GitHub Actions down again today(githubstatus.com) |
GitHub Actions down again today(githubstatus.com) |
This is why we don't use Github Actions, kids.
Seriously, its a proprietary build service that puts the keys to the kingdom in someone elses' control. Just: No!
Print this status page to PDF so you've got it handy next time someone castigates you for not using Github Actions, folks.
This time today it was caused by friendly fire by the automatic suspension of the GitHub Actions bot which is now a "Ghost" user. Since there is no CEO of GitHub to contact it we are just going to see more [1] of this again.
You might need to push a critical change soon, but now you cannot. You won't get any of these issues if you self hosted as I said 6 years ago...[2]
[0] https://www.githubstatus.com/incidents/g6ffrm0rfvz9
I'm guessing related to this? The blog post is dated 11 days ago but I just noticed a blue banner on my actions page today.
Thanks for pointing out that nobody is using that thing
Something’s wrong when my own infrastructure is more reliable than Microsoft’s.
- GitHub
- Hiring budgets
- RAM (/personal computing in general)
- Electricity
- Media/Content
- Truth
I like being able to vote with my (teams) wallet and I'm tired of staying out of convenience
Perfect timing that we post https://www.jxd.dev/writing/building-plain just as this latest incident started.
Or maybe it's before the GitHub internal devs are online and deploying changes.
The latest language models have enabled this sort of thing for me. I can integrate a mini Jenkins into every project within a 5-10 minute prompting session. This sort of code isn't hard. It's just tedious, and the LLMs absolutely rock at boring repetitive stuff. Having a win32 service start up successfully on the very first try is something I haven't experienced until 2026.
Which certainly made me shit myself, briefly.
No, it's not. Official updates = potential SLA penalties. Always requires approval.
I vibe coded a script that interacts with both Gitlab and Github via their APIs and I've been using it pretty heavily since this morning. I crossed the streams! Goodness, I didn't know it would be _this_ bad!
- So many super-heroes/super-villains
... You're off the hook this time./s
For Git, all you technically need is ssh access and some backup strategy for your server. It would be bare bones but workable. And there are of course plenty of OSS things that are a lot nicer than that.
I'm still using gh and gh actions and we are mostly below the freemium layer with that. But it is kind of slow and honestly a dedicated vm plus some high CPU/memory workers we can spin up on a need to have basis might be a lot faster. With GH outages becoming more common, my hand might be forced a bit.
In recent weeks, I've spun up listmonk (mailing list solution), matrix (as a slack alternative), and a few other things specific to our software stack. A github alternative would be more of the same. We don't need a lot.
The main objection is that with more moving parts to worry about, the workload for me also increases. Things need updating, monitoring, backups, alerting (and responding to alerts), etc. That sucks up my time and that is scarce.
Another reason for self hosting these days is that with agentic AI tools, self hosted things are a lot easier to integrate into agentic systems. If it is self hosted, you don't have to worry about API limitations, rate limitations, walled gardens, etc. All the traditional SAAS silos are becoming a problem from that point of view. The more locked down it is, the bigger the motive for moving away from it. That's why we ditched Slack for Matrix. Slack is hopelessly locked down and tedious to deal with. Matrix is super easy for this.
Technically Dropbox is just rsync.
Also https://xkcd.com/1319/ but for maintenance.
(Ofc, in a sensible universe, we just brush that off to a JS/Firefox glitch or my ISP.)
And yet, here I am. My code is not compiling, my AI isn't vibing, nonetheless I can't work! Two more hours before I can get off!
Anyway. Forgejo's response to it: https://floss.social/@forgejo/116494295922963052
No, it's not like "act," because it uses the standard Github runner, the difference is that the control plane is an emulation of api.github.com, because of this we can do all kinds of nice things:
Caching in ~0 ms. Pause on failure, so you can let your AI agent fix it and retry without pushing.
Is what it boils down to.
> codex "Fix this pipeline, use `act` to verify your changes"
I have tried to use act many times, and many times I've failed.
P.S. pause on failure is also helpful for humans, but I'm trying to be realistic about where the future of programming is going...
I started playing with proxmox VMs and containers in them (docker and tart) to see if I can build some local infrastructure to properly solve this…
The jobs runs via containers.
