fuchsiaaaaaaaaaaaaaaaaa
fuck the day has finally come where pixie.town's root disk is (basically) full.
Removing logs is enough to keep things up for now but needs actual changes
@f0x Nightmare. Not looking forward to that happening here, but we're still a ways away. What are you gonna do?
tech
@dumpsterqueer so, terra has a rather.. interesting disk setup as it's a fully kitted laptop;
sda - 240GB ssd
sdb - 500GB ssd
sdc - 2TB hdd
sda is in RAID1 with half of sdb, and that's currently the 230GB fs that is mostly filled by the ~160GB postgres files
so currently the possibilities are moving the db to hdd (meh), moving it to the currently unused other half of sdb (no longer RAID, meh), and maybe for the longer run replace sda with a 500GB ssd as well once i get paid/go around with the donation hat like last time
tech
@dumpsterqueer most of that db is synapse, it also hasn't been VACUUM FULL-ed in a long time (but that needs more free space so can't do that now)
the RAID setup is rather interesting, technically sdc is also a degraded RAID1 after a very interesting migration where I swapped out the other hdd it was raided with, for the 500 gb ssd
the joys of limited disk options in a laptop :')
tech
@f0x ahhh right synapse of course... how many years/users of synapse is that? wondering whether i should already be planning ahead for that kind of capacity--synapse and mastodon are running on the same machine in my setup, sharing a postgres instance, and the disk is only 256GB. I'm considering buying a very small shitty server and just stuffing it with HDDs 🙃
tech
@dumpsterqueer Synapse is (as always) rather inefficient in storage as well. Most of our storage is used by irc bridged rooms due to the huge amount of permission (/op *) and join/leaves (netsplits etc) events.
They can be largely compressed with https://github.com/matrix-org/rust-synapse-compress-state but that also takes ram/cpu/network cycles that currently aren't available, although I could run it overnight sometimes.
I think the server is about 2 years old now?
tech
@f0x 160GB of postgres? holy shit 🤔 That's a lot of toots.
I chuckled at RAIDing one drive with half of another, that rules, never would have occurred to me to do