This weekend I installed my own Lemmy instance, so I want to share the instructions to help others, who want to do the same.
I used the Ansible script and it was pretty easy. First I wanted to use my existing PosgreSQL server, what I already use for my Akkoma server. It didn’t really work out that well, the migrations failed and I couldn’t figure out what didn’t work. Eventually I just went back using PostgreSQL on Docker. If you don’t start modifying the script, and just use the dockerized PostgreSQL, you will have no problems with the installation.
What you need first is a cheap (or expensive, if you decide to invite million friends to your instance) VPS: I use Hetzner Cloud, which has been working for me super well for many years and I’m very happy with the service. I got the second cheapest AMD instance, with two cores and two gigabytes of RAM. Before buying the instance, you need to upload an SSH key to Hetzner. If you don’t have one, creating is easy from the command line: ssh-keygen -t ecdsa. What you need to give to Hetzner is your public key; the one with the .pub extension in your $HOME/.ssh directory. Do not give the private key to anyone. Go with Ubuntu, might work the best with the Ansible script.
You can now SSH to the instance: ssh root@<ip-address from the Hetzner control panel>.
Next what you need is a domain name for the server. Lemmy wants an A record, and being a good internet citizen, you also get an AAAA record for the IPv6 users. I use Cloudflare for my DNS records. It’s very easy to set them from their control panel. Do not set the proxy on just yet, we’ll come back to that later. You can get the IP addresses from the Hetzner panel. The IPv4 you just copy, for the IPv6 you have to replace the ::/64 with ::1.
Now you should be able to ssh to your instance with the new domain name. It’s time to follow the Ansible instructions for Lemmy, just run the script and see it’s done correctly with no errors. When you can login to your Lemmy instance as an admin, go back to Cloudflare and turn on proxying to your A and AAAA records to hide your server IP and prevent DDOS attacks.
The first time federation is a bit slow in the beginning. Go to search in your instance, and search for !lemmy@lemmy.ml. It takes a while for the result to arrive. You can SSH to your instance, and look for the logs of your Lemmy image:
root@lemmy:~# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
9e940b84cc45 dessalines/lemmy-ui:0.17.3 "docker-entrypoint.s…" 22 hours ago Up 22 hours 127.0.0.1:6719->1234/tcp lemmynaukio_lemmy-ui_1
6442d9d93554 dessalines/lemmy:0.17.3 "/app/lemmy" 22 hours ago Up 22 hours 127.0.0.1:20926->8536/tcp lemmynaukio_lemmy_1
36a030f7bf27 asonix/pictrs:0.3.1 "/sbin/tini -- /usr/…" 22 hours ago Up 22 hours 6669/tcp, 127.0.0.1:8934->8080/tcp lemmynaukio_pictrs_1
979be89076b2 postgres:15-alpine "docker-entrypoint.s…" 22 hours ago Up 22 hours 5432/tcp lemmynaukio_postgres_1
774112d48c87 mwader/postfix-relay "/root/run" 23 hours ago Up 23 hours 25/tcp lemmynaukio_postfix_1
> docker logs -f 6442d9d93554
This should start showing you the federated posts in real time. Eventually your search will show up, you can click the community open and subscribe to it. Do the same for other communities what you want to follow, federate other instances and eventually you are part of the federation. It gets faster and easier for the other users, but the beginning is a bit slow.
Congratulations, you’re now a Lemmy admin and part of the bigger federation.
You must log in or register to comment.
I used docker-compose version and had to work around a couple of bugs like needing to redundantly install Nginx and Let’s Encrypt for it to work properly with SSL and also having to add the lemmy container to the internet facing network due to the DNS not working on it and subsequently loosing federation capability. Overall a bit of a struggle, but this is common with FOSS.
Is there a way to tell it to use an existing nginx instance? I run nginx proxy manager for all my stuff and can’t forward 80/443 twice.
Well if you are using docker-compose you could probably get rid of the nginx container and only deploy the other four: lemmy, lemmy-ui, pictrs, postgres. You would then use the nginx.conf stuff you have in place for the docker container of nginx to proxy to lemmy-ui and lemmy on ports 1234 and 8536. Or if you plan to keep using the docker container for nginx then you can change the listening port in the nginx.conf of the container:
listen 80;to something different like
listen 1080;Also in the docker-compose.yml you would update the nginx ports to
1080:1080.Hope this answers your question.
Thanks! I’m new to docker-compose (been spoiled by unRAID’s community apps) but will give this a shot.
No problem, if you run into issues let me know I may be able to help. Not an expert at Docker, but I do have a handful of containers running here and there.
Thanks!
You are welcome!
If you follow
allfrom your instance, you should see all the communities from the servers you federate with.
Thanks for the post! Been thinking of spinning my own instance as well because my Hetzner server doesn’t get used that much. Though then I would have to actually make sure it stays up and that I won’t just dip out one day lol.
If you have the talent to run a server somewhere, do it for your friends and get people to the Lemmy ecosystem.
Thank you for the write up! Looks like a simple thing to do!
@pimeys@lemmy.nauk.io what disk space do you recommend?
I just went with the 40GB disk that came with the machine. If you upload a lot of stuff, you might want to use an object storage with the included
pictrssetup. A good and cheap solution is Cloudflare R2, that should cost nothing for a small Lemmy server, and is quite cheap still if you upload tens of gigabytes of stuff.Read more how to use the object storage from the pictrs repo.
