this post was submitted on 23 Jul 2023
239 points (99.2% liked)
Lemmy Server Performance
3 readers
1 users here now
Lemmy Server Performance
lemmy_server uses the Diesel ORM that automatically generates SQL statements. There are serious performance problems in June and July 2023 preventing Lemmy from scaling. Topics include caching, PostgreSQL extensions for troubleshooting, Client/Server Code/SQL Data/server operator apps/sever operator API (performance and storage monitoring), etc.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
This is fascinating
My biggest takeaway from reading through the GitHub comments though is that it seems like no one actually knows where much of the SQL comes from? As in it's possible that the bug in question is just one manifestation of old, handwritten Postres code that may or may not be optimized (Or even logical?).
I don't mean this in a critical way, as things like this are bound to happen in an open-source, federated world. However, I would think a comprehensive audit of the Lemmy Postgres triggers, queries, etc could potentially save us all from some future headaches.
That's not fascinating, that's depressing. Lemmy team lacks development skill.
I am always fascinated with these types of comments, specifically for a free and open-source software. There are lemmy instances supporting hundreds of thousands of users and trafic, feedback from both server owners and lemmy devs is almost instantaneous.
A platform like lemmy requires client side knowledge to build both desktop and mobile UI (that are performant), it requires ActivityPub knowledge to integrate with the Fediverse, it requires backend knowledge to build APIs for 100% feature compatibility with 3rd party apps. It requires DB knowledge to optimize queries, it requires devops / platform knowledge to deploy it.
And all of this is built in public.
BuT LEMmY tEaM lAcks dEvEloPmenT SkiLL – sure buddy.
What's so special about open source? Shit code is shit code.
It's more to the point that a project of this scale and range in the hands of a small group of people is expected to have some shit code somewhere.
Or there's just room for improvement and optimization, as each developer has its strengths and weaknesses, as any other professional, and a system like lemmy is very complex and really requires to cover a lot from backend to front end.
And there used to be only 2 developers.
I once check a open source implementation of a niche product from Microsoft, and it was a nightmare of unoptimized code. And Microsoft spent a lot of development resources there.
Creating lemmy as 2 people job is quite impressive. Luckily now there are resources for optimization
This is such a dumb take. Using a database efficiently is not some binary, once-off thing: you build what works based on the data you have at the time. When it works, you move on to other features. It takes analysis of real operation over time to find the bottlenecks, and discipline to focus on fixing the things that will have the most benefit to your users.
There are many successful tech companies who introduce features that create dogshit performance impacts regularly. They work because there are people looking at metrics and catching issues to fix. This is healthy.
Luckily with open source new team members or forking can address that.
Yep.