|
Blinkz0rz posted:well constructed golden images were the pinnacle of server management and i'll fight anyone who disagrees I blame microservices (and I'm not sure if that's good or bad, tbh) Good Sphere posted:no, what are the benefits? the benefits of react native are that you can smear javascript all over your mobile app like a shunted turd
|
# ? Dec 13, 2018 22:23 |
|
|
# ? Apr 24, 2024 15:40 |
|
Finster Dexter posted:I blame microservices (and I'm not sure if that's good or bad, tbh) pah, microservices are old news. we're migrating all our code to macroservices, which is where you run excel in k8s
|
# ? Dec 13, 2018 22:42 |
|
I think you mean run k8s in excel
|
# ? Dec 13, 2018 22:45 |
|
Good Sphere posted:Dear Journal: today, i was a horrible horrible programmer. i did not get anything accomplished. just total frustration trying to get UIViews to animate correctly i would argue that this is at worst neutral programming. no bad code was put into the world.
|
# ? Dec 13, 2018 23:35 |
|
i'm so sick of orm's and their lovely rear end query builders, i want to write my own sqlalchemy in something typesafe
|
# ? Dec 13, 2018 23:43 |
|
don't use a orm!
|
# ? Dec 13, 2018 23:44 |
|
sql loving owns but why does everyone hate it and want you to write in their own imbecile knockoff variant in which you can't actually do anything useful? i don't want to write raw sql either, i just want a metaprogramming language for sql
|
# ? Dec 13, 2018 23:46 |
|
if you're working in java check out jooq, it has what you want (and also other orm-ish stuff you don't have to use)
|
# ? Dec 13, 2018 23:49 |
|
Finster Dexter posted:Honestly, this probably accounts for a lot of its popularity. It boggles my mind how people continue to underestimate the critical importance of documentation, examples and good error messages. Especially for early adoption.
|
# ? Dec 14, 2018 00:01 |
|
i have never java'd in my life but i've been thinking it's probably time to learn a new language
|
# ? Dec 14, 2018 00:01 |
|
Ploft-shell crab posted:yes, that’s exactly the point. instead of ops managing several different kinds of pets, it manages just one kind of pet. so in the interest of consistent environments (which we already have because chef/ansible/salt are a thing) ops is now on the hook for making sure that a set of beautiful pet k8s masters and nodes can never ever go down. tell me again how that's better than golden images with sensible autoscaling and instance termination on app failure? Shaggar posted:y do you even have servers. just deploy to azure paas. lmao, hush shaggar, the adults are talking
|
# ? Dec 14, 2018 00:08 |
|
Finster Dexter posted:I blame microservices (and I'm not sure if that's good or bad, tbh) i blame everyone wishing really hard that they were google or netflix, cargo culting like crazy, and forcing sensible people into chasing buzzwords to enable idiots to keep going fast and breaking things
|
# ? Dec 14, 2018 00:11 |
|
I would much MUCH instead use Docker and K8s and manage microservices than have a vast monolithic application that has a million lines of code that nobody can handle adequately. Yeah, there are growing pains, but in the end, quality goes up, and that fart app doesn't eat an extra gig of memory a day because somebody tried to add a wet-fart function and hosed up a function ten inheritance levels deep. Blinkz0rz posted:I blame everyone wishing hard that they were Google or Netflix, cargo culting like crazy, and forcing sensible people into chasing buzzwords to enable idiots to keep going fast and breaking things I agree with this sentiment. Starting with a monolithic application (and keeping the idea of turning it into microservices later on in the back of your head), and then breaking it up when the time is right is the way to go IMHO. If you start with microservices, you are in for a world of hurt. Edit* Docker is cool and good for getting developers up and running quickly. Give them a Dockerfile and a docker-compose.yml and in a few minutes, they should have a development environment ready to go. FlapYoJacks fucked around with this message at 00:18 on Dec 14, 2018 |
# ? Dec 14, 2018 00:12 |
|
there's a middle ground where you can build microservices and not use docker or kubernetes and still get consistent, reproducible builds of both the app and the environment it does take a level of discipline that very few companies have though
|
# ? Dec 14, 2018 00:17 |
|
Blinkz0rz posted:there's a middle ground where you can build microservices and not use docker or kubernetes and still get consistent, reproducible builds of both the app and the environment Docker is nice though because updating is easy as gently caress. Step 1) Build Step 2) Push image to a docker repo Step 3) Pull new image
|
# ? Dec 14, 2018 00:19 |
|
ratbert90 posted:Docker is nice though because updating is easy as gently caress. golden image is nice because step 1) build step 2) push jar/war/whatever the gently caress to nexus/artifactory step 3) have packer bake a new ami it's the exact same poo poo
|
# ? Dec 14, 2018 00:25 |
|
react native’s killer feature is hot reload, which lets you modify your app while it’s running and have it update instantly this is very important when using a framework that basically only supports programming by trial and error and loving around until things appear to work
|
# ? Dec 14, 2018 00:33 |
|
i'll goog around of course, but while i'm at it, would anyone like to explain "microservices" to someone who's effective been out of the loop for ten years? like, where's the micro-, what's the line, what does a single microservice, er, serve, what's so special about em?