Another option if the disk gets full is to mount a volume from Hetzner, and use that volume for
pictrsor the database. I doubt you need that though.
Hey, have you managed to get emails working? Hetzner blocks port 25 and i cannot for the life of me figure out how to get emails to send using any of the external services.
It was easy. Just look up what smtp server you want to use. I’m on fastmail so I just followed these instructions:
https://www.fastmail.help/hc/en-us/articles/1500000279921-IMAP-POP-and-SMTP
I used smtp with starttls and configured that in the Lemmy hjson config file.
Port 25 is not used because you don’t want to send your mail unencrypted over the internet.
Thanks, i don’t want to unblock port 25 particularly, i’ve been trying to use port 587 with no luck (I get an error about connections). Not sure if the issue is with my postfix relay or with my lemmy hjson file. No matter what i try to change, i can’t seem to get a connection to work.
I’ll keep trying :)
By default they block ports 25 and 465 afaik you can request getting them unblocked after you paid your first invoice and your account is at least one month old. For some reason they aren’t blocking port 587 so you could connect to your mail server via that port if you don’t want to wait for the first month to be over using starttls and after a month switch to 465 with normal tls. And as @mrmanager@lemmy.today already mentioned you shouldn’t use port 25 since you’d be sending your mails unencrypted.
If you send a ticket to them in the cloud console, they will also unblock port 25
After many battles with getting my Cloudflare, ansible, and digitalocean droplet set up the install completed without error. However, if I turn off the proxy and navigate to my domain’s IP Address I am greeted with the NGINX success page for installation. I have tried the ports listed by running docker ps but none of the ports work there either. Is there supposed to be a page I would log in to administer the Lemmy instance or does it simply just kind of exist?
If I try to navigate with Cloudflares proxy enabled I get this message when I navigate to my URL. 404: FetchError: request to http://lemmy:8536/api/v3/site? failed, reason: getaddrinfo EAI_AGAIN lemmy
Running the logs on the container results in a whole lot of
root@lemmy2:/etc/nginx# docker logs -f 0a17a87732ddthread ‘main’ panicked at ‘Failed to load settings file: LemmyError { message: None, inner: Hostname variable is not set!, context: “SpanTrace” }’, crates/utils/src/settings/mod.rs:18:33
Did you set this value?
https://github.com/LemmyNet/lemmy/blob/main/config/defaults.hjson#L72
If you use Ansible, it takes it from the path:
mkdir -p inventory/host_vars/<your-domain>.I’m also experiencing this issue as well. From what I can guess I think its do to reverse proxy not having a custom location “/api” port 8536. I have no clue how to set this value using the ansible build since you cant access nginx webui. (very new person here) lol
I did not. I just copy pasted the entire thing. I will modify it tomorrow morning and see if anything changes. Thank you for the response.
*edit for update. That did not change much. Docker logs remain the same. As does the URL error. I might just try docker instead. I’m familiar with Docker. Just wanted to try and learn Ansible for a change.
The first time federation is a bit slow in the beginning.
Since I don’t know any better place to ask and I also setup my instance on Hetzner maybe some of you could provide me with some input regarding federation. I’m able to search new communities; the way I’m doing it right now is by searching for their handler (this !comunnityName@InstanceName thing) on my instance. For some reason if I haven’t searched for the community before no search results show up but I can switch to the community all list and see the community there. After subscribing to a community everything works nicely, I see posts, comments everything. But my main question is, if there is a way to federate a server (e.g.
lemmy.ml) in a way that I can just click on communities on my server and see every community on the federated servers without having to manually search them first?This is something that I wasn’t able to grasp from reading the lemmy docs and also didn’t found a satisfying answer to when googleing.
I think if you follow one community, it kind of federates slowly all other communities from that server too. I don’t know for sure does it work like that, but it looks like I get all the other communities too eventually.
How does it compare when hosting own instance. I have dedicated server and probably going to spin up instance just for me and friends.
Skip the Ansible and install the Docker containers by yourself. Something Lemmy should document a bit better is what version and what extensions of PostgreSQL it needs. If you have an existing server, some of the migrations might fail. Their docker image is using PostgreSQL version 15, so I think that is a safe bet to start.
And even better, if using NixOS, there’s already a recipe for Lemmy: https://nixos.wiki/wiki/Lemmy
I did exactly this a week ago to set up my instance but I opted for abit more power, 3 cpus and 4 gb memory. I figured anything under 10 bucks is not very expensive. :)
I already spend about 40 bucks a month for my instances, so wanted to save money a bit. I have more CPUs for Akkoma due to it using a bit more resources than Lemmy.
Thanks for this. I did the same yesterday on DigitalOcean, but I might migrate to my bigger Hetzner instance if I get enough signups.
A question on searching for communities though:
When I search for a community on another instance to add, I search for !community@instance.tld and nothing shows up, but when I reload the communities window and go to “all” the community shows up to subscribe to.
Do you see the community first time, or is it the same for you as above?
I sometimes have to do that, switch between all and local, and sometimes just wait a few seconds and it shows up. It’s some background process and the user interface doesn’t show anything so it can be confusing.
Yep. When you follow the logs, there’s lots of action after searching for communities and after subscribing.
OK thanks, good to know it’s not just me. I’m sure it’ll get better now we’re gaining more momentum with Lemmy.