I much prefer Woodpecker CI, which is an open source fork of Drone.io. It supports multiple Git backends like GitHub, Gitea, Forgejo, Gitlab, Bitbucket. It supports running jobs locally, on Docker, and on Kubernetes. And there's autoscalers built in for AWS, Hetzner, Linode, Vultr, and Scaleway. There's a bunch of 3rd party plugins (https://woodpecker-ci.org/plugins) for custom integrations. The UX is also very simple, with OAuth used not only for authentication/authorization but also setting up & accessing repos. The system architecture is great, with separate components that run stateless connected to a database, and a custom plugin is any program that takes environment variables and does stdio. The config file is a good balance of ugly YAML and convenience syntax like shell-style parameter expansion variables.
It probably takes less than 15 minutes to install, set up, and run WoodpeckerCI for a small team, so it's not a big investment to try out or host. With the autoscaling plugins it lets you scale your workload up to whatever size. Honestly you could run it on a laptop since it's written Go.
(to clarify for beginners: the config file docs are found in a section called "workflow syntax" (https://woodpecker-ci.org/docs/usage/workflow-syntax) and variable parameter expansion is buried deep in an environment variables page called "string operations" (https://woodpecker-ci.org/docs/usage/environment#string-oper...). poorly organized docs aside, the system itself works well)
"Microsoft’s GitHub was positioned to win the AI coding race. Outages got in the way" - https://www.cnbc.com/2026/05/22/microsoft-was-positioned-to-...
Even though it's selfhosted and we don't have a dedicated infrastructure team, I don't remember it ever being down in the last 12 years I have been working here.
Been burned too many times on that one.
Move to EC2.
Darn AWS is down.
Alright, run it on a Mac Mini in your basement. Ahh dawn, your ISP is having issues. Good thing you have a backup 5G hotspot.
Ohh no, the power is out.
Eventually you have to trust someone else.
GitHub is a tragedy of the Commons. Too many people are using it, and Microsoft isn't willing to handle it correctly.
Feels like a very good business opportunity. Minimum 50k yearly contracts, GitHub with actual uptime. GitPro ?
We can't be blocked here. Seems silly what we settled on this, but for a long time GitHub had been reliable enough for many years, but things are sliding down the pan as of late.
On my repo the jobs do not get scheduled on the PRs at all, so I assume that separation wouldn't help for todays issue.
Wait until you charge you for self-hosting runners.
Oh wait. They already tried.
You can now hire me as an overpriced consultant instead of paying Microsoft.
I don't think vibecoding at Github has much to do with it.
That makes sense. Thank you!
That being said there was a noticeable trend starting around 2022.[2] That being said they’ve also been doing a big migration to Azure. It’s likely a combination of things.
1: https://www.cnbc.com/2025/04/29/satya-nadella-says-as-much-a...
I used to use Cirrus CI as an alternative to GitHub Actions and am looking for a new alternative. I wonder if Depot could fit in the same way for my needs. I need to run builds and tests in Windows, Linux and macOS.
Hope you don't mind the public ask, it seems useful for others.
If we're using depot runners, and want to use them directly, or move off of github actions being the controller for when things run: what do you suggest?
Trigger the workflows directly on depot via CLI?
We're now considering Buildkite (apparently they have a GH actions migration tool) or self hosting something (GitLab CI, maybe even Jenkins), as it looks like that would've kept ticking over since we're still seeing webhooks being triggered today during the downtime.
https://www.blacksmith.sh/ and https://runs-on.com/
They also say that they're much cheaper than github
I don’t buy the excuse. I want to hitch my wagon to those “mysteriously lucky” competitors. (And have. And haven’t had similar issues to Github, since.)
We’d need more details around what you’re seeing. It is true that if auth across GitHub is broken than we can’t copy your actions out to be used by Depot CI. However, we have a solution in the works for that as well.
In short, Depot CI, our own engine and control plane is not dependent on upstream actions control plane. But still has to listen for commit events to know if/when to run jobs on things like PRs. This to is being removed in the future.
I like that it exists, but what a freaking mess that it's necessary and so difficult to do.
Is it true that official service status pages are updated automatically?
If the first they hear of an outage is when user requests start to fail, then that's a failure in their monitoring as well.
But effective monitoring is harder than people assume.
Isn't that what monitoring actually is? The issue seems to be in their testing, not monitoring.
