Discussion
My essentially single-user Mastodon server is becoming more expensive to run on @mastohost without doing some cleanup. My problem is that I don’t actually know how/what to clean up.
Early on I used relays to improve federation; I think there’s a TON of data in the db from that which could just be purged, but I don’t want to accidentally lose any user data. A couple of years ago we did run `tootctl statuses remove` but I might have still had a relay enabled more recently.
I got the biggest space reductions simply by enabling the media retention days setting by entering 7 as the number of days... Essentially, as you probably know, this means that cached media from remote servers is deleted automatically if it's more than 7 days old (the default is blank, which I suppose means that cached remote media just continues to build over time). I went from using about 100GB to no more than 30GB in a single day and it's hovered there ever since! 👍
@admin my disk usage issue is actually more the DB than the media!
@cassidy That's a new one for me, usually DB storage is a fraction of media storage, not this large! It may be possible that deleting old cached media also deletes associated database entries for the files, but I'm just guessing there. Other things that could be generating entries might be preview cards, could try cleaning those out as well!
@cassidy@mastodon.blaede.family @admin@thebluehydrangea.org
Does Mastodon shove media into the db as base64 or something weird? I've seen systems do that before. Also can/are you leveraging a remote object storage?
@egee I pay masto.host to deal with those details. 😅 There's a separate media db so I don't think there's anything weird like that tho.
I know in the past the database would balloon if I enabled very active relays; basically all posts from across the entire fediverse would be cached in the db regardless of whether I had or would ever see them. Which is… not ideal. I'm hoping it's a pretty simple cleanup job…
My plan so far:
• Download a backup
• Have @mastohost run `tootctl statuses remove` again to see if that improves things
20 GB for my dumb little fediverse account just seems… wrong.
Any other pointers? Should I give up and just migrate to an instance run by someone else? Should I try actually self-hosting Mastodon?? Help.
@cassidy It would be helpful to know which tables are using the most space here.
`tootctl accounts cull` and `toolctl statuses remove` should help here I guess.
Also you should do a `VACUUM FULL` on the tables after any cleanup, as pg will not reclaim the used space.
I would love to have equivalents to those commands in scheduled jobs within Mastodon, but this requires making them much more robust to be able to do it confidently in an automatic manner
bin/tootctl cache clear
bin/tootctl media remove --remove-headers
bin/tootctl media remove --prune-profiles
bin/tootctl media remove-orphans
bin/tootctl preview_cards remove
bin/tootctl media remove
bin/tootctl statuses remove
bin/tootctl accounts cull
Install bonfire.cafe
Get the full app experience