|
# ? Dec 14, 2018 00:33 |
|
Ciaphas posted:i'll goog around of course, but while i'm at it, would anyone like to explain "microservices" to someone who's effective been out of the loop for ten years? I believe canonical source on that whole thing is Martin Fowler.
|
# ? Dec 14, 2018 00:43 |
|
the main goal is to not have seperate apis in the same app i.e. in a non-micro service you'd have a ButtService and as IdiotService for your two different customer needs. but nothing actually links butts and idiots, there's no actual reason they need to be in the same codebase and deployment etc etc. in fact having them together increases the risk that they interfere with each other's code and db and poo poo. so you instead have ButtApp which handles only butts and IdiotApp which handles only idiots. you have monoliths (literally all the company's poo poo is in a single server process) at one and and full micro "literally only has one endpoint" services on, much like my posting, a spectrum. people are rarely advocating for going fully one or the other but i hope this at least gives you an idea for what people mean when talking about microservices. basically just splitting up code bases.
|
# ? Dec 14, 2018 00:47 |
|
as usual the answer is far simpler than all the gravitas that these buzzwords and marketing interest gave it made it seem thanks for the summary, i'll read the fowler link tonight (but i get the feeling i already know some 75% of the story now and the rest is just a fancy label)
|
# ? Dec 14, 2018 00:53 |
|
you can then do things like: even though butts and farts ARE related, you have two different apps that either 1) share a db (wouldn't recommend this) or more likely the butts app exists on its own and then the farts app calls the butt app. so instead of again two classes in the same app you are enforcing that they only touch in a small and well defined way. the main downside with this stuff is that the more separation you enforce the more fiddly it becomes when you do want two different things to place nicely together. plus deploying 100 things vs 1 thing. i mean if you go full micro (I've seen some people refer to this as nano) you can then have add butts be a different app to delete butts or whatever but I've not seen anybody actually do this and it kind of just feels like a strawman slippery-slope type argument. docker (and similar stuff like vms) helps microservices because you can pretend to give each app its own filesystem so they can't interact there either. docker is much less overhead as discussed earlier since you're still sharing the host's kernel stuff you're more pretending you're isolated than actually being actually isolated kubernetes (and other orchestrators) helps microservices because it makes deploying, updating them, scaling, etc less fiddly - none of this "ssh to the server, kill the pid (or running docker container), start the new version" poo poo
|
# ? Dec 14, 2018 01:23 |
|
Blinkz0rz posted:so in the interest of consistent environments (which we already have because chef/ansible/salt are a thing) ops is now on the hook for making sure that a set of beautiful pet k8s masters and nodes can never ever go down. lol, take off the big boy pants and put away all your toy languages and then maybe you can come out and play.
|
# ? Dec 14, 2018 01:25 |
|
gonadic io posted:the main goal is to not have seperate apis in the same app optional galaxy brain version as followed by our developers in the past: have separate components and services for specific functions but then compile a dll of a load of stuff you think you might need even though it's 90% irrelevant and then reference that in every single instance as a direct dependency. also just use whatever version you have at the time, don't bother recompiling it ever
|
# ? Dec 14, 2018 01:35 |
|
Powerful Two-Hander posted:optional galaxy brain version as followed by our developers in the past: have separate components and services for specific functions but then compile a dll of a load of stuff you think you might need even though it's 90% irrelevant and then reference that in every single instance as a direct dependency. also just use whatever version you have at the time, don't bother recompiling it ever truly i had not considered a system with all the convenience of microservices and all the code separation of a monolith
|
# ? Dec 14, 2018 01:48 |
|
ratbert90 posted:I would much MUCH instead use Docker and K8s and manage microservices than have a vast monolithic application that has a million lines of code that nobody can handle adequately. What if my vast monolithic application that has a million lines of code is managed in a docker compose with all kinds of poo poo containers such as an active directory container for the PHP application, That is so loving complicated new developers have to set up various env values before they run docker-compose build if not it all fucks up, But is ultimately NOT deployed on AWS with docker but instead some other poo poo ops decided to use so basically Dev Environment is not the same as staging Environment is not the same as production environment?
