• 32 Posts
  • 116 Comments
Joined 2 years ago
cake
Cake day: June 18th, 2023

help-circle

  • All right, the server has been updated to the latest stable version.

    I initially intended to leave it running for a few hours to make sure everything is stable, but somewhere along the server reboots, the bot got activated once more and errr, well… it’s been posting happily for some time now.

    I guess that answers my question whether or not it’s still compatible with the new server version.


  • Dropping by for another update:

    • Server has had Lemmy updated from 18.4 to 19.3
    • pictrs has been updated from 0.4.x to 0.5 - that also took a nearly a day
    • Just updated postgres from 15 to 16

    There are still some updates and migrations left to do, but things are looking up, I think the worst is behind us.

    Having said that, I’ll be taking the server offline now for the rest of the updates - hope this comment was properly propagated :D



  • Small update

    Image showing statistics of the file transfer, currently at 19 million checks after 4 days, 17 hours

    So it’s been over 4 days of file transferring. There are about 13.6 million files in the source buckets, and yesterday we passed 14 million file checks.
    So I’m guessing that means it needs to do multiple checks per file. I’m hoping it’s 2 checks per file, meaning the process is over half way.
    The good news that most of the data has been transferred over (source shows 727GB used, destination shows 713 GB).
    The bad news is that due to how the files are structured, it’s incredibly hard for the sync process to estimate which files it still has to move over, so it does need to check them all one-by-one.

    Ironically, me posting this image above would mean there’s 1 file more that has to be synced over. But I’m willing to let it be a casualty of this war ;)





  • It’s noble of you to want to protect those people, but the moment they post content to reddit, they have given reddit the right to duplicate and redistribute their content as it sees fit. So the way I see it, that ship has sailed, and I’m pretty sure most creators are aware of this.

    Having said that, if someone ever reports they want to take their content offline, I’ll gladly do so. So far, that hasn’t happened yet though.

    Edited to add: the lemmit service doesn’t even copy most content, only relatively small (animated) images. On account of storage (and processing) larger files costs money. So most content is hotlinked. That means that when the original is removed from Reddit, it’s removed from lemmy as well.









  • Ah, I guess I must have overlooked that part. There are several reasons for not wanting to allow signups.

    One is quite simple, cost. Right now this is running on a small, single core instance. It often stutters (especially when handling video updates), and that is not an issue, since that just means it’s going to take small while before updates are sent out. But you wouldn’t want to have that delay for actual users. Right now the costs are quite manageable, if I have to scale up in order to provide a fluent experience for its users, not so much.

    Most of the other reasons come down to the responsibility of having to provide a home to any outside users that sign up. I don’t have the interest or time to maintain a community of people, nor to guarantee the uptime that such a server would require. It also wouldn’t work. The largest Lemmy instance in existence, lemmy.world, has defederated from this instance. So any users that sign up here, would be devoid from content on there. And as you said, any other instance can decide to do so at any time (in fact, I very much suggest they do so in the FAQ).

    I could go on, but I think you get my drift.





  • I don’t know how the karma thresholds work behind the scenes, but might I suggest for the bot to do a “top for” sort instead? Like it will only repost top content for the past 6 hours only. This will also help get more quality content as well and avoid reposting low effort/quality posts.

    This is effectively already kinda how it works. For each subreddit it periodically (anywhere between every 30 minutes to every 12 hours, based on subscriber count and posts per day) requests the “hot” content feed. It then checks each post if it has at least 20 upvotes, and a 80% upvote to downvote ratio. Those numbers are configurable, but that’s what they’re currently set to - I believe they’re a good mix between filtering out the complete garbage while still making sure it doesn’t miss good content is.