this post was submitted on 22 Aug 2023
784 points (95.8% liked)
Fediverse
17776 readers
39 users here now
A community dedicated to fediverse news and discussion.
Fediverse is a portmanteau of "federation" and "universe".
Getting started on Fediverse;
- What is the fediverse?
- Fediverse Platforms
- How to run your own community
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Let the servers keep crashing, tell everyone to add new instances to help with performance, which puts 1500 rows into the database tables that used to have 50 rows and invokes a massive federation 1-vote-1-https overhead... causing more crashing... all the while ignoring the SQL design of machine-generated ORM statements and counting logic hidden in the background triggers.
... keep users off your sever as a method of scaling by crashing. It's one of the more interesting experiences I've had this year! And I spent all of February and March with the release of GPT-4... which was also interesting!
How is it going on that side, are they still ignoring your pull requests?
I've largely given up on pull requests.... for sake of sanity. But I waded back in...
I made a pull request today... and I very strategically choose to do it with minimal of features so that it would just go through... and I got lectured that JOIN is never a concern and that filtering based on the core function of the site (presenting fresh meat to readers) was a bad use of the database. I've never seen hazing on a project like this. Memcached and Redis should be discussed every day as "why are we not doing what every website does?", but mum is the word.
This is unfortunate to hear. Have you considered creating a proof-of-concept fork with synthetic data that demonstrates how much more performant a cached, filtered approach would be? I think a magnitude or two improvement of some key metrics with heavy simulated load would be quite convincing.
Of course, that would be an insane amount of work, especially if it would get ignored, but something to consider!
I already did an insane amount of work to populate a Lemmy database with over 10 million posts. It is so incredibly slow out of the box that the normal API would take days to accomplish this. i had to rewrite the SQL TRIGGER logic to allow bulk inserts.
Here is my work on that:
With this in place, 300,000 posts a minute can be generated and reaching levels of 5 million or 10 million don't take too long.
That's really cool work! It's a bit beyond my pay grade, so I can't really comment too much about it.
I had a look at the PR you mentioned, and again, while I can't comment on the contents because I am a little out of my depth, may I voice my opinion on the exchange? This is coming from a place of trying to help, since I really do appreciate all the work you've put in and are putting in, and the fediverse can really use your talents, so I hope I don't offend you.
From my reading, it didn't appear that you were being ignored/hazed, and it seemed like the devs would have been open to your improvements. From working and leading big teams, I've noticed that communication and managing emotions is often much harder than writing code. In the thread, it appeared that communication had broken down on both sides (and seemed to have been the case in prior interactions too). Since you mentioned your struggles with autism in the thread, I wonder if that played a part in the tone of the devs perhaps being misinterpreted ? This is, of course only my interpretation, and I could be completely wrong.
Ultimately Lemmy itself is an example of trying to build a community and consensus amongst a broad and diverse group of people, who will often not see eye to eye.
In any case I would like to say I personally appreciate your hard work and really do hope you're able to help make Lemmy better. Thank you!
Ok, so let's look at recent changes that they have deployed.... https://github.com/LemmyNet/lemmy/issues/3886
One of which makes entire tree of comments disappear. Do you see developers fretting over this and fixing it? Or do you see them ignoring the May 27 PostgreSQL JOIN problem.
How did such a bug go out? Do you see Lemmy developers actually using Lemmy to test things and notice these crashes and problems? Do you look at their posting and comment history? Do they actually go login over at Beehaw and Lemmy.world and see just how terrible the code performance is?
If it isn't hazing, what is it?
It's as if they build a product only for other people to use... and they don't notice any of the constant crashes, incredibly slow performance etc - and they act like nobody in the computer industry ever heard of Memcache or Redis to solve performance problems. If it isn't extreme hazing going on, then what is it?
I’ve been reading a lot of your exchanges on the Lemmy GitHub and I can tell you with a high degree of confidence that you are not the subject of a hazing ritual. What’s going on is a miscommunication issue; there’s no ill will directed towards you.
The Lemmy devs are under a great deal of stress these days due to the recent influx of activity, both on the big Lemmy instances as well as the Lemmy GitHub. You’ve clearly gone to great lengths to investigate various SQL bottlenecks in Lemmy, and this work does not go unnoticed or unappreciated.
