awful programming: feature flags and suffering

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > awful programming: feature flags and suffering

«‹›1113 »

redleader: Aug 18, 2005; Engage according to operational parameters

elasticsearch is hipster poo poo. just use a real rdbms

# ? Dec 24, 2018 06:36

Adbot: ADBOT LOVES YOU

# ? Apr 18, 2024 04:26

Luigi Thirty: Apr 30, 2006; Emergency confection port.

redleader posted:

elasticsearch is hipster poo poo. just use a real rdbms

like Oracle op

# ? Dec 24, 2018 06:54

abigserve: Sep 13, 2009; this is a better avatar than what I had before

Kevin Mitnick P.E. posted:

i would fight tooth and nail at work to use elasticsearch over timescaledb unless elasticsearch was totally unfit. that's down to us having decent experience running ES and 0 with timescaledb or even postgresql. learning how to operate a new db and dealing with its poo poo when you put a nontrivial workload on it (i.e. more than fits on a dev's laptop) is extremely expensive in man-hours

does your infra team get a lot of suggestions to implement DBs that reached 1.0 less than two months ago?

Yeah in prod use whatever you are most comfortable with. Timescale is good because I reckon more people are familiar with postgres over a document store like elasticsearch

# ? Dec 24, 2018 07:54

Powerful Two-Hander: Mar 10, 2004; Mods please change my name to "Tooter Skeleton" TIA.

Kevin Mitnick P.E. posted:

i would fight tooth and nail at work to use elasticsearch over timescaledb unless elasticsearch was totally unfit. that's down to us having decent experience running ES and 0 with timescaledb or even postgresql. learning how to operate a new db and dealing with its poo poo when you put a nontrivial workload on it (i.e. more than fits on a dev's laptop) is extremely expensive in man-hours

does your infra team get a lot of suggestions to implement DBs that reached 1.0 less than two months ago?

more than you'd probably expect as we slowly catch up on technologies everyone else has had for years like elastic which only got "authorised" like a month ago, side effect is all sorts of weird poo poo getting used

anyway the one use case I was thinking of was a specific type of financial data query but now that I've sobered up a bit I remembered that their main problem was that they had more attributes that could be queried than elastic allowed as indexed properties rather than the old time range problem.

still, I'm totally going to go "hmmm, have you considered timescale db?" at some point to pretend I'm hip, in the mean time though:

redleader posted:

elasticsearch is hipster poo poo. just use a real rdbms

# ? Dec 24, 2018 11:01

Blinkz0rz: May 27, 2001; MY CONTEMPT FOR MY OWN EMPLOYEES IS ONLY MATCHED BY MY LOVE FOR TOM BRADY'S SWEATY MAGA BALLS

elasticsearch is fine just use it for storing logs

# ? Dec 24, 2018 15:20

Nomnom Cookie: Aug 30, 2009

elasticsearch is good if you need an inverted index and don�t need transactions or much consistency at all really. elastic is pushing it for all kinds of poo poo though. gotta take over the world before funding runs out

# ? Dec 24, 2018 15:53

Beamed: Nov 26, 2010; Then you have a responsibility that no man has ever faced. You have your fear which could become reality, and you have Godzilla, which is reality.

to be fair es had amazon totally pull the rug out from under em lol

# ? Dec 24, 2018 23:24

Nomnom Cookie: Aug 30, 2009

to be even fairer giving away your product for free and trying to make money from accessories is really dumb

# ? Dec 25, 2018 02:22

Schadenboner: Aug 15, 2011; by Shine

Kevin Mitnick P.E. posted:

to be even fairer giving away your product for free and trying to make money from accessories is really dumb

TBF, the business model has worked out pretty well for ur mum?

# ? Dec 25, 2018 02:31

Nomnom Cookie: Aug 30, 2009

Schadenboner posted:

TBF, the business model has worked out pretty well for ur mum?

that is extremely debatable

# ? Dec 25, 2018 02:47

Symbolic Butt: Mar 22, 2009; (_!_); Buglord

I solved every AOC problem in less than 24h this year! :toot:

# ? Dec 25, 2018 07:56

VikingofRock: Aug 24, 2008

Symbolic Butt posted:

I solved every AOC problem in less than 24h this year!

Congrats!

