hear me out bro
instead of creating one app
you make 100 apps
and the connect them with apis
microservices brah that's the future brah
 It's All Fucked Shirt $22.14 |
 |
It's All Fucked Shirt $22.14 |
hear me out bro. instead of creating one app. you make 100 apps. and the connect them with apis
hear me out homosexual put the gun down let me pull the trigger
fr bro this boomer be posting some cring sush fr fr microservices are sus bro
>instead of making one hard to maintain complex app that will be a headache to replace in future.
>build small blocks that are simple to maintain and can be swapped out one by one in future.
>microservices brah not repeating the mistakes of boomers.
>muh future
makes sense for a highly trafficked website doesn't make sense for a startup but hr roasties think is the next bitcoin so you must turn yourself into an API master do you even apis bro you gotta dream with apis bro send the api headers to your dreams bro
Just because you're too moronic to grasp API's doesn't make them a bad idea.
Just because you're too moronic to grasp the downsides of making everything a distributed system doesn't mean it's a good idea.
Everything has downsides.
Your spaghetti code has far more downsides that's why nobody will hire you.
Nice empty response, webshitter.
is not rocket science moron it's just some fricking apis but managing a hundred of them its mechanical tiresome work and is completely worthless
fricking web devs think that understanding APIs is like some next level shit, something you need a university degree for.
Bonus points if you understand REST, dude every single protocol spec is 10 times more difficult than the entirety of REST, and you morons don't even get that right.
>simple to maintain
HAHAHA
AHAHAHAHAHAHAHAHA
I'm sorry you're moronic enough to have drank the webshit kool-aid.
What's hard to maintain about some Express.js code?
I'm not talking about any particular framework but the concept of splitting up your app into dozens of microservices. I mentioned webshit because it seems popular there. Many midwit programmers have fallen for the hype and believe that microservices will automatically make their code easier to understand and maintain, when the opposite is more often the case. In a monolith, you just call functions in other parts of the program and write everything to a detailed log file which you can read if anything goes wrong. With microservices, now you need to have timeout/retry logic, as well as network failure handling, everywhere in your code. You'll need to correlate a dozen log files with one another just to know how a single request was handled. And you won't even know which event came after which because clocks can drift relative to each other so they're reliable within one computer but not across computers. You can solve it by using stuff like vector clocks and a centralized logging servers, but this introduces even more complexity and potential failure points. I can't believe how many midwits fell for the microservices hype, which is essentially the claim that a team of programmers which can't write a good monolith will somehow be able to write a good distributed system even though distributed systems are known to be harder. Absolute bullshit.
Ok makes sense. It seems like the only benefit then is that you save money by scaling horizontally across different machines rather than the specs of one machine (which wouldn't practical if you're Twitter for example)
The logging sounds like a pain in the ass I hadn't thought about that before. Have you tried stuff like AWS Cloud Watch or other log aggregation services? Obviously I'm not big on relying on a cloud provider to do my logging but still
Yeah, I'm not always against separating out parts of the program into services. I'd just rather delay it as long as possible than do it proactively and excessively, which seems to be the current trend. Splitting up your program into dozens of services can't actually reduce complexity, it just moves the complexity from a single complicated codebase into a complicated pattern of interactions between simple codebases. I'd much rather debug a single hairy program because I can step through it in a debugger, rely on a single ordered log file, jump between its parts using IDE features such as "go to definition", etc. Plus, it can be a colossal pain in the ass to deploy a whole fleet of interacting programs written in different programming languages. So I'd prefer to run multiple copies of a monolith behind a load balancer, which is not FAANG-scale but it can scale quite far.
Regarding logging, I've always done it myself instead of using a third-party services, but I've heard good things about Datadog. My pro-tip if you have to correlate logs from many interacting services is to use a structured logging library and attach UUIDs to everything. For example, generate a random request UUID as soon as a request comes in, pass it through everything which handles the request, and always attach it to log lines generated when processing this request, for example in a field named request_id. Then you'll be able to correlate logs from different services easier with a structured log search tool.
Thanks Anon, helpful info. I forgot load balancing exists and that makes way more sense for my needs
>writing small programs that do one thing and do that one thing well
>webshit kool-aid
I wasn't aware that Unix philosophy from 1978 was webshit kool-aid.
>I found this thing called Friday Night Funking
>And I realized the future is short simple games
>connect them with apis
you mean RPCs most of which are auto generated for you.
https://github.com/wasmCloud/wasmcloud-js
try to fathom of depths of what the demo video means for apis.
it means the end of the monolith. why pay to rent GBs of server ram all month when you can instantly reserve MBs of trusted unused ram anywhere for free that DOES THE SAME THING.
you get all them microservices? Great!
put those into docker containers
put those containers into kubernetes pods, bundle into charts in helm
updates to functionality roll out daily, doesnt matter what the update is as long as its delivered automatically after going through jenkins and getting built by gitlab runners
data lives in a redis cache thats backed by wienerroachDB in private datacenters and (terraformed) Azure for on-demand scaling
throw openshift somewhere in there too
the actual codebase is all written in python maintained by poojeets to cut costs
https://circleci.com/blog/its-the-future/
Ummm Unix philosophy transisters?