The problem you’re likely running into is that the Lemmy devs are trying to address a wide array of issues, whereas you are zoomed in on some very specific performance problems. Whether or not the core devs are wrong when they say your findings are irrelevant is beside the point. What they are really saying is that they do not have the attention bandwidth to try to see what you are seeing right now.
If you find yourself unable to work with the Lemmy project, there are other fedi projects in Rust like Mitra or Kitsune which might be more receptive to your contributions. I’m personally very interested in seeing rudimentary Lemmy (Groups et.al.) compatibility in Kitsune.
Can you explain to me why it isn't social hazing?
Do you know how to read a SQL statement? I just can't grasp how it isn't social hazing. I've been reading SQL statements for decades, this is obviously a problematic one.
Can you offer alternate explanations of how 3 people could think that SQL statement isn't ... poor performing and gong to cause problems? And how an SQL statement without a WHERE clause took them months to discover and fix?
Extreme hazing is my best answer. I just can't accept that the SQL statements don't speak for themselves along with the server crashes. 57K users for 1300 servers is very... taking several seconds to load 10 posts....
Look at the date... May... this has been going on since May. If it isn't social hazing ... what is it? I keep asking myself that.
Like I said, this was my interpretation based on reading that exchange. It's difficult to convey tone or intention with text, but I didn't detect hostility from the devs, but I did sense that they were frustrated that process wasn't being followed. Perhaps they should not have gotten hung up on that, but it didn't appear to be malicious.
I do, and your arguments about the joins being problematic seemed solid. From having worked on systems with huge scale, I also agree that Lemmy doesn't seem to be big enough to be brought to its knees by the volume of posts it's processing. However, I'm far from an expert, so I don't want to suggest any certainty about the root causes, especially as I don't have the energy or inclination to dig as deep into it as I would to form that opinion.
I don't know why they weren't receptive, but perhaps they themselves felt attacked. I know that wasn't your intention, but misunderstanding happen, especially over text.
Here, you can dig into what posted days before the pull request you read:
https://github.com/LemmyNet/lemmy/issues/2877#issuecomment-1685314733
June 4:
Given that more than 8 JOIN statements is something PostgreSQL specifically concerns itself with (join_collapse_limit). I hand-edit the query with a single IN clause and the performance problem disappears. 8 full seconds becomes less than 200ms against 5,431,043 posts. And that 200ms is still high, as I was extremely over-reaching with "LIMIT 1000" in case the end-user went wild with blocking lists or some other filtering before reaching the final "LIMIT 10". When I change it to "LIMIT 20" in the subquery, it drops almost in half to 115ms... still meeting the needs of the outer "LIMIT 10" by double. More of the core query filtering can be put into the IN subquery, as we aren't dealing with more than 500 length pages (currently limited to 50).
If it isn't social hazing, then what is going on here? Why has this issue gone on since May and servers are crashing every day?
Funny, because I'm a published author and expert on messaging systems... like Lemmy. Iv'e been building them since 1986 professionally.
There was a massive thread I posted dozens of comments on that came before today's pull request... I suggest you read that too.
Did you notice them even acknowledge server crashes are happening? Do you think developers ever suggest Memcache or Redis? Or discuss how Reddit solved their scaling in 2010 with PostgreSQL?
I don't have any trouble understanding a bad SQL statement that has 14 JOINs and being told "JOIN is a distraction" after posting tons of examples.
Do we really need to spoon fed the stuff I did post?
Have you never seen social hazing in action? is it possible that I might be on to something going on psychologically besides my autism?
I can't believe anyone thinks a server should be crashing with 1 user on it.
Okay, I can't speak to whether social hazing happened or not, but I can tell you that you're making me extremely uncomfortable.
I started a dialogue, but at this point you're now sending multiple messages for each of my replies, and asking a lot from me in terms of attention. I do not wish to continue this conversation, but I wish you all the best.
Welcome to discussions with RoundSparow!
It can be a bit tiring interaction wise, but you usually can learn a lot
Haha, indeed. Any time I see an open-source discussion (especially a heated one), I'm reminded about just how much effort it takes to contribute. I'm happy to just stick to browsing memes :P
who would have predicted that Elon Musk would do all the wild things he did with Twitter. Reddit pissing everyone off in June... pretty odd how audiences are behaving in 2023 towards all this. Oh yha, Threads, that coming on the scene too. 2023 has really been odd for audiences.
