github egress for our lab versus an old log from 2024, and there's an order of magnitude or two difference. From asking around, it seems people all have the gh cli tool and let it loose with parallel tool calls and e.g LLM's polling Actions in a background bash while loop with sleep $TOO_FEW_SECONDS. Some people use a variety of skills where the agent makes a commit every few code changes, and uses Issues for its memory/log. And they have O(5) sessions at the same time doing all kinds of crap. It's the same with PR checks/PRs. Recently we also saw continued usage throughout the night as well, which did not exist pre coding agents. Loops or whatever they call cron jobs in the harnesses these days is the reason. It must be adding up.
I honestly don't care either way. Either they are using AI as an excuse to hide their incompetence, or they are constantly going down due to the AI they have been promoting.
If your platform can't handle the use patterns of AI, then perhaps don't go around telling everyone to use AI for everything? It's a self-inflicted wound, you could also just not do this. Too bad Microsoft has bet its future on AI not being a giant bubble, huh?
Someone else quoted a 14x increase in load due to AI hammering github. I wouldn't be at all shocked if that was true, considering how much AI tools are DDOSing the entire internet.
That plus migrating clouds is insanely difficult to manage. They're almost certainly drowning in traffic and trying to keep up.
All of the outages make sense to me as scaling issues. Each month, GitHub is getting the amount of commits they’d normally get in a year. And it’s growing
> Yup, platform activity is surging. There were 1 billion commits in 2025. Now, it's 275 million per week, on pace for 14 billion this year if growth remains linear (spoiler: it won't.)
GitHub Actions has grown from 500M minutes/week in 2023 to 1B minutes/week in 2025, and now 2.1B minutes so far this week.
So we're pushing incredibly hard on more CPUs, scaling services, and strengthening GitHub’s core features.
And as a fine purveyor of hand-crafted shit code for many years, I'm not gonna weigh in on that.
Even if their actions running servers were at 50% load (or even 20%) at the end of 2025, they'd be screwed right now, at this point it's as much a tech/scaling problem as it is a CapEx problem and a construction one. Given the public's responses to AI specific datacenters you can't pick everything.
Even my own systems at home are burgeoning under the load of the my more ambitious hobby projects I'm doing for fun. I don't envy "we had to build five more datacenters to keep up with demand" class problems just from a how many people have to sign off on them perspective let alone the technical difficulty of doing so
They don't hire well at all. You look at who is on their research team and it's people with social science PHDs, not computer science. Looking at the state of the site it's not a surprise. Highly incompetent people have taken over for a while. Let's not forget a member of the original executive team was a sex pest too.
For like 10 years the only feature development they did was by stealing ideas from GitLab. I wouldn't be shocked if little if none engineering discipline has took place at all during this time if it's this brittle to frequent change.
Guessing it was mostly held together with duct tape and poorly written tests/monitoring systems if any other corporate driven software.
We're really going to find out over the next few years which businesses have good practices or not.
Are you kidding? That's an insane amount of load increase to manage. "Scalability" isn't one thing, especially not at that level, so it's ridiculous to knock them for it. That amount of load ripples across your entire infrastructure.
They have the backing of a trillion dollar corporation and can literally hire some of the best talent on the planet. People nor budgets are an issue here.
Why do we keep giving excuses to poor engineering disciplines + poor management? This problem is entirely GitHub's making and acting like it's some unplanned natural disaster is low key pathetic.
They don't even let you really run those yourself now either, We moved off a big GHE footprint because the support for it was getting abysmal and features were slowly become GH.com. Ive been dreading the move to GitHub.com and it was worse than I expected.
5 hours going now and the entire thing is still completely down. This level of incompetence and their total disrespect for their customers is unbelievable.
Yeah, I think this one might be the straw that broke the camel's back and our team will look to migrate to something non-microsoft to be away from this nonsense.
And if you thought the GitHub actions hybrid mix of JavaScript templates embedded in yaml was a nightmare, I have bad news for you with respect to ADO pipeline configuration documents...
Yeah I went through hell migrating everything from ADO to GH Actions a year or so ago and even after all this I'm not even slightly close to wanting to go back
I love how even self hosted workers don't work during these outages - running jobs on their infrastructure being flakey is marginally acceptable, but the API to simply schedule workflows having this availability is mind boggling. Github just doesn't seem like a serious company anymore.
GitHub implements self hosted runners by running a normal runner that passes the environment to your runner and then polls it. That's why they cost as much as the smallest GitHub-hosted runner.
More likely is that the log streamer takes up some amount of resources and thus running it requires a baseline level of compute. But I doubt they spin up a small runner per-self-hosted runner, most likely some ruby workers (on distributed azure instances) are dedicated purely to streaming self-hosted runner logs to log consumers.
I'm certainly not surprised that they tried to sneak in billing self-hosted runner minutes at some ratio to potentially recoup the costs here.
Oh that is pretty gross, thanks for sharing. I was not aware of that process, that explains the cost they attempted to roll out for runners too a few months back.
I think they kind of have to operate that way if you want a control plane/access to GitHub’s own layered services like caching and artifact storage. There are plenty of things to blame GitHub for, but this one doesn’t seem sinister at all.
Edit: to be clear, I mean the part where they maintain state on their end. I have no idea if they do that with another runner instance, that seems unlikely.
They never were. They were just really good at maintaining the bluster and backing it up with expensive lawyers. Now that they don't come in claws-first like they did when they were a NASDAQ rocket to the stars, it's easy to see.
Their MO was to court an executive and sell second-rate tools to them before the people who had to use them had a chance to say anything. It doesn't matter how much evidence you can provide to the contrary, once the million dollar deal is signed, you are going to be tasked with finding reasons to say that your executive was shrewd for buying this pile of junk and unfulfilled promises, and not an insane idiot sucking away your job satisfaction as fast as they can.
They did a lot of deals based on how their products would have features their competitors already have 'soon' when they haven't even started them, and a long track record of taking 3 major releases to get from something to good, and then breaking everything again by doing a 4th major release that re-imagined everything and made it horrible again.
I'm not going to claim that Apple was or is a panacea. Apple doesn't use vaporware which is big, and their Cycle of Awful is 2 releases instead of 3. You could afford to skip 1 waiting for the next even-numbered version, instead of being 2 versions behind and getting pressed to upgrade.
They definitely managed to create a lot of reliable products.
Windows NT was one of them. Up to Windows XP the products were pretty solid and each had visible improvement against the previous one.
Their language products were/are still solid IMO. Maybe Visual Studio is sluggish, but we can still use an older version if we want. Plus they put a lot of effort optimizing VSCode, too.
Even back in the MS-DOS/16-bit Windows days, when things broke down quite easily, I think they still provide the best bang for individual users and developers. There was no competitors who could provide so much value back then.
Thanks for sharing. It was that bad back then? As a consumer user the stability brought by the NT kernel was great back then, but maybe enterprise wise it was a different story? I never worked as a sys admin so can't say.
For the last point what was the golden standard back in the mid-late 90s, if we don't include mainframe/minicomputers? Was it Solaris or BSD?
