Post 2357831 by luckyp3616@pl.smuglo.li
 (DIR) More posts by luckyp3616@pl.smuglo.li
 (DIR) Post #2346736 by p@freespeechextremist.com
       2018-12-27T21:50:34.647557Z
       
       7 likes, 0 repeats
       
       Now that I'm trying to debug this memory deal, it goes away.  Hilarious.Here's a thing I did notice, though:  if it doesn't anger the DB too badly, the overwhelming majority of requests (upwards of 95%) complete in under 100ms on this bottom-tier VPS box.  Good luck getting response times out of that from a Rails application like Mastodon.  Even in prod mode, you're lucky if text/plain "Hello, World!" gets under 100ms.  (Seriously, Rails adds an order of magnitude performance hit, even when stripped down to the minimum with no DB involvement:  https://github.com/luislavena/bench-micro#requestssec )Pleroma's nice.  Let's all love @lain and @kaniini and the other devs.
       
 (DIR) Post #2346817 by luckyp3616@pl.smuglo.li
       2018-12-27T21:53:22.686192Z
       
       1 likes, 0 repeats
       
       @p @lain @kaniini Huh. Maybe if it happens again we could find out the URL they use, and @lain can then patch up this major DoS vuln.
       
 (DIR) Post #2346948 by p@freespeechextremist.com
       2018-12-27T21:58:28.909362Z
       
       1 likes, 0 repeats
       
       @PussySlayer @lain @kaniini I hate using Rails; it's bad software.  Overall, I'm ambivalent:  Ruby's nice, and it'd be harder to get Ruby work without Rails, but Rails has also done more damage to Ruby than good, and pushed the language into Eternal September.
       
 (DIR) Post #2347064 by p@freespeechextremist.com
       2018-12-27T22:02:47.480756Z
       
       1 likes, 0 repeats
       
       @luckyp3616 @lain @kaniini @lain Well, like I said, just a possibility.  I've got a suspicion.  But yeah, that's what I was trying to do, figure out some sort of association between the path it's hitting (or the user or IP that was hitting it) and response times, while watching top to see memory climb.  Memory stopped climbing, though.Sometimes this is a dead-end anyway.  I was using a brute-force approach to debugging, because I'm not super familiar with Elixir.  If there's a good way to do instrumentation and monitor which endpoints do the most memory allocation, that'd be better.  The way I'm doing it is basically jiggling the wires.
       
 (DIR) Post #2347182 by luckyp3616@pl.smuglo.li
       2018-12-27T22:06:33.795090Z
       
       0 likes, 0 repeats
       
       @p @lain @kaniini @lain Huh.
       
 (DIR) Post #2347192 by m0n5t3r@gs.m0n5t3r.info
       2018-12-27T22:06:59+00:00
       
       2 likes, 1 repeats
       
       @p https://en.wikipedia.org/wiki/Heisenbug
       
 (DIR) Post #2347211 by p@freespeechextremist.com
       2018-12-27T22:08:20.349896Z
       
       1 likes, 0 repeats
       
       @m0n5t3r Ha, yes.  In this case, I don't think tailing the nginx logs could have Heisenbug'd the BEAM process.  But it's always a good concern.
       
 (DIR) Post #2347502 by m0n5t3r@gs.m0n5t3r.info
       2018-12-27T22:18:20+00:00
       
       2 likes, 1 repeats
       
       @p try putting something like atop on the machine, it takes snapshots of system parameters once a minute (assuming the leak is not of the "fills 128G and swap in 10 seconds" kind), so you could check it and cross-reference with the logs  that, or get something like influxdb or grafana going for better resolution; my interactions with rabbitmq taught me erlang is fucking alien technology, it almost never fails, but when it fails there's no way to figure out why.
       
 (DIR) Post #2347859 by kaniini@pleroma.site
       2018-12-27T22:29:23.936982Z
       
       2 likes, 0 repeats
       
       @p @lain it's probably the thread reconstructor.  there's presently no circuit breaker
       
 (DIR) Post #2347929 by p@freespeechextremist.com
       2018-12-27T22:31:58.893551Z
       
       1 likes, 0 repeats
       
       @kaniini @lain :hankhell: Hellthreads :hankhell: do tend to take a minute, which I've noticed.  But a thread gets filled in piecemeal, doesn't it?  Maybe I'm hallucinating this, but I thought that you couldn't get the whole thread from a single request.
       