There are synthetic tests, where you can generate API request calls or even simulate an entire user journey. These allow you to control the user agent, the payloads, and thus you know anything errors back are actual errors. These are triggered by the observability platform (think like running a cron-job) and thus you're not tied to user activity to see when problems arise.
There are other metrics outside of HTTP response codes too. Think like free RAM, CPU usage, disk space, etc. This is just naming some obvious ones because these types of metrics are generally bespoke to the type of application your monitoring. And with these types of monitors, you'd not just have an alert when things have failed, but ideally have alerts when an irregular trend is showing that things are likely to fail too. This latter type of monitors helps you get ahead of the problem before it become customer facing.
Then you have more traditional stuff like logs. This will also be bespoke to the application. But you'd expect errors in logs to get surfaced quickly. Assuming Github have good hygiene in what's being logged.
Tie that up with APMs, RUM, and other goodies like that and you'll have diagnostics to investigate issues when they appear.
(this is just a super high level view of observability too)
Maybe the Github Actions infrastructure isn't run like that.
edit: my oncall rotation notified on all 500s, 24/7, not just rates - https://news.ycombinator.com/item?id=48279262
If my DB health check endpoint is returning 500s for N consecutive checks over M minutes, yeah, please wake me up at 3am!
If one user hit a weird edge case in form validation and got a one-off 500, please don't! We can fix that on Monday.
Not always easy to distinguish those clearly or configure those business hours rules, but for my team at https://heyoncall.com/ that is the goal -- otherwise your team burns out fast. Waking up someone at 3am has a real cost, so you better be sure it's worth it.
Recently there was this: https://news.ycombinator.com/item?id=47252971 "10% of Firefox crashes are caused by bitflips"
Which makes me think a small amount of random issues which happen even though nothing is broken, is normal everywhere. Especially once move things around on a network, there's potential for a lot more random errors.
It does require constant tuning and adjustment though.
This is why data hoarders who have NASes with lots of space insist on running their servers with ECC RAM despite it being significantly more expensive. Because bit flips, for all intents and purposes, cannot happen. The RAM itself detects and corrects for them.
I wouldn't expect bit flips to be a significant contributor to enterprise problems.
I'm sure you're not in ops. Or in a dev org of a service with decent request rates.
What you're asking for is a service to fail silently. There's no way a service with a decent request rate to have 0 500s. Not when it still sees development.
A 50 year old bank API? Maybe...
I know all of Gmail, every GCE service I can think of, every AWS service I can think of, Amazon.com, Netflix, and Github all do not page on just a single 500.
I know none of those are particularly "high performance" though. Curious where your experience is coming from.
I had a fairly long tenure, where I maintained multiple key services in critical online payments flow. Authentication, authorization, core business and risk data, as well as some cross-cutting control plane stuff, etc. You needed one or more of our services to take a payment, serve any request from the employee dashboard - pretty much everything hit our services. The entire company ground to a halt without my team.
We paged for every single 500. In instances where a particular class of 500 was spurious or not worth fixing, we would leave it acked or mark it as noise. But typically we'd just put in a fix as soon as possible so we didn't page.
Our graceful shutdown and traffic shaping stack was great, but occasionally we'd get a few pages during deploys or failovers.
Oncall was typically not bad, but when it did get bad it was terrible. I've been involved in huge outages that cost hundreds of millions of dollars. Usually it was the fault of multiple teams having compounding runaway failures rather than one service or bug in particular.
It's inexcusable to have a customer's payments not go through. We engineered around resilience. We had strict five nines SLAs and p99 targets and evaluated our adherence with even the smallest partial outage. Hundreds of other services depended on ours, and downstream impacts were huge, so we had to keep a tight ship.
We didn't have "business hours"-only paging either as our platform was available globally, including a heavy install base in Asia.
But if it is synthetic queries sent from the monitoring platform, then you control the user agent, payload, and endpoints. So any failed requests are a symptom of a misconfiguration and/or failure that should be investigated. Albeit not necessarily as a P1 priority.
Assuming the existence of some kind of network (with zero guarantee of 100% reliability), how does this work in practice? Is each 500 treated as an event that needs investigation, even if the result of that would end up as 'a router dropped something from an internal buffer but the transaction as a whole was re-tried by a parent so the service itself recovered'?
Even if it's "DB in datacenter I tried to save to was hit by meteor" event, you can cater for this not to result in 500 (ie - DB unreachable, retry in a couple of minutes); the question is if you want to.