Elektrine lite

← Feed

@palomakop@merveilles.town

Post #980639

2026-03-29 21:53 UTC

i have kind of been ignoring this for a while because there is a lot going on but there are about 1000 reasons i don't want to keep hosting my code on github anymore. codeberg seems nice but i kind of want to just make my own forgejo instance. it's mostly personal projects anyway and i've had approximately 2 outside contributions so i'm not that worried about the friction of other ppl PRing to them. that said, does anyone with experience doing this have any advice? i don't know if i can actually *delete* my github account for Reasons but it would be nice to at least move my "fun" projects off there. i should do it soon... aaa edit: one thing i'm wondering is, for anyone running a self hosted forge, are you getting hammered with crawlers / AI scrapers? my server isn't that powerful and i am curious if i'm going to need to batten down the hatches against that.

Replies (8)

  • @palomakop it’s pretty easy to run except you’ll be overrun with scrapers and want to put something like Anubis in front of it oh and if you have public signup enabled expect thousands of them from people attempting to post spam

    Open ##1846064

  • @joergi@chaos.social 2026-03-29 22:09

    @palomakop feel free to have a look at that script, if you want to have the history and every commit with the original date linked to your new account. https://codeberg.org/joergi/migrate_github_repo_to_codeberg With the official migration, I had the commits not linked to my new account, that was not what I wanted.

    Open ##1846065

  • @ledoian@pleroma.ledoian.cz 2026-03-29 22:38

    @palomakop I run Forgejo on a cheap VPS and the crawlers were there, but I decided that I don't need to have web frontend to the complete history, so I just serve 403 for browsing individual commits based on a URL pattern match and used to make the load rather low (they all tended to look into all the past revisions). Didn't even bother with Anubis, mostly because I was lazy to learn it and didn't want to wait for the challenges. I don't know why or when, but apparently they don't scrape me anymore? It used to be a constant stream in the logs (several in a second), now I see like one request in the 5 minutes. (Thank you for making me update my mental picture of how the bot abuse goes now!)

    Open ##1846066

  • @monoxane@eno1.dev 2026-03-30 00:12

    @palomakop I've been running my own Forgejo for a year or so now and its been great, I have Techaro Anubis in front of it (like a lot of others) and haven't really had any issues with scrapers so far.

    Open ##1846088

  • @courtney@pdx.social 2026-03-30 00:20

    @palomakop I have a private Forgejo instance for my projects, with selected branches/projects mirrored to Codeberg for public. There are significant limitations to this setup but if your are ok with it, it's just. Forgejo has the ability to mirror to another git repo built in. (The limitation is this works best if your public Codeberg site is a read-only mirror. I don't have PRs enabled or actions running on Codeberg.)

    Open ##1846102

  • @bouncepaw@merveilles.town 2026-03-30 00:46

    @palomakop I used to have a Gitea instance, that was before Forgejo was a thing. Having it was a hassle to be honest, all this backing up and whatnot. And then somebody found that server and asked me if it's ok for them to use it. I thought a bit, said no and deleted that instance some time after. These days I use Codeberg for projects I would expect contributions to, and Sourcehut for more private stuff (most of repos unlisted). For self-hosting, Forgejo seems to be the most straight-forward choice. Community support, familiar UX, and built-in mirroring. I use mirroring on Codeberg, and it works really well. Crawlers and scrapers are not that much of a problem if your website is not popular, and it's probably not. That's the case for my server (I don't host a forge anymore, but I do have public services that work with git history, so I can use my setup as a comparison). I have 2 GB of memory, but didn't have problems with 1 GB too. Fail2ban will get you a long way. We Merveilles also have an instance: https://git.merveilles.town/explore/repos

    Open ##1846103

  • @samebchase@fantastic.earth 2026-03-30 01:42

    @palomakop fossil is easy to host. I host it for the wiki functionality, but at some point may host some code there as well. I don't think I have caught the attention of the crawlers. If that ever happens, I'll end up going with the recommended mitigation strategy

    Open ##1846108

  • @palomakop Yeah, I hosted some forks of popular keyboard firmware at my homelab hosted forge: https://git.pwarren.id.au/, they're quite deep and the LLM slop crawlers loved to trawl through that mire, I ended up making them private repos which now 404 to the slop crawlers, so the outgoing bandwidth is significanly less. They don't seem to stop though! Also implmented some apache config level blocking of the User Agent strings, which also helped enough. If it gets worse again I'll look at something like anubis.

    Open ##1846109