 (DIR) Post #2350384 by luckyp3616@pl.smuglo.li
       2018-12-27T23:49:58.752995Z
       
       1 likes, 0 repeats
       
       @p @m0n5t3r Ugh. Wow. Perhaps it isn't a hacker, but a weird version of quantum mechanics. Time to get Neil deGrasse Tyson on the case.
       
 (DIR) Post #2352752 by p@freespeechextremist.com
       2018-12-28T01:02:56.993054Z
       
       2 likes, 0 repeats
       
       @luckyp3616 @m0n5t3r I'd rather bring in Dick Feynman, that guy knew his way around a computer at least.
       
 (DIR) Post #2352777 by luckyp3616@pl.smuglo.li
       2018-12-28T01:03:27.657797Z
       
       0 likes, 0 repeats
       
       @p @m0n5t3r True.
       
 (DIR) Post #2353121 by p@freespeechextremist.com
       2018-12-28T01:16:41.604735Z
       
       1 likes, 0 repeats
       
       @PussySlayer @luckyp3616 @m0n5t3r  An update to this:  it's got nothing to do with any of that and is a string-gluing problem and there's a fix forthcoming and the problem should be fixed locally (for now).
       
 (DIR) Post #2355567 by luckyp3616@pl.smuglo.li
       2018-12-28T03:13:10.258057Z
       
       1 likes, 0 repeats
       
       @p @PussySlayer @m0n5t3r Ah, a string-related bug. Welp, I'm sure every dev has to debug at some point. Happens to every software package at some point. Especially with common bugs related to strings.
       
 (DIR) Post #2355623 by luckyp3616@pl.smuglo.li
       2018-12-28T03:15:17.505850Z
       
       1 likes, 0 repeats
       
       @p @PussySlayer @m0n5t3r Perhaps lower char limit until the fix is rolled out would be good for all instances.
       
 (DIR) Post #2356347 by p@freespeechextremist.com
       2018-12-28T03:45:51.092162Z
       
       0 likes, 0 repeats
       
       @luckyp3616 @PussySlayer @m0n5t3r Ah, in this case, right, our limit was already 1024 (now 2048).  But a message from another instance gets mirrored locally, right, so it's no use.
       
 (DIR) Post #2356447 by Johnny_of_the_swamp@noagendasocial.com
       2018-12-28T03:50:45Z
       
       1 likes, 0 repeats
       
       @p @lain @kaniini mastodon is trash Rails but this is BS. My day job is working on a Rails API at very large scale and we have tons of routes under 100ms with significant sql queries. The problem is shitty Rails devs.
       
 (DIR) Post #2356900 by fashywhitefem@baraag.net
       2018-12-28T04:12:10Z
       
       1 likes, 0 repeats
       
       @luckyp3616 @p Better get diversity astrophysicist to debug muh strings n’ beep boops.  (If you wanted the npc appropriate computer science expert you’d pick Touring for the fag-pass, Ida for the pink-card, or Hidden Numbers for multikulti if you’re truly married to the color.)
       
 (DIR) Post #2357623 by p@freespeechextremist.com
       2018-12-28T04:48:07.783929Z
       
       1 likes, 0 repeats
       
       @Johnny_of_the_swamp @kaniini @lain On a bottom-tier VPS, I'd be surprised if you got a response time like that.  In fact, on a bottom-tier VPS, I'd be surprised.If you think there's some shitty dev work going on in a trivial, one-file benchmark, I'm interested:  https://github.com/luislavena/bench-micro/blob/master/apps/rails-metal.ru .I've been working with Rails since the 0.9 release.  I've worked on some very large-scale Rails systems, too.  I know the problem with shitty Rails devs, because I'm the guy that cleans it up.  So I'm not talking out my ass, we don't have to drag out our credentials.I am, though, actually interested in what you have to say on this, because it is hard to find someone that knows what they're talking about in the vicinity of Rails, and you seem to know what you're talking about.
       
