58 comments

[ 2.4 ms ] story [ 89.6 ms ] thread
git push harder?
`git push --force --harder --try-really-hard` works for me.
It ran me crazy that my push wasn't working...
(comment deleted)
Alternatively you could configure an internal git server with gitweb or something. I created a screencast about git and gitolite for user authentication @ http://sysadmincasts.com/episodes/11-internal-git-server-wit...
I believe the bigger issue is when you have several other 3rd party services relying only on github like issue trackers and continuous integration/deploy... Then when this happens your workflow gets really messed up
Like Homebrew!
yap! and npm, bower, rubygems, etc...
I didn't think rubygems relies on github. And rubygems seems to be working fine to me at present.

It is true that rubygems runs on AWS and is effected by AWS outages (like much of the internet, it seems like).

It's also true that individual application Gemfiles can choose to express dependencies to gems hosted in a git repo, such as github, instead of the rubygems repository. If you don't want to be dependent on github, then don't do that. Of course, you're dependent on the rubygems repo ordinarily still, which has gone down occasionally. But the rubygems repository is independent from github.

(comment deleted)
(comment deleted)
It's no problem, Hacker News is up so I wasn't using it anyway.
Interesting how hacker news became the new replacement for http://status.<put_any_service_here>.com
It really is, I ended up checking here straight away, despite there being a link directly to the status page on the error pages of github.
It’s more karma whoring than anything else, this post doesn’t add anything except that GitHub is down and people who care about GitHub already know it, and those who don’t care, well, don’t care about this post.
I just noticed that a repository wasn't working, but hadn't put it together that the entire site might be down... so the karma whoring was useful for me.
Why don't they just host GitHub wherever the status site is hosted?
"Why don't they just make the airplane out of the same material that the black box is made from?"
(comment deleted)
(comment deleted)
Why doesn't the president ride on the dummy car?
ah homebrew its not you
Those graphs are kinda useless - detailed time series information for what's pretty clearly reduce to a 0/1 state for outages.
Well some problems are slow downs instead of outright outages, and thats what the graphs are for since "response time" doesn't make much sense when no response is being sent at all.
Run it out to "month" and you can see cyclical, daily variation of response times and a general improvement from a slower than current state over the first week.

Tangential graph criticism:

If you aren't going to put "zero" at the bottom of the Y axis, and you have graphs with only a single numbered data tick, then I have no idea whatsoever the value that correlates to a Y axis position. (except when it hits your label.) If you have fewer than 3 labeled data ticks then I don't even know if you are linear.

Starting your data at zero is a really good idea unless you really mean to be graphing the derivative, in which case, maybe you should just graph the derivative. The exception is when you have a constant offset in the concept you are graphing. Like atmospheric pressure on earth. But please give me at least three ticks and try to leave them consistently spaced if I'm every going to see another graph of them from you.

Including unlabeled ticks, evenly spaced, is a terse visual language that you are a linear graph.

Half of package managers are down (can not retrieve package) as well. Mirrors, please?
It's back up.
(comment deleted)
Status page says down for me, but site is up for me.
and..... it's back down. nvm
Clearly, we need a "Hacker News is Down" Gist so we can chat about Hacker News being down when it goes down.
Anyone know what the difference is between status.github.com and github.com/status? /status just says "Github Lives!" with a timestamp.
/status is probably what status. polls to determine if it's up. One is on the main server and the other on a dedicated, off-site one.
(comment deleted)
If you have external service dependancies on your deployment process, you're gonna have a bad time.
git in not centralised so get over it. no need to report every github issue here.
(comment deleted)
> We've recovered from an internal DNS outage, and are working to restore service to all repositories.

status.github.com

DNS, eh? Related to the recent CDN changes?

> internal DNS

Not external, which would fall into the CDN-managed territory.

Was using homebrew, composer & bundler... FAIL; github I need you!
So what? I thought the whole point of git was decentralization.
Isn't this the whole point of git? You have a local repo. So there's no reason to be on hackernews :)
I don't understand how anyone could outsource something as mission critical to their business as git hosting, issue tracking, etc to a third-party. If anything happened to github half the developer community would be hosed. Let's hope that never happens.
The same reason you outsource car maintenance to someone else: Because when there's a problem, there's a team of experts available to work on the problem, whom have far greater expertise to resolve it.
Except that, as far as git hosting goes, every developer has a complete commit history of all branches on their local machine.
GitHub has a service that is meant to be highly available. What are the chances that 1/4th of the engineering effort would be placed in to a local solution?

Downtime is inevitable, and sometimes the thing that fails will cause huge problems and it's almost entirely unavoidable.

I'd much rather trust their team, which they've built to design the best service possible, to handle the infrastructure than having to use an internal team to design a competing product (albeit internal) with the same uptime requirements.