Doing a big update today in versions. Going from 0.19.17 to 0.19.20
This will include some down time, but nothing more than 5 minutes (fingers crossed). This upgrade will start at 13:00 UTC.
Here are the release notes for the updates:
https://join-lemmy.org/news/2026-04-20_-_Lemmy_Release_v0.19.18
https://join-lemmy.org/news/2026-06-09_-_Lemmy_Release_v0.19.19
https://join-lemmy.org/news/2026-07-27_-Lemmy_Release_v0.19.20-_Reduced_Memory_Usage
This will also include an intensive update to our bot/AI scrapping deterrents. Turns out, bots and AI really love all of your posts and comments! Let’s get 'em out of here! They bring down the site multiple times per day. Working on a better fix for this.
This also may mean upstream issues like we saw in
https://lemmy.radio/post/14907033 via https://lemmy.radio/post/14120844
Over the next few days/weeks, please let me know if any federating issues appear.
73,
w0odl
Completed with notes:
Update’s done, we’re on 0.19.20.
Downtime came in at about 6 minutes, so a little over what I promised. The database took longer to come back than expected. That’s also been fixed, more below.
What changed:
Lemmy 0.19.17 → 0.19.20. Release notes are in the original post. The big one is reduced memory usage, which this little server very much appreciates.
New bot/AI scraper deterrent. I’d been running a hand-rolled version of this since the summer. We’re now on the properly supported setup, which means it gets maintained and updated along with everything else instead of me babysitting it. The rules are better too: known AI crawlers (GPTBot and friends) now get turned away at the door instead of being made to work for it, and the check that real browsers go through is significantly lighter than what I had before. Most of you shouldn’t notice it at all.
Federation. I know this is the one people care about after last month. I tested it before, during, and after the upgrade and it’s clean, comments flowing in from every instance I checked, arriving within a minute or two. That said, if you spot anything odd with posts or comments from other instances over the next few days, tell me.
Bonus fix: the database hiccup that caused today’s extra downtime was going to happen on every future restart, it just hadn’t bitten yet.R estarts went from several minutes to milliseconds. Future maintenance should be a lot less dramatic.