TC: fo-fitty To those that actually believe, really and truly believe this hype driven PoS with so many issues they had to distribute the parts because the sum of which would = failure. There is no honest person that can say k8s is a great platform. Here are 50 reasons why it sucks, and will continue to suck forever, there is no saving this shite, may god have mercy on your souls for those that continue to hype this dribble. 1) no one wants to manage a cloud inside a cloud 2) breaking api changes 3) breaking charts 4) node upgrades 5) cluster upgrades 6) cert management breakages 7) hard core folks become elitist all-in or nothing assholes 8) it makes your 🍆 shrink 9) kubectl carpal tunnel 10) private overlay networks are awful 11) service meshes are garbage on top of more garbage 12) no over provisioning of cpu or memory (mutha fucka, dont tell me how to run my OS) 13) come back in 2 years and try and upgrade your shit 14) golang sucks 15) api slows down over time due to too many events 16) people thinks its a good idea to have 50% of their compute account for k8s monitoring and metric overhead 17) they just dumped Docker, assholes! 18) who wants all their company resources on this steaming pile of dung 19) vms are making a comeback coz their easier to manage at scale now 20) good luck having an outage when the nginx container is rate limited in a production env 21) oh my god, all the developer support required, no go RTFM for homegrown or bastardize cloud implementations 22) kube-proxy sucks 23) the kubelet can go bl*w goats 24) omg someone at google wrote the ingress controller, piece of shit that just primes nginx configs with rate limiting and fugly ass annotations, why oh why o lord have thou forsaken me! 25) golang templates and helm charts 26) helm v3 27) helmfile 28) helmsman 29) every single fucking Kubernetes SDK 30) Taints and Tolerations 31) crash loop backoffs 32) k8s “pause” containers 32) base64 encoded secrets 33) performance performance performance sucks 34) folks that want to host their companies shitty web apps and apis in a ticking time bomb of the DevOps team with 1 college degree hire and 2 interns, and 1 senior guy that knows to keep his mouth shut or get whacked. 35) github issues trolling for X controller or Y package or Z whatever the fuck else 36) it gives you cancer of the anus 37) its the ultimate trojan horse deathtrap for burgeoning businesses, stay away if your solvent! 38) who really thinks anything that is capable of running “thousands” of nodes is a good idea? 39) why k8s folks wanna grab all the containers for themselves? 40) my OS scheduler works better 41) how fast do you really think service labels are? 42) how were we so stupid to fall into this trap? 43) so now I have a 1 year old setup that we spent 18 months building is now in a place that no one wants to touch and my entire new team wants to spend time building the next version - no this is not innovation dummies 44) did i mention that it shrinks 🍆, it also stretches 🍩 beyond usability save for extra storage volume 45) it smells like doodoo 46) who thought CRDS were a good idea 47) so we’re in prod but have beta apis? How does that work in the real world 48) my friends can launch a fleet of VMs in 1/10th of the time I can get a dev environment up and running 49) its the Antichrist of container orchestration 50) its shine has already faded, and the paint is peeling and the tires are dull
Who uses k8s at amazon?
it’s supported in aws
judithe I'm curious, so K8s is cool tech and all but from actually business value produced, was it useful? Was it lower ops burden on your team? did it solve problems that regular AWS services could not handle? I always hear a "tale of 2 K8s", it's either "damn this is the best thing since sliced bread" or "wow, what an expensive waste". When I look at K8s, cloud native is cool, super configurable which is great, but I wonder how many teams even need all those features. Seems like a bloat in complexity, but I could be ignorant.
It beats using buggy undocumented home-grown shite that only works with a single cloud provider though.
Wrong! Grow a pair and choose your provider, the cloud agnostic debate wont work in this argument buddy. Its the only time id/ive ever recommend to NOT use an abstraction layer. If your going all cloud, choose the best one, and use the fuck out of their cloud until it spins gold. Coz spinning your hard earned mental faculties aligning fictional what-ifs for the sake of a PoS false idol is not worth slowing down and destroying your business.
If you think it's better to write your own cloud-specific container management system, then I'm not surprised you're having difficulty understanding k8s.
I’ve never seen a post on blind I agree with more
Hats off to you for typing all that
Excellent insights. There is a whole industry of vendors who you will happily help you with all of your issues , including 🍆shrinkage but dunno how they can help tighten your 🍩. Maybe try Redhat?? TC: 800k, 16 years of k8s experience. 38 years of small 🍆
MasterCard ehh? You guys are big CF shop
16 years of k8s experience?
Are you in EKS team?
These comments are my own and in no way endorses any shitty k8s cloud provider.
Use VMs+K8s
thats still a crappy solution, why use k8s. Most orgs companies have like 30-150 apps. Containerizing them for desktop development might make sense. But man oh man, its a lot of overhead. HPA, CA, metrics agent, prometheus. Some are building k8s native apps, like integrating parts the eco-system INTO their app, how incredibly dumb is that!
You really put some work op, so what do you think is the next thing ? And where is amazon running it's infra ?
lol, did you type it all? What a fun way to spend your weekend 😂.
All typed on a mobile, thats how much Kubernetes really sucks.
omg, relax, have some fun, take care of yourself. You will probably outlive any framework out there, so most likely will be gone before you are.