Conversation

Notices

1. Embed this notice
   Nemo_bis 🌈 (nemobis@mamot.fr)'s status on Friday, 26-Jan-2024 10:52:46 JST Nemo_bis 🌈

   in reply to

   • Fabio Manganiello

   @blacklight FWIW I use a variant of https://ricard.dev/improving-mastodons-disk-usage/ on #ResPublicae.

   • clacke likes this.
   • Embed this notice
     Fabio Manganiello (blacklight@social.platypush.tech)'s status on Friday, 26-Jan-2024 10:52:48 JST Fabio Manganiello

     in reply to

     • Gabriel Pettier

     @tshirtman There are usually scheduled tootctl scripts that admins run to clean up stuff older than a certain threshold. But the cache folder itself can't simply be wiped without basically everything on the instance breaking (including emojis and avatars), and it's often very painful to fix it.

     It seems like one of those bad design decisions where the cache folder has been used as a dropbox for almost everything, including things that the software is supposed to store long-term.

     So even with the cronjobs scheduled to run every day and very little local traffic, an instance connected to a few big relays ends up having many GBs of media cache that it can't touch.

   • Embed this notice
     Gabriel Pettier (tshirtman@mas.to)'s status on Friday, 26-Jan-2024 10:52:49 JST Gabriel Pettier

     in reply to

     • Fabio Manganiello

     @blacklight I would be surprised if there is not some level of consolidation at some point, unless a community is not federating with the rest, their hosting costs will grow with the size of the fediverse, rather than its own size, so small instances will be less efficient.

     I have no experience with hosting mastodon, but the concept of a cache that you can never delete seems weird, surely things not accessed in a long time should be fine?

   • Embed this notice
     Fabio Manganiello (blacklight@social.platypush.tech)'s status on Friday, 26-Jan-2024 10:52:50 JST Fabio Manganiello

     The plans are moving to pull the plug from this instance and migrate to a new one.

     Running Mastodon really isn't fun anymore. I need 100GB of space on an S3 bucket just to store a cache that I can never delete, and 6GB of RAM constantly allocated by Redis or sidekiq just to run an instance with me and 3-4 more active users. There must be a better way.

     I have registered a new domain, configured the DNS, and I'm currently toying with Akkoma to see if it meets my needs. I may also toy with CalcKey before taking any decisions.

     It may be challenging for me to immediately replicate all the feature of this instance on a new one (especially if I opt for Pleroma/Akkoma, as I'm far more familiar with Ruby than with Elixir/Nix), but it may be a good opportunity to experiment with some new shiny toys.

     Of course, I'll personally reach out to the active users on this instance to check if they need any assistance with migration / data dumps before it shutting down.