The SQL speaks for itself, but I don't know what's going on in terms of why people are treating social media platforms like Lemmy, Twitter, Threads, Reddit this year so unusually. This SQL statement kind of thing has been covered in so many books, conferences, etc. It's like forgotten history now in the era of Elon Musk X and Reddit Apollo times.
I don't know what to say other than I can try to hire a translator or teacher to explain how this SQL problem is obvious and well understood 13 years ago. I mean, there was a whole "NoSQL movement" because of this kind of thing. But I clearly can't get people to hear past all the Elon Musk, Threads, Lemmy from Reddit ... and I'm left describing it as 'social hazing' or whatever is gong on with social media.
Lemmy has like 5 different Rust programming communities, but nobody fixing Lemmy. It's surreal in 2023 the Elon Musk X days. I think it's making all of us uncomfortable. The social movement underway.
That's quite concerning. Would you have a link to that PR? Curious to see the exchange
I already feel like I have to keep sticking my neck out to get them to question if using the ORM and a dozen JOIN statements isn't a problem.... but I guess I'll link it: https://github.com/LemmyNet/lemmy/pull/3900
As stated on my Lemmy user profile, I'm "RocketDerp" on GitHiub.
Honestly, the reason I keep making noise is because I'm sick of Lemmy crashing all the time when I come to use it... and I am on many servers that this happens. I really am not trying to piss off the developers, I even said I felt like I am being hazed, and I feel like hazing in general might explain what is going on with how much they are avoiding the elephant in the ROOM that ORM and a dozen JOIN might be the cause! Let alone the lack of Redis or Memcached addition being avoided, that's a second elephant on the second floor tap-dancing.... GitHub Issue 2910 was the straw that broke my back weeks ago, it took months for them to address it when it could be fixed in a couple hours (and it was weeks before the Reddti API deadline at the end of June.... and issue 2910 was neglected). The whole thing was a nightmare for me to watch...
Dude, chill. Even if you're right, having a meltdown on github doesn't help anybody. Go outside and take a breath.
I'm glad for you that mental control is so trival and you aren't near death in your life from your brain damage.
I just got back from dinner ant the months of hazing I've witnessed hasn't gone away. The level of social games being played with PostgreSQL in this project are levels beyond anything I've encountered in my 50+ years alive. And I've first hand seen Bill Gates and his team do all kinds of odd things to groups.
I am at a total loss to explain why such fundamentals of basic relational database are avoided in this project. If it isn't social hazing, what is it?
You're putting too much importance into this matter. If this is distressing you should let it go and think about something else.
The apologists come out of the woodwork around here who can't see an SQL statement for what it is, a charade. Anyone who has worked with SQL knows that this is bloated SQL statement and poorly engineered.
I notice the scientific facts of server crashing and SQL statements you won't discuss, but you sure dish out the social advice for me to "move along" like a Jedi mind trick. Let's talk about the human attraction to truth and honesty since you are so great at handing out life advice to people. What do you know about the works of Marshall McLuhan on media?
Repeating: Its’ as if the mere concept of Redis or Memcache never existed… and that nobody ever heard of JOIN performance problems. If it isn’t extreme social hazing, what is it?
your comments there are exceptionally aggressive. you accuse the developers of “hazing” you when they disagree with a technical decision you’ve made, and then insult them, posting wild rants. i’ve seen you do this many times. you need to step back, relax, and not take technical feedback so personally
To be fair, I think he has a point.
The way to convey it is very much debatable, but the poor database performance of Lemmy seems to be what puts LemmyWorld down so often.
As LW is the biggest instance so far, it's a bit concerning for the potential scaling of instances
agreed. but lashing out at developers who are just trying to find out why the db performance is subpar with accusations of “hazing” is incredibly aggravating and antithetical to the entire concept of FOSS. there is a reason why almost every open source organization uses codes of conduct
As I said, very debatable 😄
I've stepped back and watched them ignore the issue since May when all the servers were crashing. Every single Lemmy server was falling over while they ignored the PostgreSQL problems.
The mistakes are obvious and huge. These are not minor topics.