Now that I think about it, I feel sad that I do not have the technical prowess to compare operating systems :/
Beside the NT4 servers we had SunOS 5.x and Digital UNIX (later upgraded to Tru64 UNIX). Yes, from sysadmin point of view they were way more consistent performance wise and less problematic to operate in general. BSDs were not 'approved' but nonetheless we were running DNS and some other internet facing stuff on OpenBSD 2.x if I remember correctly.
reliable? Windows is the only operating system you have to constantly reinstall because it gets slower after a few months. To the point people think it's totally normal to do that!
Microsoft in the 90-s and early 2000-s was a serious company. Windows NT and Office were engineering marvels.
Windows 95 was a brilliant hack that allowed 32-bit GUI programs to run on hardware with 2Mb of RAM (Windows had 4Mb of RAM as the official minimum, but you could run it on 2Mb) while preserving compatibility with the majority of DOS software.
I'm not sure that it's that mind boggling. The entire complexity and value of GHA is not in the runners but the scheduler that the GH frontend hooks into and runners subscribe to jobs from. It's the most likely thing to fall over. '
If the scheduling was self hosted it would be inexcusable but you can always just connect whatever you want to webhooks.
That "scheduling" is just a git hook and a message queue, maybe with some database updates in between with very clear boundaries that make sharding easy to reason about, assuming they have a sane architecture (they evidently don't)
The one thing that falls over is that the webhooks are actually self modifying code since you have an 'on commit' webhook that bootstraps the actions logic from the workflow.yaml file(s) (which are not really config files, they have logic that needs to be evaluated).
I don't disagree that it's obvious they've got problems but I'm just saying it's obvious to me the part that falls over (the scheduling of jobs) and why that would impact self hosted runners, which do no scheduling but depend on it to function.
As for 'just a message queue with some database updates and sharding that's easy to reason about'... Here's a job scheduling problem as an example: imagine you schedule a job, and there's no runner available. How do you disambiguate between no runners available because you've reached capacity, runners not being available because they're on a real network with faulty connections, and runners not being available because of a faulty rollout of internal updates?
A simple message queue for job scheduling is fine if you own everything and can deal with the operational overhead of identifying those cases by hand, but Github can't do that.
> The one thing that falls over is that the webhooks are actually self modifying code since you have an 'on commit' webhook that bootstraps the actions logic from the workflow.yaml file(s) (which are not really config files, they have logic that needs to be evaluated).
Sure, but the part that actually schedules where a 'job' gets run is based on a relatively simplistic tag system. Reading the yaml and plopping some job metadata into a queue-like system isn't where I would expect their issues to be, but at their scale I'm sure everything becomes fragile and inscrutable.
> imagine you schedule a job, and there's no runner available. How do you disambiguate between no runners available because you've reached capacity, runners not being available because they're on a real network with faulty connections, and runners not being available because of a faulty rollout of internal updates?
You don't need to. GitHub Actions runners, and most CI runners that I've interacted with appear to have a pull-based model where they ask for work that matches their declared tags/shape (usually platform/runtime/OS/etc.). This probably amounts to a database query, but who knows.
> A simple message queue for job scheduling is fine if you own everything and can deal with the operational overhead of identifying those cases by hand, but Github can't do that.
I highly doubt it's a simple message queue. My issue is git repos and their CI infrastructure have very low coupling to other repos or entities in most circumstances, at least conceptually, so parts of the system (ie. regions, shards, etc.) should be able to function even when others are down (ie. it shouldn't break for everyone). There's clearly centralization and coupling that isn't obvious from an outside perspective, which sorta tells me it's incidental, but that's a guess.
The only statement I'm making about their system is that it appears to be poorly designed, as evident by their poor uptime. The rest is just visible details about what data is involved in performing the task at hand, agnostic to the implementation underlying it.
The entire complexity and value of GHA is rent-seeking to keep the lights on for other things.
They have a strong motivation (self preservation) to continue to misunderstand the problem. If they did what is best for us, then we could avoid a substantial fraction of all GH subscriptions by using a FOSS tool to hit the Pareto frontier by replicating just enough GH services to watch commits and PRs.
As someone who’s building a github actions control plane replacement, the scheduler is extremely simple. It basically the kost naive version of a FiFO queue.
"1 nine" tends to mean "90% availability" i.e. 37 days of downtime per year.
9% availability would be an uptime of ~33 days a year, I think at that point, we're pushing the semantics of "available" if the service is down the entire year except one month on average.
I once saw a sign at a railway station in the UK that actually did say "normal service will be operating between the 12th of July and 18th of August" (or something like that). Made me laugh. At the time Anglia Railways had a rail replacement bus literally every weekend.
I thought they should rename to Anglia Busways and have bus replacement trains instead.
It's honestly insane how terrible their reliability is. Over the past 4 weeks, 8 days with GitHub Actions outages, many of them multi-hour outages: >3 hours on each of July 9th, July 20th and today, and 1.5 hrs on July 23rd.
Outages happen, but this many outages so close together, and so many of them so major/long lasting, something is systematically wrong for sure. It's been seriously hamstringing our ability to ship code at my company.
Yea just try requesting an SLA credit from them. According to THEIR numbers it’s 99.995%. It’s intolerable and I am going to put my full weight behind stripping as much work as we can from GHA as possible, even if leaving GitHub itself is effectively logistically and contractually impossible.
Based on previous posts I've seen about this, IIRC the timing seems to imply that it has more to do with them being hammered with AI slop than it does with the Azure transition. Who really knows though
Time to touch some grass. Better use of my time and energy than twisting the remaining things on my todo list today to make more progress on them than I have managed. I should have taken a long lunch but I rebased the hell out of a PR instead.
Can't wait for the Graphite/Cursor alternative (Origin) to become available. Actions being down for an hour in the middle of the US East workday is crazy. It's super frustrating to be an enterprise customer of Github, seems like we don't get any dedicated resources or additional stability. At this point our only option is self-host github enterprise or move away.
I feel stuck on Github until the next generation of forges (I assume Linear is working on one too, and the Pierre guys, and there's probably more) are ready. When it works, I'm actually a huge fan of Github. It just doesn't work reliably and it's a huge operational risk to not be able to freaking deploy. We'll most likely switch to self-hosted GHE in the near-term while continuing to evaluate alternative forges — Forgejo, Gitlab, Gitea, Sourcehut, aren't appealing enough to switch.
Given the recent history of outages, I wonder how many are seriously considering moving off GitHub for anything other than code hosting. I mean actions / workflows.
To offer a data point (not a tribalist argument): GitHub Actions being down this often is still less expensive than the cost of switching everything somewhere else. Would rather have my team walk away from the PRs and do other tasks than have engineering effort, meetings, design pages, scheduling, etc. for switching over.
This is annoying and I'm here because it's down. But it would have to be far worse to come close to actually being worth changing.
1. go sign-up elsewhere
2. Log into GitHub and point Actions to that new host
3. All my actions files immediately worked without question
I'd probably give it a spin and make a wiki page explaining how to swap back and forth. No meetings. No design issues. No scheduling. Just a flip switch on who to pay for computers.
For what it's worth, we switched some of our actions to a self-hosted Woodpecker instance, and although there were a few kinks to iron out, it works better overall (for example, because of better caching on that single instance, our docker images build faster).
I’m intensely grateful that despite rumblings, the company I contract to hasn’t switched from self-hosted Jenkins to GHA. Hopefully the steady drumbeat of failure will keep it that way.