# ? Dec 25, 2018 09:59

FlapYoJacks: Feb 12, 2009

Symbolic Butt posted:

I solved every AOC problem in less than 24h this year!

Alexandria Ocazio-Cortez can solve her own problems thank you very much. :colbert:

# ? Dec 25, 2018 15:43

Finster Dexter: Oct 20, 2014; Beyond is Finster's mad vision of Earth transformed.

Blinkz0rz posted:

elasticsearch is fine just use it for storing logs

When I last used es, we used it as a full-text index for our nosql backend. It was pure dogshit in terms of uptime and was extremely unstable. Since then, I've only ever used it as part of elk and it still manages to be a huge pita.

# ? Dec 26, 2018 18:15

CRIP EATIN BREAD: Jun 24, 2002; Hey stop worrying bout my acting bitch, and worry about your WACK ass music. In the mean time... Eat a hot bowl of Dicks! Ice T; Soiled Meat

how in the world are you screwing up an es cluster?

how are you getting any sort of downtime?

# ? Dec 26, 2018 20:21

Blinkz0rz: May 27, 2001; MY CONTEMPT FOR MY OWN EMPLOYEES IS ONLY MATCHED BY MY LOVE FOR TOM BRADY'S SWEATY MAGA BALLS

CRIP EATIN BREAD posted:

how in the world are you screwing up an es cluster?

how are you getting any sort of downtime?

running it on aws, pumping hundreds of thousands of log-lines per second through it, then losing a node and/or having a shard/shards reallocate

es is great when it works but man it takes a lot to keep it operating as a near-realtime store for operational logging from a few hundred microservices + sidecar services

# ? Dec 26, 2018 23:03

Arcsech: Aug 5, 2008

Blinkz0rz posted:

running it on aws, pumping hundreds of thousands of log-lines per second through it, then losing a node and/or having a shard/shards reallocate

es is great when it works but man it takes a lot to keep it operating as a near-realtime store for operational logging from a few hundred microservices + sidecar services

if you're getting downtime from that i suspect you are substantially underprovisioned

this is one of the problems with es: it's actually pretty good at handling more steady-state load than it really should, then people get surprised when it gets a little bit more load and it falls over, even though they've been redlining it for months. sizing is hard

that plus the whole "the jvm gets fucky if you give it too much memory" thing, which effectively limits the size of each node, which is really easy to cut yourself with

# ? Dec 26, 2018 23:17

Blinkz0rz: May 27, 2001; MY CONTEMPT FOR MY OWN EMPLOYEES IS ONLY MATCHED BY MY LOVE FOR TOM BRADY'S SWEATY MAGA BALLS

yep that's exactly right. it's hard to tell what an appropriate capacity for an es cluster is based on volume/storage needs until you tip it over, adjust, tip it over again, and then keep adjusting. there are some ok tools for understanding shard allocation and replication but no real silver bullet for any of it.

# ? Dec 27, 2018 00:16

distortion park: Apr 25, 2011

my experience with ES is that it's good at what it does but it's really hard to use as a primary data store. As long as you can recreate the indexes from a more reliable source it's fine

# ? Dec 27, 2018 18:28

eschaton: Mar 7, 2007; Don't you just hate when you wind up in a store with people who are in a socioeconomic class that is pretty obviously about two levels lower than your own?

any idea why poll(2) would not actually return -1 with errno set to EINTR when a signal occurs? this is on NetBSD 8

I�m using an INFTIM (infinite) timeout, and it does return for activity on the file descriptor I care about

I�ve also verified my signal handler�s getting called, by having it write(2) a message to stderr from within the handler, so it�s not a matter of the signal mask blocking them or something

I feel like there�s something I�m just forgetting about this stuff, it�s making me wish for CFRunLoop�

# ? Dec 28, 2018 09:12

MrMoo: Sep 14, 2000

Try pollts instead?

