Lemmy.world down, probably following the upgrade. A reminder to move to smaller instances for a better experience
Lemmy.world down, probably following the upgrade. A reminder to move to smaller instances for a better experience
A reminder to move to smaller instances for a better experience
A reminder that this constant advice people blindly parrot to install and flock to smaller instance has now created something like 1000 new servers in 50 days that are poorly run and already going offline as quickly as they went online.
Github Issue 2910 is the kind of PostgreSQL problems that the developers ignored for months and people still defend the developer choices to have the code doing real-time counting of every single comment and post for numbers nobody needs to needs done in real-time.
PostgreSQL is voodoo to this project, they do everything they can to avoid going to !postgresql@lemmy.ml community and asking for help, learning 101 about how to fix their SQL TRIGGER logic like Github Issue 2910 spelled out June 4.
You can also just have multiple accounts. I have one on lemm.ee just for this reason.
Quoting myself from a previous post:
First of all, it's really fine to stay on LW for now, no need to rush anything. But if at some point you have some time for this, then read the following.
So, to pick your instance, you can have a look at https://lemmy.fediverse.observer/list, filter by "1m" to see what are the most popular ones. As you can see, with a 27433 monthly users, Lemmy.world is by far the most popular, which is why you might experience some issues from time to time.
You should have a look at the next instances on the list. Short story: lemm.ee, sh.itjust.works, lemmy.one, sopuli.xyz and reddthat.com are solid choices.
You are looking at instances with quite a lot of people (the more people help with filling your "All" feed), just not the most populous one (lemmy.world), the original one (lemmy.ml), and instances that are too specific, either due to country or specific focus.
Long story:
spoiler
- lemmy.ml is the original insance, also quite crowded, not really the best choice
- lemm.ee can be nice, you can have a look at it and see how fast it is for you. The admin communicates a lot and is very helpful.
- sh.itjust.works had some rough time in the last few days. You might also not like the name, that's okay.
- beehaw.org does not federate with the big instances, so if you go there, you will be in their own space. It can a valid choice, but please have a look at their guidelines first, they tend to moderate a lot. Can work for you, or not.
- feddit.de, lemmy.ca, discuss.tchncs.de, feddit.uk, aussie.zone are country specific instances, so probably not interesting to you if you are not from there
- lemmynsfw is a NSFW instance, probably not the one you want to move to
- programming.dev is an instance focused on programming
- lemmy.blahaj.zone is a pro queer instance
.
To migrate your settings (including subscriptions and blocked instances), you can use that script: https://github.com/CMahaff/lasim
when I start writing this comment, the post is 47 minutes old. if I understand the linked page properly, lemmy.world has been functional (all green checkmarks) for the past 10 minutes which is the furthest back the data goes. All the other instances are all green except for lemmy.one which is all red. I am assuming that 47 minutes ago, lemmy.world had red boxes?
Maybe a different link would have explained the point better but I don't really see how a 30 minute (??) server outage during an upgrade is compelling to avoid a large instance. Are you suggesting it's better to use a server whos admins don't upgrade? If not, is there really any size of server that would meaningfully avoid this kind of occasional disruption? Seems to me that the dynamism of the environment will inevitably lead to various problems. That's part of the experience. TBH threadiverse uptime on the whole is pretty impressive for such a ragtag groups of admins and devs.
I have accounts on some smaller servers but they have their drawbacks too. Using a bigger server is more convenient because the people and content is already there. It's easier. I didn't plan to use lemmy.world but I ended up making account there to use sometimes.
I think in a year or so the situation might be different. I see the ideological point and I would like it to be true. Maybe the technology will catch up. I think it would be nice to be able to programmatically seed content, but maybe that would be obnoxious to admins.
Of course I have one or two other accounts, but I personally like Lemmy.world. They serve as a necessary stress test that shows the devs and admins how to optimize further, and I just like learning admin practices at this scale of a userbase from a work perspective. Plus I don't want to be on an instance so small they can't or don't know how to handle compliance stuff and evaporate if something like that comes up. Not saying I know how to handle all of those situations, that's the job of someone else at work.
Lemm.ee is amazing with a top admin
As I said elsewhere
There are certain things that are memory intensive and CPU intensive. If you have 10k on one server doing that it really adds up. However having them across a wide range of smaller servers, its not such a big deal.
As a user, you literally lose out on nothing not being on lemmy.world. You can partake in all the same conversations, communities and everything. In fact when lemmy.world is down, you can still see everything and when it comes back up, your posts will synchronize. There's genuinely no upside to being on lemmy.world. That's the way the system was designed.
Not sure people will listen though. I will always talk up the amazing admin I have on lemmy.tf, but it's also worth mentioning that I have a bunch of communities hosted on other instances and each and every one of them is amazing.
The instance lemmy.kya.moe that I created my alt on performs well too if anyone else is looking for an alternative :D
The major issue I see with Lemmy communities and Kbin magazines, is that they both rely on a single server to be up and running to even work. Sure, you can cross-post to several communities at once, but that generates one different thread per community. Add to that the fact that to even start to implement a single distributed thread properly cross-posted to several communities, they must take into account that a given user or server may be defederated from one community but not another - should the user not receive the message if at least one server bans the user, or should the user be allowed to receive it if at least one server allows the user to receive it?
Lemmy's machine-generated ORM SQL and hand-made flawed PostgreSQL TRIGGER logic is so bad, bloated. The developers on GitHub brag about "high performance". It's unbeliable.
In reality, small instances work because it has so many SQL performance problems that it mostly only is stable with little posts and comments in the database. They dd everything they could to avoid using Lemmy itself to discuss !lemmyperformance@lemmy.ml topics and hang out on Matrix Chat to avoid using the constantly-crashing servers they created.
If you go to a server with no users creating comments and posts and only has a tiny amount of data, it does crash a lot less.
We're back!
While this thread has some interesting points in it, the majority of it is chaotic and confrontational. I'm closing this, as I believe we need to have a bit of a cooldown.