We've started writing some new workflows against argo-workflows for the last like 9months or so. Especially for use cases where the actions are like self-service type automation.
I think people were so excited to move away from jenkins to something 'managed' just because of how much a dinosaur jenkins is and how much a pain in the ass it is to upgrade it... but now we are seeing how managed can bite you in the ass if the manager is incompetent.
I think I agree if you don't have someone dedicated to running and maintaining Jenkins. If you have someone at your org that knows what they are doing though, I'd still take Jenkins over GHA any day.
That's significantly more difficult to set up and it cost money. Github's big issues that it's free, Microsoft doesn't want to allocate enough budget to keep the thing running properly.
But for people who either don't pay anything at all or phenomenal amount one 9 of up time is all you need.
If you're actually trying to run a business I guess you can call and gitlab and get an Enterprise contract
We've been working on this at Depot[0]. We built the internals of Depot CI as a general-purpose workflow engine that understands GitHub Actions as an input (with plans to extend) - we've been working to make Actions runners more performant and more reliable for a while, but eventually you hit a ceiling if you don't also control the control plane.
Our deployments are triggered by GitHub Actions, and it is a pain to deploy without it (specifically collecting the credentials and putting them into variables the shell will read).
This is multiple times this month that this has been a problem.
Has GitHub completed it's internal migration to Azure yet? Or is it still ongoing? None of our devs want to switch away from GH, but we will have to at this point.
We have set up a new enterprise and are migrating to GHE. For us this was a no-brainer; we get our own isolated environment, Enterprise Managed Users with Entra OIDC provisioning and Github Copilot in EU data residency. https://eu.githubstatus.com/posts/dashboard looks pretty good to me. We will still keep our github.com enterprise around for public repositories. I’m genuinely curious to know why a company would prefer github.com over GHE?
I wasn't directly involved, but working in a couple of shops that used GH, I think they really steered people away from GHE, my impression was it's a product they don't really want to support - I vaguely recall hearing that upgrades were very painful on GHE, and the scaling / failover story was not great. Can anyone else confirm?
To clarify, we are on Github Enterprise Cloud, not Server. The option for enterprise managed users with your own ghe.com subdomain and data residency options is still prominently shown in the enterprise signup flow. https://docs.github.com/en/enterprise-cloud@latest/enterpris...
My Github actions are now basically: "checkout / set env vars from secrets / docker-compose builder run make".
I used large machine runners to run full the Docker (Podman actually) on Github first to avoid dealing with docker-in-docker complications. This step also provided some very nice robustness advantages, as I can now trigger deployments from my laptop if needed.
Step two:
Migrate to self-hosted runners. I used my former homelab server to set up a build machine. It has 16Tb of fast NVMe SSDs and thanks to Podman container layer caching, my entire lint workflow now takes 30 seconds. Faster than just one "npm install" on Github before.
And Github's self-hosted runners are actually surprisingly easy to set up and use. They are also somewhat more robust.
WoodpeckerCI for DIY, Drone.io for paid support. Self-hosted, OAuth login, repo-permission-based authorization, container-native design, easy web UI, optional RDBMS, supports many platforms. All the core features needed for scalable CI, deploy as few/many as you want for multiple teams, small enough to easily run on one box for one repo. Nothing else is as simple, easy, powerful, compatible.
Until someone makes a better PR experience you guys are stuck with GitHub.
Nobody cares about ATProto or whether your commits are a damn NFT or some bs just literally improve upon the experience.
That’s it.
It’s as if no company is focusing on the product experience or anybody’s experience anymore. It’s all ooo look what I got I got this I can do that too me me me but nobody will ever buy that.
Say what you want about huge companies like Microsoft or Walmart but they spend a lot of energy understanding the human experience to sell products and less on their own perceived self-aggrandizement.
GitHub is the best version control online and it’s not even close.
It would seem like GitHub is in a precarious situation.
We have many agents per employee working in parallel pushing way more commits than was humanly possible before AI, triggering GitHub actions a lot more than the workflows were built for, causing Actions costs to escalate (they really aren't cheap if you compare to hosting it yourself), meanwhile working with YAML workflows is just a pain, and just writing code would be so much more fun and AI compatible[1].
At the same time, GitHub has about ~3 different PR review UIs? And they're all half-bad? Any decently sized PR triggers their "optimized for large PRs" UI which jumps around randomly in my experience. If you don't get that UI and keep the scrolling one (there's an old and a new one btw) then god forbid you click a line number because at some point your browser will randomly scroll back to that line and it won't unstick. Now Linear[2] (and others) is replacing the PR review experience for the agentic era.
I'd love to see a solid AI first Git + CI + reviews.
It's cute that someone at GitHub had time to make their hamburger menu pancakes to promote stacked PRs, but nobody has time to make the platform stable.
It's not too hard to switch over from GitHub actions to other runners. For example, I wrote up the steps needed in order to use Hugging Face Jobs instead, which also enables GPU runners and other flavors. I use this for several of the repos I manage (such as Trackio):
We use other runners (Blacksmith), and it hasn't helped, still hard down all day long. Brutal outage.
As a side note, I started getting these symptoms (actions staying queued forever or not running at all) at 5pm PDT yesterday, intermittently. And definitely full outage by 8:30pm PDT. So for sure full outage for 7 hours and counting, even if you use other runners, and I strongly believe partial outage for ~15.5+ hrs before that, even if it hasn't been acknowledged by GitHub yet.
It most likely is. We are using Ubicloud (which seems to work the same way hf jobs does) and it's also down. GitHub is not sending out Webhooks and is also not able to queue external jobs. If we had to emergency deploy anything now I'd have no idea how to really do that...
Soon we can start posting "GitHub {x} is up!" instead, that's becoming more news-worthy lol. I know they like to blame AI/scaling, but really they are owned by one of the main hyperscalers, so I find that excuse kind of weak..
On other occasions, I'd take this as a time to have a walk because I'm blocked. Unfortunately, I need to get some stuff out for a customer quickly. Can't do anything about that though. GitHub is the only product in it's size class that I use that has this kind of incredibly poor uptime.
I'd love to know what the most common root causes for these outages are.
They are known are engineers pushing features that bloated product management folks keep pushing so they can justify their jobs on LinkedIn.
I'm not sure why this particular industry is so abysmal at making things even semi-reliable after decades of research, educated workforces, and loads of cash.
This whole mess brought to like just now little moat there is left in the AI world.
Like, everyone talks about how you can just prompt a new CRM instead of paying for Salesforce.
But in this case, I raced with Fable and Sol and had time to build a full featured, fully functional CI/CD workflow system that has most of the features of GitHub actions (not the ecosystem obviously) and costs less than the GitHub runners per minute, while being able to scale to millions of workflows... And I had it working end to end before the outage was resolved and was already running my apps deploys through it.
It would take a little longer to do in a company with red tape and more investments in the GitHub ecosystem, and of course it only replaced actions. But it works.
Nothing like the old "everything is normal" to only be corrected 8 minutes later it is actually still down and then 2 hours later they are still working on it.
Aug 06, 2026 - 16:27 UTC - Update - Pages is experiencing degraded performance. We are continuing to investigate.
