Nevar forget the Quadro Tracker, the device sold to police agencies allegedly able to track anything, which instead of a circuitboard was revealed to just have ants glued to paper with epoxy https://en.wikipedia.org/wiki/Quadro_Tracker
object storage saga: more documentation
There's official Mastodon documentation on setting up an nginx object storage proxy, too! https://docs.joinmastodon.org/admin/optional/object-storage-proxy/ No doubt the other config I found is based a lot on it. I'll compare the two and try to get to the best settings 🤞
object storage saga: plot twist!
I installed MinIO and got it running, but turns out MinIO no longer has a gateway function and B2 is now S3 compatible anyway?? Okayyy.... going to try it directly over the coming days then, with the main headache now being that I have to set up the host myself on nginx. Fortunately there's a pretty helpful-looking guide. https://thomas-leister.de/en/mastodon-s3-media-storage/
I am finding I'm benefiting more and more from @darius's work on the Hometown fork of mastodon: exclusive lists and local-only posts are such great features.
MinIO installation (advice welcome)
Looks like these were the instructions I needed on Debian: https://min.io/docs/minio/linux/operations/install-deploy-manage/deploy-minio-single-node-single-drive.html
Installation seems to be successful, the systemd service file was created, and I created the group & user. I'm not foolhardy enough to self-host object storage and only plan to use MinIO as a router to B2, per the gist seen here, https://gist.github.com/fwenzel/84a6f2f6f34d28d6693d8751f23819e2 so I'm guessing I just need an environment, an empty mount, and get the service up so I can hook it up to the B2 bucket and then hook up Mastodon to MinIO.
technobabble, no advice needed
Very timely too, because I have somehow Failed At Crontab and have had to delete media caches manually.* Today I logged into terminal JUST before we had another overflowing disk & attending crash situation, at 34 GB used out of 38 GB.
Also once the disk situation is better I'm looking forward to trying out more apps, like maybe a blog? Who knows!
* I'm more used to systemd timers but this server isn't set up to use systemd with the --user switch which would have been the cleanest solution, and some digging convinced me setting up user-level systemd jobs would be more trouble than it's worth especially since we're moving to objective storage anyway. Since periodic media purges will still be needed after the move I'll figure out crontab at more leisure afterward.
Went through the minIO site and realized I don't need Docker or Kubernetes to set it up! https://min.io/download#/linux Looks like operation: Object storage is a go 🤔
Longer-term I plan to move media files off-server to object storage. Once done this would make the storage situation even more comfy. It's just a little more involved than I thought because evidently Mastodon doesn't by default talk to the B2 buckets service I use and needs to be routed through a local copy of minio, which according to my research only runs on docker? As I have never docker'd before I'm just trying to understand what this thing lol. Anyway, I'll get there eventually and in the meantime we're not in imminent danger of running out of space.
We're currently using 24 GB out of 38 GB of space and I expect to clear out more space by culling remote accounts that are no longer there and removing unreferenced remote statuses, both computationally heavy operations that I'll be running during my daytime when the server is less active.
I set up a crontab to delete federated media from other servers every day, changing the rule to keep remote media for one day and not seven, because federated media fills up storage very quickly now that servers are more active. Which is a good thing, just needed to adjust to the change! This should get rid of the occasional downtimes we've been having, which were from the disk filling up.
Because a few people have tried this already:
Please DO NOT try to replace another fedi instance with a GoToSocial deployment on the same domain, with the same username. This will almost certainly cause issues with federation due to public/private keys changing, and it is not something we support.
In the future, we’ll try to provide a proper migration path between GtS and other fedi servers. But in the meantime we recommend deploying GtS at a subdomain that hasn’t been in use yet, like
Your enthusiasm about GtS is appreciated! But if you try to drop-in GtS in place of a preexisting fedi server, and things break, we will not be able to help you with them.
The typo STAYS
I'm the tech support person on writeout.ink. Avatar image is a butterfly sitting on the mouth of an ink bottle.
A small instance for writers.