Over the 15-year history of CoasterBuzz, I'm very guilty of letting it age without any updates. The Internets have a weird fashion cycle that tends to be pretty short, and stuff gets dated even if you thought you were ahead. I suppose it doesn't entirely matter (no other way I could explain Craigslist), but I didn't want to go another three years without updating the site.
So I beat that deadline by nine days, and got the "new" site up last night. I say "new" because, if I'm being honest, it's not really new. The forums are a new version, finally, but that's it in terms of new features, really. I ditched all of the mobile views and consolidated it into one responsive design, mostly because the new forum version used it. I think most of the site looks much better, and it's super fast for club members in particular (because the ads go away). The home page, however, is a disaster. It looks pretty terrible.
That was kind of the m.o. for this release. I just wanted to get it out there even if it's a little broken. People have opinions and aren't afraid to share them, so why bother trying to predict what they'll like? This was a pretty low risk update, because the only schema changes involved a couple of forum tables. With things being that easy, there was no reason not to just get it up and start evolving it.
In previous major releases, I've gone on feature streaks and made new stuff, so hopefully I'll be motivated enough to do the same thing this time around. At the very least I need to do something with that home page in the short term.
From a nerd standpoint, I'm happy to see that rendering times for the home page are well under 100ms now, and the forum pages have been in that neighborhood as well. It has been a little inconsistent, probably because I should be running in a bigger VM with more RAM. I'm just too cheap to spend the extra $74 per month when ad revenue is so completely shitty. Google is averaging pages at 149ms, so that's pretty solid. I'm also not entirely sure if there's a latency issue with the database, but it seems to have plenty of overhead if I'm to believe the metrics there.
I would still like to get out to doing multiple instances, mostly for redundancy. That will cost twice as much as well, but I haven't yet thoroughly tested the Redis cache that will sit in between. I have a bit to do if I want to make this stuff scale out, but the fundamentals are there.
I can say it didn't take three years for a re-do, and I can sleep at night knowing that's partially true.