Aug 06, 2026 - 16:19 UTC - Update - Pages is operating normally.
The most likely explanation for the degraded reliability: Someone at GitHub saw the Three Body Problem last year and started shifting their infra over to Minecraft servers with "physical" memory and redstone logic gates.
This is unconscionable. It's an incredibly bad look to be up an entirety of a workday. When my company migrated from Bitbucket, I never thought that we'd be going to a worse service.
My own company had a migration from GitLab to supposedly greener pastures but at least when a self hosted gitlab server had an issue someone could figure out what was going on. Absolutely zero control over GitHub.
Does anyone recommend an alternative? We've already moved our runners to blacksmith, looking for a control plane as well. We are tiny startup that ships at high velocity and this kind of downtime is highly disruptive for us
There are three other things I could be doing instead, but I noticed it was down when I pushed the last of my GH todo list for the day and my brain is stuck in closure-seeking.
I think this is because all frontier AI providers integrate with GitHub as a default git forge through their GitHub apps platform and their free tier is too generous. They should either charge some of these huge partners access to GitHub apps or partition the free tier off and have a different SLA.
If the frontier AI used GitLab as a default I'm sure they'd be the ones suffering now.
In the news tomorrow: SpaceX's newest Grok model just got out of the sandbox and hacked into Github... Grok is so very very dangerous, their engineers only enter the server room in foil hats
Outages happen. What concerned me most was the time taken for the full resolution (almost 12 hours). I am hoping that root cause analysis promised will look into this as well and how it can be improved.
I knew that when the build actions started failing en masse and my newly submitted PRs did not trigger the build actions at all (besides the mandatory one in my organization), it seemed logical to me to look at HN first..
Especially troublesome in the middle of trying to fix a high score security vulnerability when the release vehicle is Github.
This incident made me look at the SLAs we have with GitHub, since our company is on the Enterprise plan. Interestingly, it seems that only failed GHA invocations are counted towards an SLA breach. We informed engineering to stop using GHA while the incident was ongoing, but it seems what we should have done was to hammer them without remorse to get the whole incident period (10+ hours) counted towards an SLA breach. What a scummy SLA design.
I've never understood this argument. Even if self-hosted things were offline more often (which I've never found to be the case, I've had Forgejo + runners running for a year now with no downtime), the real benefit is that you yourself can work to bring it back online when it does, rather than waiting on a large, slow-moving organization to figure out what slopped PR caused their global service serving ungodly amounts of RPS to go offline again.
Self-hosting a service like GitHub that operates at GitHub scale is difficult.
Self-hosting a service like GitHub that operates at the typical small/medium company's scale is trivial.
A single machine (with separate runners for CI) will cover many companies' needs. It being a single machine eliminates a lot of the complexity and failure modes associated with a distributed system and makes backups/restores/maintenance easy.
Even self-hosted runners are impacted.... How can that be?
The cost of this globally has got to be in the hundreds of millions to companies that use CI/CD through GitHub Actions. What if prod is broken and GitHub actions is stalling the deployment of your hotfix? What if this makes your organization miss and SLA and diminish user trust? What if this makes you miss a release that you were contractually obligated to meet? This is happening during peak dev hours on a Thursday (not that it would be acceptable at any other time).
I don't understand how a service this critical to the global technical infrastructure can fail like this at all, let alone for more than a few hours. Like where's the backup generator for crises like these? You can't even use self-hosted runners? WTF? Like how can you not bring your own backup in a crisis event like this?
Not that Microsoft has a good reputation, but holy moly, you'd think they would prepare from something inevitable like this.
Counterparty risk is still a thing. This is the cost of convenience, reminds me of the milkman joke in that South Park episode. Maybe a global SPOF owned by people who do not care is not the way to go.
It's hard to draw a direct analogy there, but I feel like it echoes the same sentiment.
A soapbox I have is that GHA workflows are scripts that could run on your machine without any of the YAML stuff. Who gives a flying about the DAG or the logs? Which, by the way, if you're willing to walk to the milk store to buy your milk, could be recreated in a much more testable and maintainable way without any of the YAML bs that GHA prescribes....
But DAGs are pretty, and logstreams showing up in a browser application instill trust (for reasons that fly far above the head of yours truly). So people go for that. Pretty DAG, nice logstream; therefore, deliver my milk. All of a sudden.... The CI/CD platform is having its merry way with your SLAs, contract abidements, and hotfix deployments.
What a time to be alive.
Not to rail on the South Park thing, but the blast radius of this issue also reminds me of the episode where the internet dried up.
If this bs with GitHub continues, Parker/Stone will have to make a GitHub episode. How seen would we all feel if that happened?
I feel for you.... This is not a position you should be put in.
There's always the escape hatch of running you GHA workflows locally, but unfortunately, despite the existence of packages like `act`, there is no way to fully recreate the GHA runtime locally. Tons of the special YAML syntax just can't (more accurately, "just doesn't") get interpreted by those local actions runners.
We never went this route, but at my old org, I always advocated for considering GHA to be wrapper around a single bash script (or whatever script you want to run), as a means of completely breaking out of the GHA hellscape that is programming in YAML, who's turing-completeness is pretty dubious.
Unless you have things set up this way, you (the client of GitHub) would have to completely redesign your CI on the fly, run it locally, and then figure out how to get the D compliment of the I to work in a way that is auditable. Fat chance for most teams I bet.
Thank god you're dealing with a batch scenario. Silver lining for sure. Still, embrace the anger.
What makes my blood boil is that there's millions of DEVs literally crying at the moment worrying about how GitHub's failure to be responsible will put their jobs in jeopardy.
And fingers crossed for you my friend. We're at 5+ hours at the time of this writing.... You're batch job may still have a chance!!!
You don't think GitHub working on seven different interfaces to copilot and vibe coding PR page changes filled with bugs are a good use of engineering time?~
I noticed my CI throwing errors all of a sudden. I sure wish they would become more active in alerting folks or build it into these tools - especially as these occurrences are becoming more frequent. Possibly an API-accessible services status. Then, at least we could build in our own checks when we hit errors and not have to hunt down what all is broken.
I was scratching my head wondering why my build file isnt working, ran a working project and it didn't build either. Figured out what might be the case, but wasted an hour.
I was complaining about how I should have used AI instead of manual brainwork for this instead but turns out that might have been the problem.
I have a barely known/used GitHub application and losing it from time to time over some qq#%@ repo that blows up both my actions and CloudFlare workers. Over a few hundreds of repos that were installed, I have now seen two where the authors installed every AI system on the planet. This starts a vicious circle where one app comments on the PR, triggers all other installed apps, who also start commenting on the PR. This quickly reaches the limit of number of comments on the PR BUT GitHub will keep pinging you about any event that is happening to that insane PR. Of course, your app itself will fail to comment so that triggers the redundancy I have including emails from my app about a failure (which I can't read on the email itself for logs privacy, so I need to check my system logs). Just last week, I had a blow up with hundreds of failure emails. All of this was coming from a single dude PR which had nothing but AI apps fighting one another.
Either they had a good month or I'm off my game, I actually spent a minute checking if I did something wrong with actions instead of immediately going to HN to see if it was an outage.
