I stood up a Yunohost and installed Mastodon a few months back. I had issues with storage and exponential growth as a result of federating with other instances.
It was just too much work keeping the storage at a minimal level for a single user instance, so I ditched it.
Is there anything like that I need to consider before I try my hand at Lemmy?
These things hurt me while setting up my Lemmy instance on my Raspberry Pi 4 via Docker.
- The instance name must be less than or equal to 20 characters in length (database limit)
- The lemmy and lemmy-ui docker images must be arm64 (for my Ubuntu 22.04 setup on my RaspberryPi)
- The certbot image needs to be added to the docker-compose from the docker install instructions and “depend_on” the nginx image
- I needed to disable the 80->443 redirect in my nginx config in order to get my initial cert (maybe there’s another way)
- The lemmy container needs its own network to allow it to access the internet (permitting searching)
One smaller gotcha is that lemmy-ansible only supports apt. README doesn’t mention that at all.
Disk space is a bit of a concern since I started hosting this instance. The Lemmy database is 12.2GB and pictrs is 1.8GB after a little over a month. I set all images to convert to WEBP and scale down to save on disk space but the database is growing pretty fast. From my understanding, they already fixed this and we’re just waiting for 0.18.3 to release.
Mastodon caches images from remote instances, Lemmy doesn’t. It links, which has it’s own issues (images disappearing from remote posts if the instance goes down). Everybody I’ve seen complaining about storage space have been open Lemmy instances with loads of users potentially uploading images to the pictrs sub-system. There’s even been requests to limit image storage for individual users.
If you’re thinking of using the YunoHost version of Lemmy, be aware that the pictrs subsystem is removed due to technical issues, I would absolutely recommend using the ansible install. But that might defeat the purpose of easy self-hosting for most YunoHost users…
From a non-technical aspect, the biggest issue (or at least used to be) on Lemmy was the ideological attacks and isolation by the fascist core community of any non-conforming instance. That had a chokehold on the lemmyverse to the point I saw no point running a Lemmy instance unless I self-censored my sometimes negative opinions of the Chinese government. Seeing the core developers being willing to hold back their own project was the biggest surprise when self-hosting Lemmy.
Thank you for responding.
I agree, the Ansible route is very daunting and does negate the user friendly appeal of Federated social media. But, it’s early days and I am hoping there will be an emergence of one-click installs soon.
This is digressing somewhat but what’s the deal with Lemmy / China / CCP? I’m a relatively normal 40 something. I just want to use the internet in a normal fashion.
Lemmy is created by what many describes as “tankies”, fundamentalist communists. The lead devs are, in my impression generally supportive of the Chinese government. They run their own instances, dedicated to their own world view and interests. I don’t have the slightest problem with that.
Hopefully due to a lack of foresight their instance was presented as a flagship instance on the official landing page and other places, effectively making it the representative for Lemmy for a long time. With their moderation practices, they in large parts drove away what I consider reasonable users regardless of personal ideology.
This created an monolithic instance that dwarfed every single instance besides one; the other dev-run instance which seems like a containment instance for those too fundamentalist to stay on the “main” server. For a large part Lemmy was a downright nasty monoculture that would harass even lead developers for daring to interact with code-contributing users on instances deemed “problematic”.
With a very generous blocking practice, these monolithic instance basically shut out anybody not agreeing with them from the majority of content - including non-political. This in turn gave other instances less of a reason reason to stay on the Lemmyverse which reinforced the monoculture.
The devs did take steps to rectify this problem: they offered a year of free Lemmy hosting for anybody willing to admin a general purpose instance, recruited other general purpose instances to promote on join-lemmy.org. I respect that. But at that time it seemed too late to change the direction. When I realized the bad reputation Lemmy as a project had amongst some of the larger fediverse influencers, I realized that Lemmy would not be viable until it got a massive and sudden influx of new users to shake up the overall federated lemmy community. And now here we are!
Kind of hilarious that a bunch of CCP fans end up building a censorship-resistant social network that would totally be banned in China.
That last point is interesting. How did they fix it? Or did they?
A pictrs storage limit per user? Separate subsystems without interaction so not possible last time I checked. There are quite some quirks surrounding lemmy and pictrs, you could dig through lemmy.ml for posts about it.
Sorry, I meant the ideological attacks and isolation by a fascist core community of any non-conforming instance. I imagine you mean eventually, a large main group of instances could block off small instances being part of the fediverse? How did Lemmy devs solve this? Or did they even?
How did Lemmy devs solve this?
When the devs released an update that allowed Lemmy to communicate properly with Mastodon and Friendica, a lot of the echo-chamber effect disappeared. But the dominance of larger instances kept up, sure anybody could find a smaller instance or even spin up their own and be auto-featured on join-lemmy.org - But with practically everybody subscribing and posting to existing and established communities it was hard to challenge the monopoly and get a foot in the door.
I’ve seen instances with hundreds of posts relegated to irrelevance, no upvotes, no comments. Because they weren’t able to get the word out so people could have a choice to subscribe. Because they were blocked from the big announcement forums.
The only way to find them was to use the lemmy landing page and manually browse all instances to hunt down interesting subs. If I’m blocked from posting on the “new forums” section of lemmy.world, how would you ever know about my new forum? It would be dead in the water.
Or did they even?
Wider federation with Friendica/Mastodon++ was always planned. The reddit influx is by chance. I feel that the things that have changed the wider lemmy culture for the better have been by chance or as a by-product. Not by design.
And that’s why I believed, and still belive, the Lemmyverse needed some externally driven changes. You have no idea the difference between now and 6-12 month ago. Being able to just post somewhere without being afraid of a misinterpreted comment could communaly block hundred of fellow instance users from a majority of the total lemmyverse content has been… Illuminating on the plights of people living in totalitarian states.
If you gave up running a single user instance of Mastodon then I really don’t think you should be running a Lemmy instance. It’s a fediverse compliant system that will have many of the same issues as Mastodon.
I run a small 10 person Instance of Mastodon and it runs very well on a small server of 4GB Memory and 100GB disk. Not sure why you had issues with a 1 man instance.
Out of interest why would you want to run a 1 man Lemmy Instance?
I’m not alone.
https://github.com/mastodon/mastodon/issues/20255
Compute was never an issue. I had my instance hosted on Hetzner and it ran just fine.
My concern is the storage, as described above, constantly checking disk space, cronjobs, purging etc etc
And whether I host a single user or a multi user instance is neither here or there. My question still stands.
Cloud storage is expensive. Physical storage is cheaper in the long run. Might be a good candidate to actually self-host on-prem. Throw a bunch of SATA SSDs at it, which are astonishingly dirt cheap nowadays and probably fast enough for a medium-sized instance; maybe add an NVMe disk cache if you want to be fancy.
One of these months (or years…) fiber will finally be available in my area, at which point I’m going to move most of my servers from the cloud to on-prem. But for now I can’t, because 10 Mbps upload is a pretty tight bottleneck…
I never had to check disk space, cron jobs or do any purging.
Where did you run you Mastodon Instance. I run mine at Digitial Ocean, been running for nearly a year and have not encountered any issues and I’m running on the cheapest server they have.
