In preparation for the Reddit June 12-14 Black-out
As the reddit mods gets ready for the June 12-14 black-out, there some anticipation that an influx in user base will shift over to many of the lemmy instances as user seek out a home to post their internet memes and discuss their interests.
In anticipation of this increased volume I will be growing our current instance from
16 CPU
8 GB ram
to
24 CPU
64 GB ram
This server is currently equipped with SSDs that are configured in a raid 10 array (NVMEs will come in the next gen that get deployed)
Earlier today I also configured some monitoring that I'll be watching closely in order to have a better understanding on how the lemmy platform does under stress (for science!)
I'll be sharing graphs and some other insights in this thread for everyone that is interested. Feel free to ask anything you might be interested in knowing more of!
EDIT: I'll be posting and updating the graphs in this main post periodically!
Last updated: 6:21AM ET June 12th
Just joined you from that featured site, it just works. People are complaining about how hard it is to switch, I genuinely don’t know what the difficulty is
Joined from recommended a bit less than an hour ago. Tried lemmy.ml, got rejected, went for this one cuz it was top recommended and I figured with the name and the user count the likelihood of rejection was smaller.
I am from reddit migration.
it's 100000000000 times better, e.g. more accessible with my screenreader, on reddit, I needed to use special app, called dystopia which handled voiceover more gracefully.
A little update for all of you interested. I allocated the additional resources to the VM and will post some updated graphs once they update with the new configurations.
For those who are like me and like looking at graphs here are some prior to the upgrade.
so if I'm reading this right, less than 10% CPU capacity and about 1/8th ram at peak times, before upgrades? gotta give you credit where it's due, that thing looks ready to take some abuse.
The only part I don't have graphed yet is the disk IOs. I'm going to need to invest a little more time to get that metric going captured.
Say you through 1000s of active users to this instance.. what would cause the bottle neck first? CPU, memory, network, disk? I'm thinking probably disk due to database optimization that need to be reworked on lemmy afterwards CPU and then memory.
Hello Lemmy! I've just made an account to get my foot in the door incase the Reddit execs don't roll back after the blackout, let's sit back and watch the fireworks.
Thanks @TheDude. I work in enterprise network/systems/cloud operations as a network/security engineer. Would love to contribute monetarily or with time.
I'm being a little private about the total amount of disk assigned to this instance for now. I do plan on sharing these details when I have implemented a viable solution. That being said I can tell you that with the amount of current users and activities this instance is growing by about 20GB per day in disk size.
It can definitely run on a Pi instance. The storage medium is going to be important as it will need to house a PostgreSQL database. Would be fine for a few users but not sure how many it would be able to handle concurrently.
It's got 4 cores at 2.16GHz (I think 🤔), 4GB of RAM. Regarding the storage, that can be arranged. Currently there's a 32GB SD card in it, it can be swapped for a 512GB one, no prolem.
So, how many users can an instance on a Pi like that serve?
I was looking at that earlier too. I found LemmyNet on GitHub but haven't had a chance to test it. But "can it run" is also a fundamentally different question than "can it run smoothly for end users as it gets hit with a huge influx of users," and given the larger context of a potential Reddit exodus, that scalability concern is probably not negligible.
I've never hosted anything big but cloud fanatic is cheaper than digital ocean. I migrated from the latter to the former earlier this year to save. And then someone told me AWS is basically free up to a certain usage point. $200 free credits/mo? I can't remember what the "free" part implied. As always, do your own research, but you could join one and switch servers later on I guess.
+1 Cloud Fanatic. They were Server Cheap until a few months ago, but I guess that name turned people off. I don't have any heavy usage so I cant really speak to that aspect, but I've never been over billed or unable to connect in 3 years.
I've recently migrated a forum community (vbulletin -> discourse) using a vpn by Contabo (germany hosting). Their price are incredible and while you don't get an official SLA, it's around 99.5%.
It probably doesn't need that much but I'm curious in seeing how much It can handle until it can't I have more resources available but it would make more sense after this size to start separating the different components of Lemmy into their own individual instances across multiple servers instead of trying to scale it vertically. Might require a little more time than I have but I'll see what I can manage!
I don't have a lot of experience administrating rust apps, but usually single instances hit bottlenecks when trying to utilize resources on a large enough server.
Like others have said, I'm happy to help if there's anything you need to keep this going. I have many years of Linux and Windows admin experience, observability/instrumentation, and I can code to an extent.
Yeah I'm trying Lemmy anticipating that Relay Pro may stop working soon. Searched for my sub reddit subjects, seemed to work though some of them seem a bit empty so far.