Hi folks, I wanted to feedback on my recent update to 0.19.3.
Since I host many services, I’m using docker for deployment, more precisely https://join-lemmy.org/docs/administration/install_docker.html.
Since I dont use ansible (yet), I had to replace most of the content of the compose file to make it work. So far this has been successful.
But updating atm is a great gamble which I’m not a fan of. Other services do this way better, especially when hosted in docker. I can even let them run in latest-mode and nothing ever breaks. Granted, they’re far more mature than lemmy but I’d like to see a normal admin being able to host this so lemmy can grow.
My pain points are:
- I would like to have the option to just use docker without manually putting together the compose file when not using ansible
- pictrs breaks on every update since my server icon is a png and it for some reason says the picture is broken although I can upload and view it after. I need to delete the reference in my postgres database manually to get everything working again
- I analyze the compose files every time I update to see if any changes need to be added to my version. This time, I stumbled across the lemmy container hostname. This didnt break the whole thing but sorting by anything didnt work anymore, neither did using voyager. Commenting out the hostname made everything work again.
The point I’m trying to make here is that we should have a working docker compose file for every update. Pros can still change everything they like. If there is a reason that we must use ansible please state the reason clearly in the documentation. I’m fine with doing so if there is a compelling reason.
Also, I use docker for years in production now and I would like to help if this is needed. Let me know if you’re interested.
Have a good one.
You must log in or register to comment.
I use docker and whilst any included docker compose file is appreciated, I only use them as an example. I’ve always curated my own docker compose to fit my environment.
You should raise 2 GitHub issues for the PNG icon and hostname issue if reproducible.
I appreciate the suggestion. I‘m very sick atm and only took the time to troubleshoot because I wanted to doomscroll on the couch in peace. :) I‘ll raise the issues when I‘m feeling a little better. Otherwise the bug description will be incomprehendable.
About the compose file: I can see where you‘re coming from. I also change the stuff to fit my config but I would like to have the option to align myself with the devs as far as possible to have easier troubleshooting. I’m not one to read all release notes and research any incorporated services pros and cons since I dont have this kind of time. I check if the new versions may need new parameters and do compose up -d, start troubleshooting.
The more unique your setup, the more problems will be unique as well. As an admin by trade I always put availability first and customization second.
Have a good one.
I‘m very sick atm and only took the time to troubleshoot because I wanted to doomscroll on the couch in peace. :) I‘ll raise the issues when I‘m feeling a little better. Otherwise the bug description will be incomprehendable.
Ugh, I feel for you! I hope you get to feeling better soon!
Thank you very much. Havent been this sick for quite some time. I hate it a lot.
Have a good one.
Also host my own instance, got absolutely burned by :latest being “literally the last commit” and not “latest stable release.”
I sort of want the devs to take a page out of mastodon’s book and make the microservices (pictrs, UI) operate out of the same base image.
Afaik that’s docker’s fault. They always mark the latest uploaded image as latest. But also I thought it’s well known that :latest shouldn’t be used for any kind of software, unless you are a developer
That’s not quite the case actually, the image builder doesn’t have to update the
latesttag, but that’s the usual default configuration since it’s easier for out-of-the-box usage such as newbies, demos, etc.
I can relate. Thankfully I caught the warning not to use latest. I‘m still kinda missing an explanation for this as it seems odd. Maybe its because of how early the versioning still is.
In any case, I‘m kind of on board with splitting the services as is docker custom and also to make high availability and horozontal scalability easy. I think mastodon is a lot further in development. Also, I kinda enjoy the rather stubborn nature of lemmy in terms of trends. It could go both ways but so far, lemmy did pretty good.
Feel free to reach out here or on matrix if you want to share experiences as an instance admin. I‘d be curious.