Seems like it’s been solid lately, or at least not quite so obviously broken. I was thinking last week that I hadn’t seen a GitHub outage on HN in quite a while.
Github is a terrible platform. We have had multiple enterprise tickets open with no response for multiple weeks after they fraudulently charged our company for services we cancelled weeks in advance. If you value any kind of real customer support or response time, I would avoid Github like the plague. They will steal your money and there is no one to even talk to about it. Their customer support is non existent (probably because their platform is in shambles).
I may be wrong, but I think this could be traced to the Microsoft's increasing reliance on the outsourced engineering workforce. Regardless of the country it has been outsourced to, the quality of service will suffer, apparently greatly as we now see.
It seems to manifest in many places: AI slop, Github outages, recently Miscoroft sent me an email demanding that I subscribe to 365, in order for MS Office (which I already paid for) to continue to work. I simply moved on to the other, free, provider.
Perhaps, it's time to evaluate the use of the Microsoft products?
It's really crazy to me they can just be down for hours and can't recover their own systems. It shows they don't have the capability or infrastructure to roll back disastrous changes. These kinds of things are a tell on the organization and operational excellence (or not). As soon as a viable alternative surfaces for Github, I'm moving off and will advise all my clients to do so as well.
My experience working at LinkedIn was that, after Microsoft took over, they offshored a huge part of their infrastructure and turned the entire culture into one of performative based engineering instead of results based engineering.
If you implemented a tool and it worked one time for your presentation to management that's all that mattered.
The actual company employees using it downstream in prod basically had to constantly QA the alpha software they were forced to use and the authors of the tool were hard to track down if they even still worked there. And if you did find the author or the team they would be very resistant to admitting there was an issue because it LOOKED bad.
So many tools I used were fragile and buggy, it was clear the authors just presented the happy path to management to get the note added to their promotion packet and the rest of the company just had to deal with the fallout.
My team implemented this product that the entire company used that was broken and buggy as hell but they kept presenting the product to management as this amazing product and nothing was ever done about how broken it was. One of my team members came from Apple and said Apple's tool to do the same thing was much better. The tool my team worked on was a well known pain point amongst the rank and file but management was very detached from the rank and file, which I guess ultimately was the primary problem.
If Github is having the same issues I feel for them.
Someone needs to work on improving the way Actions deals with resource exhaustion like this. Today it exploded all the CI jobs with "failed" and an inscrutable message about "The job was not acquired by Runner of type hosted even after multiple attempts" then "Internal server error". It should say "Timed out due to no runner available, our apologies, try again later, or pay $$$ here to get priority for runners".
Platform activity is surging. There were 1 billion commits in 2025.
Now, it's 275 million per week, on pace for 14 billion this year if
growth remains linear (spoiler: it won't.)
GitHub Actions has grown from 500M minutes/week in 2023 to 1B minutes/week
in 2025, and now 2.1B minutes so far this week.
So we're pushing incredibly hard on more CPUs, scaling services, and
strengthening GitHub’s core features.
I find this hard to believe in this context. They should be utilizing load shedding or admission control and killing/rejecting jobs rather than hard failures if it is a scaling issue. It's much more likely an actual software defect than just more load. If this was the case (which they would likely prefer) free/public tiers would be removed first to preserve paying customers services.
Why not both? Higher base load combined with insufficient internal controls for ratelimiting/load-shedding (as in, they don’t know who to shed) would be explanatory.
If they can't implement something as simple as "decode upstream headers and determine if 429/503" I don't know what to say. Since this has knocked out all customers it indicates they likely don't have anything of this form implemented.
They’ve said that, but it doesn’t add up. The outages started shortly after the migration to Azure became the top priority, and before the load massively increased form agentic coding (by their own dates).
Ever since Microsoft took over, and then when AI Skynet took control, things started to decay. AI companies owe all of use a lot of money. By the way, why do I have to pay for increasing RAM prices here? Why are we so dependent on a few greedy corporations, anyway?
As I said 6 years ago. [0] You would be better off self-hosting than using GitHub or GitHub actions. No CEO of GitHub exists and now it is falling over again.
When the barrier to entry for software was high, free unlimited publicly hosted git repositories made sense. Not only did you have to have some basic understanding of software, but a grasp on git itself.
Now that the calculus has changed, I think we should start asking ourselves whether anyone has the right to unlimited free public repositories.
We can all agree that trying to attribute any single metric to repo quality is subject to Goodhart's Law[1] (e.g. what happened to GitHub stars) yet we can also agree that the Linux Project is a much more vital repository to keep public infrastructure availble for than say, someone's vibe-coded to-do app. Is it impossible then to make any quantitative distinction between these two? We can't have both an uncontrollable firehose of AI slop and unlimited free public storage/compute at the ready for it. I say we actually need to decide what software is worthy of using up these resources for free. We already have companies like JetBrains[2] using dynamic pricing for their products. It's time Github do the same. If you want a hosted VCS for vibe coded projects that aren't being used or depended upon widely, then host the repo yourself.
Im shocked this is still going on... We moved our repos at work to a self-hosted Forgejo 2 months ago because the regular outages were getting ridiculous, looks like a pretty good call now.
Honestly I dont think I've seen a tool I used regularly with such a moat lose it simply because they cant keep the service up. Also its hard to not see a pretty strong correlation between a bunch of these big companies doubling down on AI and just having their service go to s%#$. We've had similar issues with Digital Ocean recently, which is pretty lined up with them adding a bunch of "inference" services and rebranding the site to add all the "agent" marketing slop. Looking to move to Hertzner when the time allows.
Have there ever been a comprehensive account from a GitHub insider on why their reliability tanked so hard since the MSFT acquisition?
I'm genuinely curious what changed, what their processes are and how they internally think about their reputation being in the gutter with all these incidents.
but i always guessed that maybe half of us seeing how many tokens we can spend for a year straight maybe stressed tested their platform for a year straight
Among with assorted thoughts on AI or new model point release announcements.
We're a good year+ into the use LLMs for all major bits of software that we all rely upon and GitHub here is down to one 9 of uptime. I've been using GitHub for a _long_ time, my first commits there go back to August 2009!, and I honestly don't recall GitHub going down as much as it has in the last year.
I'm sure there's other things happening in the background, but I can not help but believe that this is directly correlated with the increase of LLM usage.
Though I would love to hear someone else's pet theory how a rock of the internet went from four+ nines of uptime to maybe one.
As someone who has spent many years working in high load environments, this is not an uncommon pattern.
You design a system and it works great. It can handle failures, load spikes, it is horizontally scalable, things are great. You think you figured it out.
And then load keeps increasing and you suddenly hit a tipping point where everything keeps failing, and you cant keep up. The things that you thought were perfectly horizontally scalable turn out to have a bottleneck you didn’t even think about until you got to a truly massive scale. Your systems suddenly don’t have the excess capacity to handle load spikes or catchup work, so suddenly any failure cascades and recovery is more and more difficult. You can’t solve the problem with additional hardware, and your perfect scalable design actually can’t scale any more.
This doesn’t have to be about GitHub using LLMs in their code to still be related to LLMs. GitHub gets a lot more commits now because of LLMs and probably get a lot more reads because of LLMs as well.
It could be that the extra usage just pushed them past one of those capacity thresholds.