 (DIR) Post #2357669 by p@freespeechextremist.com
       2018-12-28T04:49:47.183517Z
       
       1 likes, 0 repeats
       
       @fashywhitefem @luckyp3616 Ida?  Ada?If we're doing diversity card, we do have :turing: here.  No Ada.  Grace Hopper is superior and the genius and I will fight anybody. :grace:
       
 (DIR) Post #2357742 by luckyp3616@pl.smuglo.li
       2018-12-28T04:53:25.238950Z
       
       0 likes, 0 repeats
       
       @fashywhitefem @p Not everything has to do with skin color. Sheesh.
       
 (DIR) Post #2357831 by luckyp3616@pl.smuglo.li
       2018-12-28T04:57:13.169295Z
       
       2 likes, 0 repeats
       
       @p @fashywhitefem Afa? That reminds me of that tranny bitch that is ruining Linux.
       
 (DIR) Post #2358173 by fashywhitefem@baraag.net
       2018-12-28T05:09:42Z
       
       1 likes, 0 repeats
       
       @p @luckyp3616 Thanks, how silly of me to disassemble her name like that, don’t know what I was thinking.
       
 (DIR) Post #2358193 by fashywhitefem@baraag.net
       2018-12-28T05:10:33Z
       
       1 likes, 0 repeats
       
       @p @luckyp3616 Who carried the ns length of copper wire or whatnot?
       
 (DIR) Post #2358245 by p@freespeechextremist.com
       2018-12-28T05:13:09.678229Z
       
       2 likes, 0 repeats
       
       @luckyp3616 @fashywhitefem Ha, in fact, the "Ada Initiative" was named after Ada Lovelace.  It is a dark joke that they named their organization after a rich honkey woman that never did any coding but sure talked about it a lot.
       
 (DIR) Post #2358312 by fashywhitefem@baraag.net
       2018-12-28T05:15:42Z
       
       0 likes, 0 repeats
       
       @luckyp3616 @p You’re right, not everything is about race, but the NDT tokenism is entirely about race, on your part.  I just supplied the list of generally approved names you should have dropped, since you tripped into the faux pas of the oversell.They do say that it’s forbidden to criticize the glorious leader, but it’s even more forbidden to mention that it’s forbidden.  Don’t worry, you can resume your generally clumsy conversation unmolested.  Ciao.
       
 (DIR) Post #2358332 by p@freespeechextremist.com
       2018-12-28T05:16:52.331180Z
       
       0 likes, 0 repeats
       
       @fashywhitefem @luckyp3616 Indeed.  She kept at her desk a length of rope that represented a microsecond (983'6"), and suggested it be tied around the necks of programmers that forgot how long it was.
       
 (DIR) Post #2358442 by fashywhitefem@baraag.net
       2018-12-28T05:21:55Z
       
       1 likes, 0 repeats
       
       @p @luckyp3616 Yeah, I watched a video of a talk where she pulled a similar prop, only made of copper, from her bag.  That’s how the name got lodged in there, and then there were an insufferable number of Asian girls who played trophy for the skids borrowing the name.  I was specifically aiming for the approved list.
       
 (DIR) Post #2358530 by luckyp3616@pl.smuglo.li
       2018-12-28T05:28:40.332438Z
       
       0 likes, 0 repeats
       
       @fashywhitefem @p Nobody talks about the engineers back in the 1800s and 1900s that lead up to the modern hardware. Most of them happened to be of European descent.
       
 (DIR) Post #2358777 by fashywhitefem@baraag.net
       2018-12-28T05:39:24Z
       
       0 likes, 0 repeats
       
       @luckyp3616 @p Take that Jared “happens to be white” Diamond apologia and piss right off.  It’s hard to imagine how someone could be so ungrateful.  You have an artifact that can literally transform logic into action and you’ve turned it into a vehicle to shit on its creators.  Disgusting.
       
