3 ms·
The post where they describe what they're doing sure makes it look like that. From what was described, this is a massive refactoring and recalculation effort to
by HCIdivision17 10y ago
The post where they describe what they're doing sure makes it look like that. From what was described, this is a massive refactoring and recalculation effort to pay down years of accumulated technical debt. Occam's Razor is probably good enough: they had tons of complicated rules in the system that goobered up their calculations, and they needed to get the large amount of ineffectual cruft out at some point.
There's even a clear note that the /top stuff should go back to normal after the recalc is finished and it's back up to speed.
Frankly, I was a bit shocked by the scope of it. Seems like a huge and terrifying update.
- jonlucc 10y agoI'm not sure if there's a commonly accepted way to handle the logistics of this, but I'd love to see a blog post from them about how they manage this on a site with so many active users at all times. I guess you start with backing up the production DB to an offline location. The fact that votes can be placed on old posts makes it difficult, I'd assume.
- HCIdivision17 10y agoI'll see if I can find it, but there was a comment in an AMA on the topic where someone asked if they have a test server and they said no. Not because they're goofy noobs at the business, either, but apparently it's just too impractical. A lot of the testing really needs the load to be properly exercised, and simulating that is sufficiently hard to probably not be worth the effort. So this update? Updating on Production, baby! Edit: I bet they do have test servers, but nothing full scale. Which really may emphasize your point more: I would love to know how they tested it and merged it in when there's only so much they can do offline!