· 3 minutes to read · in how-to, tutorial, lemmy, docker, web

Lemmy Selfhost Install

My experience and problems ran into starting up my own instance

Notice ~~!

this page is very far from where i want it rn
these are the notes i made when i first learned to install lemmy on docker using ansible
i plan to redo this page when i have time. but for now i hope this helps if you ran into some of the same problems i did

at the time of writing lemmy does not compile and host their own image for arm. so you will need to use someone elses. i found masquernya on docker was compiling and hosting an arm version of lemmy and lemmy-ui so i used that.

ansible with docker / arm64 raspi mods to make it work

  • had to manually install a few things on first time setup
  • docker cant connect without /etc/resolv.conf having external name servers (8.8.8.8, 8.8.4.4) before local adresses
  • /etc/resolv.conf keeps being reset
  • /etc/nginx/sites-enabled/DEFAULT removed
  • on ansible lemmy.yml docker images for lemmy need to be set to arm images (masquernya/lemmy, masquernya/lemmy-ui)
  • on ansible templates/docker-compose.yml needs "platform: linux/arm64" in almost every service. "platform: linux/arm64/v8" for lemmy/lemmy-ui

trying to pull images from the repo to docker kept saying it could not resolve. this is because on my version of debian the dns resolver by default was only looking on my intranet. adding cloudflare’s or googles dns to my /etc/resolv.conf before my local address fixed that issue. but a more permanent solution is to set that external dns resolver to the list of used dns inside of, in this case, NetworkManager. no im not going to tell you how to do that… look it up

Docker containers

{WEBSITENAME}_proxy_1  
{WEBSITENAME}_lemmy-ui_1  
{WEBSITENAME}_lemmy_1  
{WEBSITENAME}_postgres_1  
{WEBSITENAME}_pictrs_1  
{WEBSITENAME}_postfix_1

transfer images to target and load docker save {WEBSITENAME}_proxy_1 | gzip | ssh lemmy@{SERVERIP} 'gunzip | docker load'

issues installing:

  • python docker “This environment is externally managed”
    FIX: remove Externally Managed folder sudo rm /usr/lib/python3.x/EXTERNALLY-MANAGED

  • letsencrypt “Some challenges have failed”
    FIX: port forward 80 (http) & 443 (https)

Migration

ok so you are transfering both the container images plus their respective volumes

docker volume transfer script
[https://github.com/ricardobranco777/docker-volumes.sh/]

not every volume needs to be migrated
i cant remember which ones. i hope to fill this tutorial out later when i have time to test install again

  • stop containers on donor server

pnr = probably not required

  • commit containers to repos pnr

  • docker commit $CONTAINER $CONTAINER for each pnr

  • save repos and send to host server to load pnr

  • docker save $CONTAINER | ssh $USER@$HOST docker load pnr

  • save volumes for each docker service
    docker-volumes.sh $CONTAINER save $CONTAINER-volumes.tar

  • copy volumes to host
    scp $CONTAINER-volumes.tar $USER@$HOST: (that leading colon is important)

on host:
run ansible install compleatly
stop all docker services
load volumes, but only volumes that are not empty

{WEBSITENAME}_pictrs_1-volumes.tar  
{WEBSITENAME}_postfix_1-volumes.tar  
{WEBSITENAME}_postgres_1-volumes.tar

docker-volumes.sh $CONTAINER load $CONTAINER-volumes.tar
start all docker services

Email DNS records

this is not a complete list of email records required for full auth
but this is what started allowed email to start working for me

dmarc record:
v=DMARC1; p=quarantine; pct=25; rua=mailto:dmarcreports@{WEBSITENAME}; ruf=mailto:dmarcfailurereports@{WEBSITENAME}; aspf=s;

spf record:
v=spf1 ip4:192.168.0.1 a:{WEBSITENAME}