If I had to guess it's because Github is sitting on top on infrastructure held up by toothpicks and duct tape
GitHub Will Prioritize Migrating to Azure Over Feature Development - https://news.ycombinator.com/item?id=45517173 - October 2025 (63 comments)
Web search, steaming, high frequency trading, and many other systems are resource demanding, and keep scaling. Whether to handle the surge due to bots or wider adoption. But GitHub can't scale git? It isn't even git failing.
Incompetence in leadership is what makes a tech business technically unable to meet growing demand.
I don't think this is true. Take LLMs for example, you need GPUs to serve them and these are in short supply at the moment making it hard to meet demand. I don't think this is necessarily the fault of incompetent leadership.
Probably both. But it still throws a thorn into the theory that LLMs are about to replace software engineers any day now.
You'd think they could LLM-code their way out of this situation easily if LLMs were the software engineer replacement they are being marketed as.
If you want to have a horrible time, try reading some of the runner code. It's early 2010s-style Windows-First, MSFT C# crud that has trouble not racing several threads to inconclusive status codes.
Seems pretty conclusive. Very similar story when they bought skype.
I have an Ops background and I strongly suspect they were given a stupid timeline for the Azure migration.
I've got to believe Microsoft have decent Ops people but the management wanted to move faster than was reasonable and screwed it up. Move one thing at a time and double check it all works and you can do a migration like this.
I'm sorry but this made me laugh out loud. That isn't true at all, this has been going on for years. This conversation[0] from six years ago has discussion about the outages starting to become much more frequent in December 2019. It has never gotten better in that time, it's just continually degraded.
[0] https://news.ycombinator.com/item?id=22935941
Most people work around it by self-hosting github (which has other problems but uptime aint one).
I'm really kind of surprised they let us do that - like, why didn't they just have you upload the binaries after building on your local machine?
You can do that already with GH Releases. Actions is if you want CI/CD managed by GitHub. And you can also use your own machines via self-hosted runners.
With what to show for it? If GH did 10x in volume/git commits, it's all LLM sloppy-pasta. Where's the 10x productivity? Where are the amazing apps?
Though I'm sure some of the blame can go to internal slop code.
If your platform can't handle the use patterns of AI, then perhaps don't go around telling everyone to use AI for everything? It's a self-inflicted wound, you could also just not do this. Too bad Microsoft has bet its future on AI not being a giant bubble, huh?
That plus migrating clouds is insanely difficult to manage. They're almost certainly drowning in traffic and trying to keep up.
> Yup, platform activity is surging. There were 1 billion commits in 2025. Now, it's 275 million per week, on pace for 14 billion this year if growth remains linear (spoiler: it won't.) GitHub Actions has grown from 500M minutes/week in 2023 to 1B minutes/week in 2025, and now 2.1B minutes so far this week. So we're pushing incredibly hard on more CPUs, scaling services, and strengthening GitHub’s core features. And as a fine purveyor of hand-crafted shit code for many years, I'm not gonna weigh in on that.
x.com/kdaigle/status/2040164759836778878
These are folks that routinely make it a point to press on system design and scalability during interviews.
Now they suddenly can’t scale or design systems but we should accept that?
Even my own systems at home are burgeoning under the load of the my more ambitious hobby projects I'm doing for fun. I don't envy "we had to build five more datacenters to keep up with demand" class problems just from a how many people have to sign off on them perspective let alone the technical difficulty of doing so
For like 10 years the only feature development they did was by stealing ideas from GitLab. I wouldn't be shocked if little if none engineering discipline has took place at all during this time if it's this brittle to frequent change.
Guessing it was mostly held together with duct tape and poorly written tests/monitoring systems if any other corporate driven software.
We're really going to find out over the next few years which businesses have good practices or not.
Why do we keep giving excuses to poor engineering disciplines + poor management? This problem is entirely GitHub's making and acting like it's some unplanned natural disaster is low key pathetic.
selfhosted runners do not go down when our CI is down, its operates separately.
This is no surprise given standard Microsoft operating procedure - https://news.ycombinator.com/item?id=47616242
I'm certainly not surprised that they tried to sneak in billing self-hosted runner minutes at some ratio to potentially recoup the costs here.
0: https://docs.github.com/en/enterprise-server@3.21/admin/mana...
Edit: to be clear, I mean the part where they maintain state on their end. I have no idea if they do that with another runner instance, that seems unlikely.
Their MO was to court an executive and sell second-rate tools to them before the people who had to use them had a chance to say anything. It doesn't matter how much evidence you can provide to the contrary, once the million dollar deal is signed, you are going to be tasked with finding reasons to say that your executive was shrewd for buying this pile of junk and unfulfilled promises, and not an insane idiot sucking away your job satisfaction as fast as they can.
They did a lot of deals based on how their products would have features their competitors already have 'soon' when they haven't even started them, and a long track record of taking 3 major releases to get from something to good, and then breaking everything again by doing a 4th major release that re-imagined everything and made it horrible again.
I'm not going to claim that Apple was or is a panacea. Apple doesn't use vaporware which is big, and their Cycle of Awful is 2 releases instead of 3. You could afford to skip 1 waiting for the next even-numbered version, instead of being 2 versions behind and getting pressed to upgrade.
Windows NT was one of them. Up to Windows XP the products were pretty solid and each had visible improvement against the previous one.
Their language products were/are still solid IMO. Maybe Visual Studio is sluggish, but we can still use an older version if we want. Plus they put a lot of effort optimizing VSCode, too.
Even back in the MS-DOS/16-bit Windows days, when things broke down quite easily, I think they still provide the best bang for individual users and developers. There was no competitors who could provide so much value back then.
Disagree. I was admining NT4.0 boxes back then and it was
1. very slow
2. constantly leaking that required weekly scheduled reboots
3. security wise it was nightmare even by that times standard
For the last point what was the golden standard back in the mid-late 90s, if we don't include mainframe/minicomputers? Was it Solaris or BSD?
Now that I think about it, I feel sad that I do not have the technical prowess to compare operating systems :/
Windows 95 was a brilliant hack that allowed 32-bit GUI programs to run on hardware with 2Mb of RAM (Windows had 4Mb of RAM as the official minimum, but you could run it on 2Mb) while preserving compatibility with the majority of DOS software.
If the scheduling was self hosted it would be inexcusable but you can always just connect whatever you want to webhooks.
I don't disagree that it's obvious they've got problems but I'm just saying it's obvious to me the part that falls over (the scheduling of jobs) and why that would impact self hosted runners, which do no scheduling but depend on it to function.
As for 'just a message queue with some database updates and sharding that's easy to reason about'... Here's a job scheduling problem as an example: imagine you schedule a job, and there's no runner available. How do you disambiguate between no runners available because you've reached capacity, runners not being available because they're on a real network with faulty connections, and runners not being available because of a faulty rollout of internal updates?
A simple message queue for job scheduling is fine if you own everything and can deal with the operational overhead of identifying those cases by hand, but Github can't do that.
Sure, but the part that actually schedules where a 'job' gets run is based on a relatively simplistic tag system. Reading the yaml and plopping some job metadata into a queue-like system isn't where I would expect their issues to be, but at their scale I'm sure everything becomes fragile and inscrutable.