|
# ? Dec 14, 2018 01:59 |
|
Blinkz0rz posted:so in the interest of consistent environments (which we already have because chef/ansible/salt are a thing) ops is now on the hook for making sure that a set of beautiful pet k8s masters and nodes can never ever go down. please explain how your k8s workers are anything more than disposable resource piles that you grind up on a whim actually don't, i don't care, i just wanted to say disposable resource pile
|
# ? Dec 14, 2018 02:27 |
|
Blinkz0rz posted:so in the interest of consistent environments (which we already have because chef/ansible/salt are a thing) ops is now on the hook for making sure that a set of beautiful pet k8s masters and nodes can never ever go down. if you live in a world where you have stateless components and can easily scale up in an asg, then by all means go that route. I would argue that baking containers is preferable to baking images, but it doesn’t really matter that much there are plenty of operational environments that are not cloud or are stateful where you can’t just let aws magic take care of everything. if you have to have pets, there’s no reason why those pets shouldn’t be k8s
|
# ? Dec 14, 2018 02:28 |
|
Ciaphas posted:i'll goog around of course, but while i'm at it, would anyone like to explain "microservices" to someone who's effective been out of the loop for ten years? well people were starting to figure out how to make a webshit backend. and they realized that there were problems sometimes, or that if they use the right tone of voice they can make things sound like problems at least microservices solves this problem by getting everyone to do webshit backends in a different way that no one is good at so once again there is lots of low-hanging fruit to pick and blog posts to write
|
# ? Dec 14, 2018 02:34 |
|
DELETE CASCADE posted:if you're working in java check out jooq, it has what you want (and also other orm-ish stuff you don't have to use) jooq owns and everything we've deployed in the last 8 years has jooq in it. code generation is awesome.
|
# ? Dec 14, 2018 02:48 |
|
Ploft-shell crab posted:if you live in a world where you have stateless components and can easily scale up in an asg, then by all means go that route. I would argue that baking containers is preferable to baking images, but it doesn’t really matter that much fair
|
# ? Dec 14, 2018 03:46 |
|
your pets should be cats imo
|
# ? Dec 14, 2018 03:59 |
|
cats are good as hell
|
# ? Dec 14, 2018 04:19 |
|
k8s falls down real bad with storage. its tolerable if you just need some EBS but i'm not going to try anything more complicated i.e. DBs in k8s. in practice it works fine anyway to maintain a base AMI and let DB instances take 10 minutes to enter service after launch, so just use terraform/CF and chef/puppet/ansible for DBs. dunno how ansible is supposed to work with ASG though
|
# ? Dec 14, 2018 04:29 |
|
Captain Foo posted:your pets should be cats imo what an awful thing to say
|
# ? Dec 14, 2018 04:35 |
|
AggressivelyStupid posted:cats are good as hell
|
# ? Dec 14, 2018 05:08 |
|
catching up, herePleasureKevin posted:hey everyone, faked my death and moving to japan in glorious Japan, convenience store food is good and Pleasure Kevin is, I hope, well Jabor posted:Implying someone works at the turd-shunting yard is a zero-cost insult, so it's not surprising that it just got optimized away in hackbunny's mental model lol Shaggar posted:lol, take off the big boy pants and put away all your toy languages and then maybe you can come out and play spotted: the rare horny shaggar Kevin Mitnick P.E. posted:microservices solves this problem by getting everyone to do webshit backends in a different way that no one is good at so once again there is lots of low-hanging fruit to pick and blog posts to write a classic from kevin mitnick, posting engineer
|
# ? Dec 14, 2018 06:21 |
|
whew, think im finally starting to get the hang of tex scoping & expansion, a littlecode:
Carthag Tuek fucked around with this message at 06:51 on Dec 14, 2018 |
# ? Dec 14, 2018 06:24 |
|
Kevin Mitnick P.E. posted:k8s falls down real bad with storage. its tolerable if you just need some EBS but i'm not going to try anything more complicated i.e. DBs in k8s. in practice it works fine anyway to maintain a base AMI and let DB instances take 10 minutes to enter service after launch, so just use terraform/CF and chef/puppet/ansible for DBs. dunno how ansible is supposed to work with ASG though The CTO here (who worships k8s and kelsey hightower) spent a week fiddling with getting a basic postgres db set up in k8s. After finally being blocked unless I had db in aws, I just spun up an RDS instance for each cluster because jfc you aren't kidding about storage being a garbage fire on k8s. That's my k8s storage story I guess
|
# ? Dec 14, 2018 06:25 |
|
|
# ? Apr 24, 2024 15:40 |
|
in other news, I'm falling in love with MATLAB. The language is awful and really encourages making write-once monstrosities even though it provides better ways for organizing code, some of which pay off and make you look smart to the people who only ever write scripts. I like using their extremely robust standard library to reimplement a single function from a toolbox that costs $1000. I like chewing on physicsy data and doing stats or signal processing poo poo and I get paid for that now. I am the idiot hellfucker.
|
# ? Dec 14, 2018 06:29 |