June 12 - Server Capacity Update

https://sh.itjust.works/post/49787

June 12 - Server Capacity Update - sh.itjust.works

I’m aware that some of you have been getting some errors loading this instance. This was a configuration that I needed to be adjusted which has since been. Do be patient if we run into other issues, I’ll be continuously working in the back-end with others to better your overall experience. Still a lot of capacity available!

if anyone starts getting any weird errors please do let me know but everything still seems to be running smoothly from my side.
I tried to create a new community a couple times and it hung a bit ago. I'll try again tomorrow.
That's not normal. Let me know if it happens again
Yeah still hanging for me (even on a different computer) :/ In looking at the dev tools, it doesn't seem to do a POST or anything. The only thing it does is load the svg for the 'thinking' icon on the button.
Yeah, it hangs sometimes. Log out, close the tab, clean cache, then log in again, should work 👍.
Great to hear, and thanks for the update!
Users should spread out across instances - that is the entire point of being distributed and federated.

Absolutely. At the same time, when you have a bunch of users migrating from a centralized platform who don't quite get the fediverse stuff yet, they're likely going to go to well-populated instances.

Hopefully as things progress messaging around this point will penetrate with new folks (like me!)

Would love to hear some more details on the misconfiguration if you want to share @[email protected]
just updated the thread post to include more information for your curious beautiful mind
Not sure if you already did or the default config does it but if you only have one backend in an nginx reverse proxy it might also make sense to configure the max_fails and possibly the fail_timeout options so nginx won't consider your backend down for a few seconds every time it receives a TCP error connecting to that (single) backend. max_fails=0 in the named upstream section (see http://nginx.org/en/docs/http/ngx_http_upstream_module.html#upstream ) is what you want here.
Thank you for this great server!
Thanks dude. The performance here is really excellent in my opinion. Had a couple of page errors but nothing that couldn't be fixed by reloading.
Thanks for addressing it so quickly!
Thanks for the update! Any plans for a separate channel outside of sh.itjust.works like Discord or Mastodon, just in case to give people a heads up if the server goes down or is in maintenance?
I like the idea of a status account on Mastodon. That’s probably a little more accessible than Discord.
Matrix would also be a viable alternative to Discord.
@TheDude does Lemmy support a distributed configuration with multiple database and app servers, or are you limited to a single instance of everything?
The official supported deployments are single instance based, that being said things are already broken up into separate docker containers so it should be pretty easy to do. Would need to do some testing before hand. If this instance continues growing this way I'll need to look into scaling horizontally instead of vertically
Thanks for the info! Keep up the great work man.
Software engineer here - check out CockroachDB, a distributed SQL database compatible with PostgreSQL clients.
PostgreSQL Compatibility

A summary of CockroachDB's compatibility with PostgreSQL

CockroachDB Docs

If this instance continues growing this way I'll need to look into scaling horizontally instead of vertically

Could you elaborate what this means?

Better to have more servers as opposed to one ultra powerful server because the ultra powerful server is expensive. Also, you pay for all the power even during times you don't need it whereas you can take down or add more weak servers as necessary.
Ah, that makes sense. Thank you for sharing.
Glad to have an update on this, Thanks!
Thanks @TheDude! Scaling systems like this is always a challenge and you really get to learn the performance quirks of the code. Thanks for all your work.

And through it all, the Dude abides.

Thanks for hosting, thanks for the expertise, thanks for the response time!

I joined for the server URL, and definitely stayed once I saw his username. Long live the Dude!
Thanks for the great server! I've read your sticky post saying the storage was the most unexpected ressource going up. Can you share what kind of storage space your talking about? I could in the future be interested in spinning up an instance, and I haven't seen any documentation or info on that.
Are things scaling well then? Is it still alright to recommend new people come here or would you rather hold off for a bit now?
I figured out the storage issue last night, the instance is only at about 20% utilization so we should be good to take on a good amount more. We probably will need to do some more tweaking as we grow but for now its looking pretty good!
Wow. 20% usage at 2.4k users is pretty damn amazing.
its not so much about how many registered users as it is about how many online users. The online user count is what adds fuel to the fire!
Oh wow so it's really the transactions per minute, not so much the baseline set of data?
Yeah, basically the problem is number of calls (requests) per minute. If it's not too many, everything works fine, if it overloads, you get 500 errors 🤷.
Including value to federated communities, yes?
I'm not sure about that, but most probably yes, since if you post on federated communities, your username pops up as [email protected] and the post is saved on your instance (from what I know).
In Montreal, do we need to send you some hard drives?
Good to know. Keep up the good work, you're the man.
Think long term. Don't take too many new users or you'll end up with loads of new content (mainly images). Storage is not expensive nowadays, but you can't load new disks to the arrays indefinitely.
Can't we just link to images hosted elsewhere? Like a certain site I once knew
We can, but then you might loose history. This is long known problem on forums. Trying to cut on disk space by linking to image hosts. Sometimes the images survive, most of the time they don't. If that doesn't bother most of the user base, fine, but certain communities would like to keep a history of the media, since sometimes they share valuabe information, like maybe schematics or art.
I guess it's a balance. I remember the imageshack purge but I also don't want to overload the instance.

Photobucket was next after that, not to mention countless others. Imgur has been pretty consistent over the years, but after 2 purges and hundreds of lost images (schematics in most cases) on my forum, I'm scared to death of trusting another image hosting company.

Basically, we still allow externally hosted images on the forum, but only for temp things, like buy/sell, stuff like that. Everything else is attached on the forum.

Imgur is deleting all images that were uploaded without being signed into an account.
Didn't know that, thanks for the info.
And all nsfw too iirc
What a dude. Thanks for your work
congradulations, you have now beat reddit itself for uptime, lmfao.
Funny enough I’d say Shit Doesn’t Just Work - it takes your hard work. Thanks for helping fill the Reddit void @[email protected]
Thanks for the update, i really like the transparency
Thank you for hosting this and keeping us informed!
Thanks. Glad I’m here.
Really glad to have a home here. Thank you for your work!!!!

Just out of curiosity, would it help if when posting images we use services such as imgur (or alternatives)? I’m assuming if there is storage issues those type of posts are the biggest culprit.

Thank you for hosting this server @TheDude :)

Yes 100%. Using external image hosting service would reduce how much storage is being consumed