> imagine you schedule a job, and there's no runner available. How do you disambiguate between no runners available because you've reached capacity, runners not being available because they're on a real network with faulty connections, and runners not being available because of a faulty rollout of internal updates?
You don't need to. GitHub Actions runners, and most CI runners that I've interacted with appear to have a pull-based model where they ask for work that matches their declared tags/shape (usually platform/runtime/OS/etc.). This probably amounts to a database query, but who knows.
> A simple message queue for job scheduling is fine if you own everything and can deal with the operational overhead of identifying those cases by hand, but Github can't do that.
I highly doubt it's a simple message queue. My issue is git repos and their CI infrastructure have very low coupling to other repos or entities in most circumstances, at least conceptually, so parts of the system (ie. regions, shards, etc.) should be able to function even when others are down (ie. it shouldn't break for everyone). There's clearly centralization and coupling that isn't obvious from an outside perspective, which sorta tells me it's incidental, but that's a guess.
They have a strong motivation (self preservation) to continue to misunderstand the problem. If they did what is best for us, then we could avoid a substantial fraction of all GH subscriptions by using a FOSS tool to hit the Pareto frontier by replicating just enough GH services to watch commits and PRs.
Sitting on Github these days is the same as sticking to twitter a decade ago, expect next mecha hitler, I suppose.
9% availability would be an uptime of ~33 days a year, I think at that point, we're pushing the semantics of "available" if the service is down the entire year except one month on average.
I thought they should rename to Anglia Busways and have bus replacement trains instead.
I have sympathy for the on-call team trying to resolve it, most of us have been there done that.
But seems something is systematically going wrong at GH
Yes, we call it: Microslop.
Outages happen, but this many outages so close together, and so many of them so major/long lasting, something is systematically wrong for sure. It's been seriously hamstringing our ability to ship code at my company.
What is happening at GH?
Rate of change trying to keep up with new challengers? Over-reliance on AI? Engineers trying to debug slop?
They're at 93.91% uptime over the past 90 days, according to https://mrshu.github.io/github-statuses/ , and that doesn't even include today's outage yet.
A glorious one nine of reliability.
In large part, the move from AWS to Azure. Azure's just bad.
Source: me
This is annoying and I'm here because it's down. But it would have to be far worse to come close to actually being worth changing.
If I could right now:
1. go sign-up elsewhere 2. Log into GitHub and point Actions to that new host 3. All my actions files immediately worked without question
I'd probably give it a spin and make a wiki page explaining how to swap back and forth. No meetings. No design issues. No scheduling. Just a flip switch on who to pay for computers.
I think people were so excited to move away from jenkins to something 'managed' just because of how much a dinosaur jenkins is and how much a pain in the ass it is to upgrade it... but now we are seeing how managed can bite you in the ass if the manager is incompetent.
But for people who either don't pay anything at all or phenomenal amount one 9 of up time is all you need.
If you're actually trying to run a business I guess you can call and gitlab and get an Enterprise contract
[0] https://depot.dev
This is a man who's spent a significant portion of every day for the last 15 years on GitHub.
>My personal projects and other work will remain on GitHub for now.
People always talk about "leaving GitHub" but it's all talk.
This is multiple times this month that this has been a problem.
Has GitHub completed it's internal migration to Azure yet? Or is it still ongoing? None of our devs want to switch away from GH, but we will have to at this point.
Step one - migrate my build workflows to Docker.
My Github actions are now basically: "checkout / set env vars from secrets / docker-compose builder run make".
I used large machine runners to run full the Docker (Podman actually) on Github first to avoid dealing with docker-in-docker complications. This step also provided some very nice robustness advantages, as I can now trigger deployments from my laptop if needed.
Step two:
Migrate to self-hosted runners. I used my former homelab server to set up a build machine. It has 16Tb of fast NVMe SSDs and thanks to Podman container layer caching, my entire lint workflow now takes 30 seconds. Faster than just one "npm install" on Github before.
And Github's self-hosted runners are actually surprisingly easy to set up and use. They are also somewhat more robust.
Step three:
Swap Github for something else.
Nobody cares about ATProto or whether your commits are a damn NFT or some bs just literally improve upon the experience.
That’s it.
It’s as if no company is focusing on the product experience or anybody’s experience anymore. It’s all ooo look what I got I got this I can do that too me me me but nobody will ever buy that.
Say what you want about huge companies like Microsoft or Walmart but they spend a lot of energy understanding the human experience to sell products and less on their own perceived self-aggrandizement.
GitHub is the best version control online and it’s not even close.
Also GitHub just sucks.
Github is just the laziest default. It's not _terrible_, but it's also not great.
Now I'm stuck twiddling my thumbs with PR checks stuck/failing...
We have many agents per employee working in parallel pushing way more commits than was humanly possible before AI, triggering GitHub actions a lot more than the workflows were built for, causing Actions costs to escalate (they really aren't cheap if you compare to hosting it yourself), meanwhile working with YAML workflows is just a pain, and just writing code would be so much more fun and AI compatible[1].
At the same time, GitHub has about ~3 different PR review UIs? And they're all half-bad? Any decently sized PR triggers their "optimized for large PRs" UI which jumps around randomly in my experience. If you don't get that UI and keep the scrolling one (there's an old and a new one btw) then god forbid you click a line number because at some point your browser will randomly scroll back to that line and it won't unstick. Now Linear[2] (and others) is replacing the PR review experience for the agentic era.
I'd love to see a solid AI first Git + CI + reviews.
[1] Cloudflare CI https://blog.cloudflare.com/ci-workflows/
[2] Linear PR reviews https://linear.app/changelog/2025-01-23-pull-request-reviews
https://huggingface.co/blog/github-ci-hf-jobs
As a side note, I started getting these symptoms (actions staying queued forever or not running at all) at 5pm PDT yesterday, intermittently. And definitely full outage by 8:30pm PDT. So for sure full outage for 7 hours and counting, even if you use other runners, and I strongly believe partial outage for ~15.5+ hrs before that, even if it hasn't been acknowledged by GitHub yet.
I'd love to know what the most common root causes for these outages are.
I'm not sure why this particular industry is so abysmal at making things even semi-reliable after decades of research, educated workforces, and loads of cash.
Like, everyone talks about how you can just prompt a new CRM instead of paying for Salesforce.
But in this case, I raced with Fable and Sol and had time to build a full featured, fully functional CI/CD workflow system that has most of the features of GitHub actions (not the ecosystem obviously) and costs less than the GitHub runners per minute, while being able to scale to millions of workflows... And I had it working end to end before the outage was resolved and was already running my apps deploys through it.
It would take a little longer to do in a company with red tape and more investments in the GitHub ecosystem, and of course it only replaced actions. But it works.
There is no moat.
Aug 06, 2026 - 16:27 UTC - Update - Pages is experiencing degraded performance. We are continuing to investigate.
Aug 06, 2026 - 16:19 UTC - Update - Pages is operating normally.
I should eat lunch.
If the frontier AI used GitLab as a default I'm sure they'd be the ones suffering now.
Especially troublesome in the middle of trying to fix a high score security vulnerability when the release vehicle is Github.
I think we need to ask for some money back, this is BONKERS.
Any more of this, and we'll go full Gitlab, like we should have years ago!