 (DIR) Post #2358823 by p@freespeechextremist.com
       2018-12-28T05:41:31.151835Z
       
       0 likes, 0 repeats
       
       @fashywhitefem @luckyp3616 She used to bring out little copper wires, nanoseconds.She's the first person to enlist the help of the computer in programming itself, essentially creating the first primitive compiler.  Cutesy shit about nanoseconds aside, that was a major step forward.
       
 (DIR) Post #2358840 by p@freespeechextremist.com
       2018-12-28T05:42:12.340736Z
       
       0 likes, 0 repeats
       
       @luckyp3616 @fashywhitefem Everyone talks about :babbage:.
       
 (DIR) Post #2358897 by fashywhitefem@baraag.net
       2018-12-28T05:45:24Z
       
       1 likes, 0 repeats
       
       @p @luckyp3616 https://youtu.be/JATjMZJuX1c
       
 (DIR) Post #2358989 by p@freespeechextremist.com
       2018-12-28T05:51:37.877042Z
       
       0 likes, 0 repeats
       
       @fashywhitefem @luckyp3616 :brandt:
       
 (DIR) Post #2369680 by Johnny_of_the_swamp@noagendasocial.com
       2018-12-28T15:35:57Z
       
       1 likes, 0 repeats
       
       @p @kaniini @lain Honestly, I'm not sure. Just a vanilla Rails 5 app on my 2015 Macbook will serve a Devise backed login page in sub 10ms and most of that is spent on rendering the erb file. If I had to wager a guess as to what is going on here, it might be related to using Puma. I'm not sure how this ever became popular outside of the JRuby world. Real Chads use Unicorn. You'd really would need to just jam in ruby-prof here and figure out what the hell it is wasting time on.
       
 (DIR) Post #2369689 by kaniini@pleroma.site
       2018-12-28T15:36:37.452872Z
       
       1 likes, 0 repeats
       
       @Johnny_of_the_swamp @lain @p untag pls
       
 (DIR) Post #2394630 by p@freespeechextremist.com
       2018-12-29T09:13:46.672781Z
       
       1 likes, 0 repeats
       
       @Johnny_of_the_swamp Yes, I use unicorn and rainbows as well.  They were written by a colleague.  I don't know if EW's still maintaining those projects, but he's a good coder.  His blog is now email:  https://80x24.org/On a 2015 laptop, you can get reasonable response times, but probably not on a single-core VPS with next to no RAM, and the actual times don't matter, just the times relative to the other frameworks, other languages, etc.  I don't think the benchmark in this case is showing anything out of line with what I've seen in the wild.  It uses puma/rack as a baseline, and anything else you use is going to be overhead on top of rack, unless you're deploying to WEBrick or something.
       
 (DIR) Post #2408162 by Johnny_of_the_swamp@noagendasocial.com
       2018-12-29T19:28:54Z
       
       1 likes, 0 repeats
       
       @p yea, my experience does not touch on small VPSs. More like beefy EC2 and Xen hosts. Gets down to the classic tradeoff of efficiency vs developer productivity. 2018 was the year I got serious about Golang and it is wonderfully fast but it is much easier to get a juniorish developer writing Ruby than it is Go.
       
 (DIR) Post #2419335 by p@freespeechextremist.com
       2018-12-30T01:37:39.315246Z
       
       0 likes, 0 repeats
       
       @Johnny_of_the_swamp I really like Go.  :pike: and :ken: have very good taste in languages.  I've actually written a lot of Limbo, which, if you remove the semicolons and replace 'spawn' with 'go', is nearly the same language.  (Nearly!)  But same creators, so not a huge surprise.I think you can make Ruby fast, but I don't think you can make Rails quite as fast.  The TL;DR is that I think the framework overreaches and in some places it picks entirely the wrong abstraction.  I wanted to throw up when DHH said "omakase", and I don't have a lot of confidence in Yehuda Katz, having interacted with him some.The main issue I have with Go is that the tooling is a little overblown.  For small stuff, I really miss some of the string-mangling that Ruby's so good at.I have another issue with Go, though.  It tends to work so nicely and everything behaves on the first try.  Although it's a productive language, the result is that a larger proportion of my time is spent debugging JavaScript because the Go stuff just works.