@Tubsta @ironicbadger iirc they were like halfway through rolling it out when Microsoft bought them lmao
@Tubsta Yep still 0% uptime from my IPv6 only server point of view.
@ironicbadger I expect them to start measure the SLA properly after the acquisition - such a change looks rather than a change of methodology rather than immediate effect of a merge (fuck Microsoft anyway, but not for that)
@ironicbadger is that because people left, or tech investments stopped, or something like bits of housing were moved to azure?
@ironicbadger Microsoft has always had a peculiar reverse Mida's touch, transforming into junk everything they touch - remember Zune, Windows Phone, etc.
@goleztrol @ironicbadger Internet Explorer 😂
Visual SourceSafe, the only VCS/SCM capable of thrashing your whole present and past source history just opening the application.
And this is why we keep local repo copies for production repos. :)
@ironicbadger lmao yeah that checks out
@ironicbadger wait, they went from 100% uptime on their own servers,which were constantly on, to 99,5% as lowest uptime on a scalable infrastructure? That's 3,5 hours max spread over a month. And you're taking the piss on them for that? That's wild. Maybe focus on their link to Israel in the Gaza war or something.
@luceos @ironicbadger Critical services tend to measure availability (informally) in "nines". 99.999% is good enough for most purposes; run your critical system at 99.9% availability and your contract is immediately over.
3.5h downtime in a month is ridiculous. It's teenager-computer-in-a-closet availability.
@mbpaz @luceos @ironicbadger I actually didn't know that and was asking myself the same question! so thanks for that extra context
@ironicbadger interested if you can map that to growth/shrinkage in users and/or volume or data being used.
@ironicbadger searching for "github downtime 2016" I don't trust the old values
@ingoschi @ironicbadger yeah, my first thought looking at this isn't that MS ruined the uptime, it's that status monitoring improved shortly after MS bought it. I can't say whether or not that's accurate, but this graph looks...specious
@ironicbadger The inverse Midas touch
@ironicbadger @isotopp what’s the data source behind it? Official status page? Because that has been notoriously inaccurate long before microslop took over.
@ironicbadger @isotopp Thanks! So it is indeed the data from the official status page. Yes, since microslop took over reliability has gotten even worse, but it was bad before as well. They just often did not mark the services as down on the status page, even though they were.
So, the general conclusion here is still correct, but the data for the visualization is not reliable IMO.
@ironicbadger Could anyone explain to me, how is this possible?
I would imagine they would keep things running on the original hardware etc. which I wouldn't expect to fluctuate like this.
@Jourei @ironicbadger Move fast and break things!
@Jourei @ironicbadger Yeah, this is likely just a bad graph. Before vs after the change was likely a change in either the definition of, or the measuring of "down time", and the two end up not comparable.
I suspect the data went from tracking as "most/key things working=up" to "anything not working=not up"
@Jourei @ironicbadger "restructuring"
They fired huge chunks of the existing staff and terrified the rest.
@Jourei @ironicbadger No, when MSFT came in they tried to change everything at once. If anything services fee can be paid back into MSFT pipelines instead of other service providers, they wanted it yesterday. They brought in more than 1000 people, which was more than all of GitHub at the time, to dilute the original teams.
@Jourei AFAIK they started to use Azure more and more. That would explain the initial stability after the buy event. Also they might have shifted from stability to ship feature after feature. Before Microsoft bought them, their feature set was rather stable.
@ironicbadger behold the jagged teeth of the dog that eat all the dogfood!
@ironicbadger Where was this chart published?
@ironicbadger @a_different_jlh when did they really start measuring though, the source of the graph also shows data before the invention of even git itself: https://www.githubstatus.com/uptime?page=200
@murb @ironicbadger The historical lookup seems to load a page for any value in the URL, but the backwards button stops working after page 40, so it looks like the earliest data provided via that portal are from July 2016
@murb @ironicbadger There is some discussion of data validity (including the author's perspective) in this reddit thread https://old.reddit.com/r/github/comments/1rnvhs9/githubs_historic_downtime_scraped_and_plotted/
@a_different_jlh @ironicbadger ah good point. I was just modifying URL params, I'd expected a 404 with no info available.
@ironicbadger Acquire and ruin. This is the corporate way.
@ironicbadger hmm, quite a few opinions on that chart 😀 tl;dr it’s complicated.
Most of GitHub services, until I left a couple of years ago were not on Azure.
@andymckay @ironicbadger what's complicated about it? it used to be good and now it's bad
@aburka @ironicbadger ah I was replying to a different post about it being Azure or Azure management. I did Mastodon wrong, sorry.
I will say after the acquisition when GitHub became rapidly more complex as features were added, the definition of downtime requiring a status change became a lot more strict and focused (for some teams). A bit more loose and easy beforehand.
@andymckay @ironicbadger yeah I can believe there are many causes, mismanagement and forced development speed as much as technology changes
@andymckay @aburka @ironicbadger The change here is so stark that I did wonder if this actually reflected a policy change on how downtime is tracked.
@harris @aburka @ironicbadger yes that’s one of the factors.
Status is an indicator controlled by the company. As GitHub sold more into businesses and became more crucial to them eg: Actions part of deployments - uptime became more critical and monitored more.
In Actions we built lots of sensitive metrics and quickly became more and more transparent about downtime, customers noticed it.
@harris @aburka @ironicbadger status was also one metric for all of GitHub, it was split into 6 or more around the time Actions was coming out. So more granularity too.
Surprised copilot isn’t on the status page yet.
That is the least of the problems with it, got that damn AI stealing everyones hard work
Why settle for Five Nines of reliability when you can get Nine Fives?
@DaveMWilburn @ironicbadger get the sweet 55.5555555% availability
@kaito02 @DaveMWilburn @ironicbadger works 50% of the time, all the time.
@DaveMWilburn @ironicbadger Those are rookie numbers. Always aim for 1!
@DaveMWilburn @ironicbadger 9.9999% is also five nines
@ironicbadger big oof (for smallish values of oof) but I’d also like to see that graphed against active usage volume to get the full picture
(yes, regardless, a “hyperscaler” should be able to handle the volume)
@ironicbadger Geez, this is like what happened when they first tried to convert Hotmail to run on Windows servers.
@jmelesky @ironicbadger That was fun.
A graph demonstrating that the GitHub website's uptime suffered considerably after being acquired by Microsoft.
The graph shows average uptime of the website by-month from April 2016 to January 2026. Months that earn 100% are colored green, while months that miss that goal are colored either red or yellow, depending on some undisclosed metric of severity.
A line is marked in November 2018 where Microsoft acquired GitHub. Before that, no months were visibly worse than 100%. There's one red dot, but it's not visibly different otherwise. For the first year after the acquisition uptime is worse but acceptable, and stays above 99.95%: six out of 12 months earn 100%.
After October 2019 the story flips completely. Only one month has earned 100% uptime since then, and the remaining months vary wildly from 99.98% to about 99.70%. May 2023 was the worst, crashing almost as low as 99.5% uptime.
Honestly, it looks like a seismograph that's started recording an active earthquake.
@AcornSquashbuckler @ironicbadger When did they Rewrite It In React? 🤔 🤡
Azure?
If you bought MSFT 5 years ago, you would be "as you were" ( and a bit of a loser ).
https://g.co/finance/MSFT:NASDAQ?window=1Y
Check out these other losers:
https://www.google.com/finance/portfolio/413d47c7-2932-4196-8b7b-3d073a637f14
@ironicbadger as a post msft acquisition survivor this tracks