# ? Dec 28, 2018 15:45

eschaton: Mar 7, 2007; Don't you just hate when you wind up in a store with people who are in a socioeconomic class that is pretty obviously about two levels lower than your own?

yeah, that�s going to be my next step

# ? Dec 28, 2018 21:22

hackbunny: Jul 22, 2007; I haven't been on SA for years but the person who gave me my previous av as a joke felt guilty for doing so and decided to get me a non-shitty av

eschaton posted:

any idea why poll(2) would not actually return -1 with errno set to EINTR when a signal occurs? this is on NetBSD 8

the default on BSD is to automatically resume syscalls interrupted by signals. make sure to use sigaction (relevant flag: SA_RESTART) instead of signal to set the signal handler

# ? Dec 29, 2018 03:58

eschaton: Mar 7, 2007; Don't you just hate when you wind up in a store with people who are in a socioeconomic class that is pretty obviously about two levels lower than your own?

it turns out that wasn�t the issue after all

after the call to poll (and then pollts) I had one branch I didn�t cover with a log, and of course that was where the hang actually was: poll indicated POLLOUT for my descriptor, so I called DNSServiceProcessResult on its associated DNSServiceRef, which then blocked in recv

I think I just need to solely request POLLIN for that descriptor

# ? Dec 29, 2018 04:53

eschaton: Mar 7, 2007; Don't you just hate when you wind up in a store with people who are in a socioeconomic class that is pretty obviously about two levels lower than your own?

that was almost it, I also had to invert the sense of the signal mask passed to pollts

all the docs just say �signal mask,� they don�t say whether the mask is to indicate signals to block or signals to allow; turns out in this case it means signals to block

with this my little �register the mDNS advertisements in this file� daemon will now reload when it gets SIGHUP like every other daemon

# ? Dec 29, 2018 05:35

Powerful Two-Hander: Mar 10, 2004; Mods please change my name to "Tooter Skeleton" TIA.

Well, pretty sure I've created a monster here. the mega generated form pages used to capture data in the app now allow an item Foo to be of a type called "orderedlist " that can then parent a list of IFoos of arbitrary length and type , that can themselves parent IFoos, and auto render an editor modal to add or edit existing items by storing a json serialisation of every item in the list in a hidde input that is passed to the editor and deserialised when needed

needless to say, the page is getting pretty loving large when you have a lot of stuff in the list

all this because usrs want to be able to make hundreds of changes before saving instead of saving as they go

# ? Dec 29, 2018 14:35

MrMoo: Sep 14, 2000

8am I found out this news via spider tingly sense. This is a NYSE relaunch (~9:36am) and yet their own system thinks it's trading as normal pre-open. Took them 1� hours to publish that NYSE wasn't actually trading.

# ? Dec 29, 2018 15:51

hackbunny: Jul 22, 2007; I haven't been on SA for years but the person who gave me my previous av as a joke felt guilty for doing so and decided to get me a non-shitty av

eschaton posted:

all the docs just say �signal mask,� they don�t say whether the mask is to indicate signals to block or signals to allow; turns out in this case it means signals to block

oh yes the mask blocks, "everyone knows this". think of signals as an abstraction of interrupts: there's a "mask" of pending signals (not literally a mask because signals carry a little more data than a simple pending/clear bit - see siginfo_t) and a per-thread (originally per-process) mask of blocked signals. the fact that pending signals are a mask is very important: if a signal is already pending, sending it again will have no effect. unless the OS implements sigqueue (I believe they all do nowadays) and the source of signals uses that instead of kill

# ? Dec 29, 2018 17:19

carry on then: Jul 10, 2010; by VideoGames
(and can't post for 10 years!)

what is it about awful.app users and the trump thread? like half the bug reports and feature requests come in from people who read basically only that

# ? Dec 29, 2018 20:02

DONT THREAD ON ME: Oct 1, 2002; by Nyc_Tattoo; Floss Finder

carry on then posted:

what is it about awful.app users and the trump thread? like half the bug reports and feature requests come in from people who read basically only that

they probably want to feel like they have power over something in this stupid world

# ? Dec 29, 2018 22:52

DONT THREAD ON ME: Oct 1, 2002; by Nyc_Tattoo; Floss Finder

i'm working through this years AOC in various languages and right now I'm doing one in haskell.

i've done a lot of FP since the last time I messed with it ages ago. the language is much smaller than i remember -- it's tiny compared to scala. feels much less powerful in some ways, specifically around typeclasses.

anyhow haskellers why isn't this doing what I want:

code:


type GuardId = Int
data Shift = Shift GuardId [LogEntry] deriving (Show)

groupById :: Shift -> Shift -> Bool
groupById (Shift a _) (Shift b _) = a == b

*Day4> let shifts = [(Shift 1 []), (Shift 2 []), (Shift 1 [])]
*Day4> groupBy groupById shifts
[[Shift 1 []],[Shift 2 []],[Shift 1 []]] // Why isn't this [ [Shift 1 [], Shift 1 [] ], [Shift 2 [] ] ]  ???

e: oh, it needs to be sorted first

double edit: also parsec is cool. way better than regex.

