GitHub down 2016.10.34 6:06 GMT

42 points by noajshu ↗ HN
unicorn on each pageload

28 comments

[ 3.2 ms ] story [ 59.3 ms ] thread
Good thing we like to deploy at 2AM.. oh wait..
You may not realize that Central Europe is already awake and working.
(comment deleted)
I flagged this because I don't think that this is actually interesting, will drive discussion, or in general meets the site's guidelines.

I furthermore don't think upvotes/downvotes playout well with this sort of content. It's obviously low-to-zero-value, but it attracts "hey, me too" type upvotes of people saying they observed this super obvious and dumb thing too. I think that factor ends up with a very low-quality post in general.

People who are interested in github will find out by using github and seeing a very clearly identifying unicorn with complete message and status page.

This provides no value or discussion. The likely post-mortem will be interesting, but until then, nothing to see here.

I actually agree with you. Don't understand why you are being downvoted.
> I don't think that this is actually interesting, will drive discussion, or in general meets the site's guidelines.

I disagree with this. The guidelines refer to Anything that good hackers would find interesting and I certainly think a major outage at the heart of open source software is interesting in that it drives discussion around similar experiences, implications, pre-emptive action and mitigation.

The post-mortem is interesting and drives discussion. This is logically and historically true.

If you look at the other comments here, and in past threads like it, it's obvious that these threads do not.

A single service has been down for less than 30 minutes...

If it were the first time, or anything special about the outage, maybe. Not much new to be said now.

Interestingly, only the site was down, most probably just unicorn. No new workers were created, or the existing workers were busy. I was still able to do a git clone using the git@github.com:TEAM/REPO.git format.
(comment deleted)
(comment deleted)
I noticed that for 1-3 minutes after the outage began (I was between pageloads and noticed the outage immediately) the 503 page did not load after waiting >30s. Afterwards the 503 page loads in < 1s. I wonder what failover configuration Github uses, such that this delay would occur.
It's probably just the CDN caching the 503 page.
Do you mean that the CDN takes a few minutes to "kick in" and start serving that page?
(comment deleted)
(comment deleted)
34?
He possibly meant "04".
I can no longer edit the title -- is this because the post is flagged?
The title has the wrong date.
And the wrong format. ISO 8601 rules.

Or at least write out the month. That makes it unambiguous regardless of what weird date ordering is used.

(comment deleted)