I would be curious if anyone else on HN is responsible for running a GH Enterprise instance and whether or not the extra liability is worth it.
Pros: Not affected by GitHub.com downtime; stability has generally been good for us; you have full control over the org and user names on the instance (instead of having to pick globally-unique ones on github.com); can sync licenses between GHES and github.com if you want to use some private repos on github.com; can raise your API rate limit as high as you want; probably cheaper to provide your own storage.
Cons: New features from github.com can take months (or longer) to be released to GHES; many third-party integrations only work with github.com not GHES; if something goes down, it's now your fault; Actions requires bring-your-own self-hosted runners; autoscaling your self-hosted runners is non-trivial.
The last point is probably the biggest negative for us. We're trying to use GitHub Actions more, but I don't really want to have to manage a pool of autoscaling runners. (I'm doing it, though.) If GitHub provided an add-on service to connect GitHub-managed cloud runners to GHES, I'd use it. (Needing to always keep code private and on-premises is not a requirement for us.)
The sibling comment about needing to use a partner may speak to this, but unlike GitLab GHE is also not `docker run ...` and tada, so you're right to be wary of the pain
https://www.githubstatus.com/history
On one hand GitHub seems a little more proactive than most when it comes to raising incidents and updating status pages, but we’re getting a little too used to pausing our deploys in the middle of the day.
It's not too late to join the serious projects like Blender, ReactOS, RedoxOS, etc who are happily self hosting and are not going all in on GitHub.
I posit that they even have better uptime by just self hosting for one year vs GitHub's uptime record since they got acquired by Microsoft. Looks like 'centralising everything' to GitHub did not age well at all, as I have warned years ago. [2]
[0] https://news.ycombinator.com/item?id=34982456
https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...
i dislike github quite alot but this does not achieve anything. no one else is saying to go all in on github so what is this even reply to?
Because I can. How many more times does a service need to go down in less than a month to deem it unacceptable for any use? Unless you think paying for downtime is acceptable?
> i dislike github quite alot but this does not achieve anything.
That is not my problem since I saw the problem years ago and chose to self-host instead. Rather than others complain to GitHub, they might as well self-host instead for reliability.
If you can setup a simple SaaS website, you can easily self-host a Gitlab / Gitea instance for 1 to 5 people. It's not that hard.
> no one else is saying to go all in on github so what is this even reply to?
So you didn't read this then?
>> We're all in on Github actions these days but there was a time when we used to maintain our own Jenkins instance. [0]
>> At this point, our self-hosted CI and code review tooling has had significantly better uptime for the past year. [1]
The schadenfreude for GitHub's unreliability and some of its fans celebrating monthly down time only makes it hilarious to watch.
That can also be a nightmare.
Do you want git hosting, CI/CD, etc. to be among your core competencies? Because that's what this entails. You need a full team to staff it and be on call.
I guarantee you that those who self host will have just as many breakdowns. Five nines is hard, especially when it's not your main job.
https://twitter.com/AzureDevOps/with_replies
Edit: Must be something bigger than just the MS/Github hosted agents. Can't trigger builds or deploys on my self hosted agents either. Control plane issues perhaps.
Paulie 'Walnuts' Gualtieri : Get the fuck outta here.
Tony Soprano : Yeah. Nice, huh? He was with the Interior Ministry. Guy's some kind of Russian green beret. This guy can not come back to tell this story. You understand?
Paulie 'Walnuts' Gualtieri : I hear you.
[the telephone connection is lost - Tony swears, and Paulie hangs up]
Paulie 'Walnuts' Gualtieri : [turning to Christopher] You're not gonna believe this. He killed sixteen Czechoslovakians. The guy was an interior decorator.
Christopher Moltisanti : His house looked like shit.
Seen how often Azure has problems? That's the future now.
Just don't 'centralize everything to GitHub' like someone else said before since we now know that with this foresight made years ago [0], it is a very bad idea in the long term.
Just sayin... Bitbucket pipelines reliability over the last months has been looking pretty sweet compared to GH Actions...
Bitbucket: https://bitbucket.status.atlassian.com/history?page=1&filter...
We are taught early on that if something doesn't work, it is because something is wrong in our code.
Keep the f$!# away from it. I have suggested moving away from GH entirely, but my peers aren't too keen on migrating to a different CI system yet again.
Consider self-hosting everything.