DONT THREAD ON ME fucked around with this message at 23:19 on Dec 29, 2018

# ? Dec 29, 2018 23:00

Symbolic Butt: Mar 22, 2009; (_!_); Buglord

I bet you'll have fun with day 8 in haskell

# ? Dec 29, 2018 23:10

DONT THREAD ON ME: Oct 1, 2002; by Nyc_Tattoo; Floss Finder

nice maybe i'll skip to that one

# ? Dec 29, 2018 23:21

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

carry on then posted:

what is it about awful.app users and the trump thread? like half the bug reports and feature requests come in from people who read basically only that

i have no doubt in my mind that half the forums traffic is the trump thread

# ? Dec 29, 2018 23:33

pokeyman: Nov 26, 2006; That elephant ate my entire platoon.

Krankenstyle posted:

i have no doubt in my mind that half the forums traffic is the trump thread

I think it�s this plus the trump thread exercises some pathological usage patterns e.g. having a million twitter embeds in a single post

# ? Dec 30, 2018 01:11

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

yea ive noticed that taxes the app a bunch lol

ps thx for the app again

# ? Dec 30, 2018 02:17

DONT THREAD ON ME: Oct 1, 2002; by Nyc_Tattoo; Floss Finder

Symbolic Butt posted:

I bet you'll have fun with day 8 in haskell

you were right, this was a good challenge. it obviously wants you to use recursion, but it's not super straightforward. here's my solution if anyone has tips:

https://gist.github.com/daviswahl/84068630a48174374ee52f5a01409ad4

# ? Dec 30, 2018 02:29

DaTroof: Nov 16, 2000; CC LIMERICK CONTEST GRAND CHAMPION
There once was a poster named Troof
Who was getting quite long in the toof

DONT THREAD ON ME posted:

i'm working through this years AOC in various languages and right now I'm doing one in haskell.

Symbolic Butt posted:

I bet you'll have fun with day 8 in haskell

hey, this is fun. thx for bringing it up, first time i've seen it

i'm doing them in ruby cuz im a douchebag

# ? Dec 30, 2018 09:07

urea: Aug 18, 2013

DONT THREAD ON ME posted:

you were right, this was a good challenge. it obviously wants you to use recursion, but it's not super straightforward. here's my solution if anyone has tips:

https://gist.github.com/daviswahl/84068630a48174374ee52f5a01409ad4

day 8 is basically designed for parser combinators

code:

num = A.decimal <* (void A.space <|> A.endOfInput)

parseTree = do
  numChild <- num
  numData <- num
  flip Node <$> A.count numChild parseTree <*> A.count numData num

# ? Dec 30, 2018 09:19

Adbot: ADBOT LOVES YOU

# ? Apr 18, 2024 04:26

gonadic io: Feb 16, 2011; >>=

DONT THREAD ON ME posted:

i'm working through this years AOC in various languages and right now I'm doing one in haskell.

i've done a lot of FP since the last time I messed with it ages ago. the language is much smaller than i remember -- it's tiny compared to scala. feels much less powerful in some ways, specifically around typeclasses.

anyhow haskellers why isn't this doing what I want:

e: oh, it needs to be sorted first

double edit: also parsec is cool. way better than regex.

Groupby is really dumb (as in, implemented simply and isn't very powerful, check out its source to see it). If you want arbitrary bags by a given key you need Data.Map.fromList (with a custom Ord instance) or like you say sorting it so the keys are all adjacent.

Parsec is great yes it wasn't the first but imo it brought a resurgence of parser combinators into the industry now most languages have them. gently caress yacc and Co and especially gently caress regex

gonadic io fucked around with this message at 10:18 on Dec 30, 2018

# ? Dec 30, 2018 09:55

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > awful programming: feature flags and suffering

«‹›1113 »