I think they are hazing the entire World Wide Web, Reddit users, etc. How else can you explain such basic SQL problems that they have allowed to go on for so long?
Its' as if the mere concept of Redis or Memcache never existed... and that nobody ever heard of JOIN performance problems. If it isn't extreme social hazing, what is it?
Thanks for sharing!
This sounds like it would be interesting if I could understand it lol
It’s not
It's really odd how many people around here think the server crashes are perfectly normal and are glad to see newcomers driven away.
They are perfectly normal. Unlike giant corporations, the people who run Lemmy don't have the money to support a fleet of failover servers that take over when the main server goes offline. That's basically the only reason you don't see lots of downtime from major corporations: investment in redundancy, so when something breaks, a perfect copy takes over. Server crashes happen all the time for major corporations, you just never see them due to investment in redundancy.
That's the difference between a community and a company. One takes actual investment from the community as a whole, and the other ruthlessly exploits for profit.
That has nothing to do with the issue I'm talking about. Every server with the amount of data in them would fail. Doesn't matter if you had 100 servers on standby.
The Rust logic for database access and PostgreSQL logic in lemmy is unoptimized and there is a serious lack of Diesel programming skills. site_aggregates table had a mistake where 1500 rows were updated for every single new comment and post - and it only got noticed when lemmy.ca was crashing so hard they made a complete copy of the data and studied what was gong on.
Throwing hardware at it, as you describe, has been the other thing... massive numbers of CPU cores. What's needed is to learn what Reddit did before 2010 with PostgreSQL.... as Reddit also used PostgreSQL (and is open source).
Downtime because you avoid using Redis or Memcached caching at all costs in your project isn't common to see in major corporations. But Lemmy avoids caching any data from PostgreSQL at all costs. Been that way for several years. May 17, 2010: "Lesson 5: Memcache;"
As I said in my very first comment, server crashing as a way to scale is a very interesting approach.
EDIT: Freudian slip, "memecached" instead of Memcached
That's a much more... coherent explanation than your original one, friend. I wouldn't have argued this point if you had started here.
If anyone bothered to actually look at the SQL SELECT that Lemmy uses to list posts every time you hit refresh it would be blindingly obvious how convoluted it is. yet the community does not talk about the programming issues and instead keeps raising money for 64 core hardware upgrades without recognizing just how tiny Lemmy's database really is and how 57K users is not a large number at all!
I mentioned "ORM" right in my first comment.
Damn, so many joins :/
How could this monster be optimized though?
First optimization is to not fetch every field and prune it down. For example, it gets public key and private key for every community and user account - then does nothing with them. That's just pushing data between Rust and PostgreSQL for no reason. That kind of thing is pretty obvious.. the huge number of things listed after "SELECT".
The whole approach is what I recently described as: make a JOIN fusion implosion bomb, then wait for null columns to fall out
There are short-term and long-term solutions. Right now there is already a new feature that will add one more JOIN that is pending merge.... "instance blocking" by each single user.
Based on the server overloads and resulting crashes, I think some obvious solutions would be to remove post_aggregates table entirely and just throw more columns on the post table... I've seen people do stuff like that. But really you have to have a concept of core foundation.
To me the core foundation of Lemmy data is that people want fresh meat, when world events get into a frenzy, they want to F5 and get the LATEST post and the LATEST comments. Data should have a big wall between the most recent 5 days and everything else. It's the heart of the beast of human events and a platform like this.
From that perspective, that fresh posts and fresh comments mean everything, you can optimize by just doing a INNER SELECT before any JOIN... or partition the database TABLE into recent and non-recent, or some out-of-band steps to prepare recent data before this SELECT even comes up from an API call... and not let PostgreSQL do so much heavy lifting each page refresh.
If I remember, I'm gonna look into that tomorrow when I'm not on a phone screen. Not that I could contribute anything, but this seems like a good opportunity to learn some advanced stuff. Thanks for your answer!
Community stuff can work well if done right. For example you don't see Debian repositories constantly crashing.
I don't follow your comment, are you suggesting I said something negative about open source project communities? I was talking about the Lemmy social media communities who actually comment and fund the 64-core server upgrades without asking why the site crashes with only 57K users.... the people who comment and post on Lemmy.... not the "open source" programmer community, but the social media community of Lemmy.