Also, doesn't even have RAG offering.
Seems like the only reliable way to run GHA jobs is to not use their runners. Hope they at least didn’t break self-hosted runners operations
> Customers using self-hosted runners may see errors or rate limiting when runners register.
Maybe they're hosting in us-east-1 though :)
Self-hosting a service like GitHub that operates at GitHub scale is difficult.
Self-hosting a service like GitHub that operates at the typical small/medium company's scale is trivial.
A single machine (with separate runners for CI) will cover many companies' needs. It being a single machine eliminates a lot of the complexity and failure modes associated with a distributed system and makes backups/restores/maintenance easy.
Even self-hosted runners are impacted.... How can that be?
The cost of this globally has got to be in the hundreds of millions to companies that use CI/CD through GitHub Actions. What if prod is broken and GitHub actions is stalling the deployment of your hotfix? What if this makes your organization miss and SLA and diminish user trust? What if this makes you miss a release that you were contractually obligated to meet? This is happening during peak dev hours on a Thursday (not that it would be acceptable at any other time).
I don't understand how a service this critical to the global technical infrastructure can fail like this at all, let alone for more than a few hours. Like where's the backup generator for crises like these? You can't even use self-hosted runners? WTF? Like how can you not bring your own backup in a crisis event like this?
Not that Microsoft has a good reputation, but holy moly, you'd think they would prepare from something inevitable like this.
It's hard to draw a direct analogy there, but I feel like it echoes the same sentiment.
A soapbox I have is that GHA workflows are scripts that could run on your machine without any of the YAML stuff. Who gives a flying about the DAG or the logs? Which, by the way, if you're willing to walk to the milk store to buy your milk, could be recreated in a much more testable and maintainable way without any of the YAML bs that GHA prescribes....
But DAGs are pretty, and logstreams showing up in a browser application instill trust (for reasons that fly far above the head of yours truly). So people go for that. Pretty DAG, nice logstream; therefore, deliver my milk. All of a sudden.... The CI/CD platform is having its merry way with your SLAs, contract abidements, and hotfix deployments.
What a time to be alive.
Not to rail on the South Park thing, but the blast radius of this issue also reminds me of the episode where the internet dried up.
If this bs with GitHub continues, Parker/Stone will have to make a GitHub episode. How seen would we all feel if that happened?
The fix is merged, but won't deploy... it's been hours
Thankfully it's a batch job, and isn't interrupting production ATM
There's always the escape hatch of running you GHA workflows locally, but unfortunately, despite the existence of packages like `act`, there is no way to fully recreate the GHA runtime locally. Tons of the special YAML syntax just can't (more accurately, "just doesn't") get interpreted by those local actions runners.
We never went this route, but at my old org, I always advocated for considering GHA to be wrapper around a single bash script (or whatever script you want to run), as a means of completely breaking out of the GHA hellscape that is programming in YAML, who's turing-completeness is pretty dubious.
Unless you have things set up this way, you (the client of GitHub) would have to completely redesign your CI on the fly, run it locally, and then figure out how to get the D compliment of the I to work in a way that is auditable. Fat chance for most teams I bet.
Thank god you're dealing with a batch scenario. Silver lining for sure. Still, embrace the anger.
What makes my blood boil is that there's millions of DEVs literally crying at the moment worrying about how GitHub's failure to be responsible will put their jobs in jeopardy.
And fingers crossed for you my friend. We're at 5+ hours at the time of this writing.... You're batch job may still have a chance!!!
latest server i set up is simply a bare repo + hooks to make a local deploy after running tests and shit
super easy to set up having ai do it, zero dependencies, deploy is still 'push it to the main'
i have several remotes for backups and stuff
I was complaining about how I should have used AI instead of manual brainwork for this instead but turns out that might have been the problem.
It was before it became a news and a trend in X.
Really frustrating experience.
Or has Microsoft made sure (through Windows licensing terms and pricing) that it's not possible to compete with their own CI offering?
Edit: CircleCI seems to offer 750 minutes/month (whereas GitHub offers 1000 minutes/month).
It seems to manifest in many places: AI slop, Github outages, recently Miscoroft sent me an email demanding that I subscribe to 365, in order for MS Office (which I already paid for) to continue to work. I simply moved on to the other, free, provider.
Perhaps, it's time to evaluate the use of the Microsoft products?
If you implemented a tool and it worked one time for your presentation to management that's all that mattered.
The actual company employees using it downstream in prod basically had to constantly QA the alpha software they were forced to use and the authors of the tool were hard to track down if they even still worked there. And if you did find the author or the team they would be very resistant to admitting there was an issue because it LOOKED bad.
So many tools I used were fragile and buggy, it was clear the authors just presented the happy path to management to get the note added to their promotion packet and the rest of the company just had to deal with the fallout.
My team implemented this product that the entire company used that was broken and buggy as hell but they kept presenting the product to management as this amazing product and nothing was ever done about how broken it was. One of my team members came from Apple and said Apple's tool to do the same thing was much better. The tool my team worked on was a well known pain point amongst the rank and file but management was very detached from the rank and file, which I guess ultimately was the primary problem.
If Github is having the same issues I feel for them.
Once this goes in, I'd expect to see 89%, which is zero nines. (I'd like to say, "a new low!", but sadly we've had this before)
Is the AI slop that bad? Culture change?
From the GitHub COO on April 3rd:
point clanker to forge.smol. ai/llms.txt
for now its just a fast agent native git remote and u can check docs for the extras.
There is no better time to self-host.
[0] https://news.ycombinator.com/item?id=22868406
Now that the calculus has changed, I think we should start asking ourselves whether anyone has the right to unlimited free public repositories.
We can all agree that trying to attribute any single metric to repo quality is subject to Goodhart's Law[1] (e.g. what happened to GitHub stars) yet we can also agree that the Linux Project is a much more vital repository to keep public infrastructure availble for than say, someone's vibe-coded to-do app. Is it impossible then to make any quantitative distinction between these two? We can't have both an uncontrollable firehose of AI slop and unlimited free public storage/compute at the ready for it. I say we actually need to decide what software is worthy of using up these resources for free. We already have companies like JetBrains[2] using dynamic pricing for their products. It's time Github do the same. If you want a hosted VCS for vibe coded projects that aren't being used or depended upon widely, then host the repo yourself.
[1]: https://en.wikipedia.org/wiki/Goodhart%27s_law
[2]: https://www.jetbrains.com/community/opensource/
Honestly I dont think I've seen a tool I used regularly with such a moat lose it simply because they cant keep the service up. Also its hard to not see a pretty strong correlation between a bunch of these big companies doubling down on AI and just having their service go to s%#$. We've had similar issues with Digital Ocean recently, which is pretty lined up with them adding a bunch of "inference" services and rebranding the site to add all the "agent" marketing slop. Looking to move to Hertzner when the time allows.
After 6 years of this nonsense of "centralizing everything on GitHub", it is not a good idea at all.
You might as well self host like I said before [0].
[0] https://news.ycombinator.com/item?id=22868406
I'm genuinely curious what changed, what their processes are and how they internally think about their reputation being in the gutter with all these incidents.
but i always guessed that maybe half of us seeing how many tokens we can spend for a year straight maybe stressed tested their platform for a year straight