|
for windows users who dont understand containers: imagine if your application shipped with the msvc library you needed
|
# ? Apr 6, 2024 22:45 |
|
|
# ? May 3, 2024 01:19 |
|
The Management posted:containers: imagine instead of shipping an app you could ship an entire OS (minus kernel) that runs in its own special little partition that can’t interact with the main OS. containers are fancy bsd jails which themselves are fancy chroots
|
# ? Apr 6, 2024 23:12 |
|
CRIP EATIN BREAD posted:for windows users who dont understand containers: imagine if your application shipped with the msvc library you needed incomprehensible
|
# ? Apr 6, 2024 23:50 |
|
Jonny 290 posted:rotor, i present to you a dilemma - what if we ran the code on kubernetes, but kubernetes is running on an optiplex under greg's desk a cluster of optiplexes under multiple gregs's desks greg is now ha
|
# ? Apr 7, 2024 00:04 |
|
Jonny 290 posted:rotor, i present to you a dilemma - what if we ran the code on kubernetes, but kubernetes is running on an optiplex under greg's desk isn't this every chik fil a?
|
# ? Apr 7, 2024 00:08 |
|
chick phil eh?
|
# ? Apr 7, 2024 00:33 |
|
Jonny 290 posted:rotor, i present to you a dilemma - what if we ran the code on kubernetes, but kubernetes is running on an optiplex under greg's desk
|
# ? Apr 7, 2024 00:35 |
|
why are you upset? sounds like greg's problem to me
|
# ? Apr 7, 2024 00:55 |
|
Jonny 290 posted:rotor, i present to you a dilemma - what if we ran the code on kubernetes, but kubernetes is running on an optiplex under greg's desk this was literally the first version of battle.net (w/o the kubernetes part obvs)
|
# ? Apr 7, 2024 02:21 |
|
Asleep Style posted:containers are fancy bsd jails which themselves are fancy chroots Solaris zones did what we call containers 20 years ago
|
# ? Apr 7, 2024 04:23 |
|
The Management posted:Solaris zones did what we call containers 20 years ago and where is solaris now
|
# ? Apr 7, 2024 04:25 |
|
Jonny 290 posted:and where is solaris now a stark warning from history for those with the eyes to see it
|
# ? Apr 7, 2024 04:46 |
|
hows Rancher? is it also a giant pile of timing windows and polling loops? also terrible, but differently?
|
# ? Apr 7, 2024 15:18 |
|
fresh_cheese posted:hows Rancher? is it also a giant pile of timing windows and polling loops? never heard of it
|
# ? Apr 7, 2024 15:26 |
|
oh. its just an opinionated kubernetes nmind - i thought it was its own thing.
|
# ? Apr 7, 2024 15:29 |
|
fresh_cheese posted:hows Rancher? is it also a giant pile of timing windows and polling loops? it’s really only a management platform sitting on top of k8s. it does nothing to change its nature. it has some opinions on the load balancing and ingress you should deploy but it’s not hard to change those opinions also, it’s p deece when you want to click instead of type
|
# ? Apr 7, 2024 15:32 |
|
shitface posted:also, it’s p deece when you want to click instead of type lol noob
|
# ? Apr 7, 2024 15:40 |
|
CRIP EATIN BREAD posted:lol noob eh. makes it easy to check poo poo. but for fucks sake don’t ever try to upgrade a helm package with it. it has a mysterious ability to gently caress this up
|
# ? Apr 7, 2024 15:46 |
|
the abstractions it provides are better than whatever your in-house ops bozos would cook up
|
# ? Apr 7, 2024 15:48 |
|
my homie dhall posted:the abstractions it provides stopped reading here
|
# ? Apr 7, 2024 15:51 |
|
my applications are so wild and free they cannot be contained, op. let alone orchestrated.
|
# ? Apr 7, 2024 16:45 |
|
Steve Jobs would have loved kubernetes because he liked playing orchestration
|
# ? Apr 7, 2024 18:08 |
|
fart simpson posted:stopped reading here Most software presents an abstraction of some kind but I agree with not reading about software
|
# ? Apr 7, 2024 18:13 |
|
i call my approach to software test “strategic ignorance “
|
# ? Apr 7, 2024 18:17 |
|
my take is that adding abstractions to k8s makes it harder, not easier. a good multi-purpose helm chart or two is all you really need e: and yes i know helm charts are an abstraction, but they're a nice thin one well-read undead fucked around with this message at 18:31 on Apr 7, 2024 |
# ? Apr 7, 2024 18:17 |
|
my homie dhall posted:the abstractions it provides are better than whatever your in-house ops bozos would cook up then the same in-house bozos will come up with any excuse to use k8s (or eks, or aks) instead of other managed services by aws/azure that offer even higher levels of abstraction that are better than whatever your in-house ops bozos would cook up.
|
# ? Apr 7, 2024 22:22 |
|
all I know is I used to run my stuff on one t3.medium and now after “rearchitecting” 5 t5.larges and a gazillion AWS queues is apparently still not enough for this beast
|
# ? Apr 7, 2024 23:25 |
|
FormatAmerica posted:then the same in-house bozos will come up with any excuse to use k8s (or eks, or aks) instead of other managed services by aws/azure that offer even higher levels of abstraction that are better than whatever your in-house ops bozos would cook up. agreed, running kubernetes on public cloud that has managed services like that makes no sense
|
# ? Apr 8, 2024 04:17 |
|
FormatAmerica posted:then the same in-house bozos will come up with any excuse to use k8s (or eks, or aks) instead of other managed services by aws/azure that offer even higher levels of abstraction that are better than whatever your in-house ops bozos would cook up. my homie dhall posted:agreed, running kubernetes on public cloud that has managed services like that makes no sense generally, where practical, I do like to keep things like databases externalised from the cluster and these services can work well for that, no doubt. I highly suspect that many, if not all, are just running in k8s with the detail hidden from you though. but anyway what managed services will scale your application nodes automatically and transparently? what equivalent to hpa do the cloud providers offer? (genuine question. I don't know) ime the experience you have with k8s is largely down to the competence of the original architect and build engineers. provided they have designed and setup the architecture constraints correctly, provided robust pipelines for upgrades/deployments etc., k8s become a pretty automated self-healing substrate for applications (except when it's not) shitface fucked around with this message at 04:32 on Apr 8, 2024 |
# ? Apr 8, 2024 04:28 |
|
shitface posted:generally, where practical, I do like to keep things like databases externalised from the cluster and these services can work well for that, no doubt. I highly suspect that many, if not all, are just running in k8s with the detail hidden from you though. but anyway what managed services will scale your application nodes automatically and transparently? what equivalent to hpa do the cloud providers offer? (genuine question. I don't know) aws has autoscalers for ec2 instance groups that you can connect to various metrics in various ways and i'm sure there are similar things in other public cloud providers. they're actually what the cluster autoscaler uses under the hood if you're running eks (or otherwise running k8s in aws) but they suck to set up. most stuff in aws sucks to set up. i'd rather use the k8s control plane any day of the week, and it's just a bonus that that control plane is portable to other providers the thing isn't that kubernetes is this beautiful thing that's simple and perfect, it's that other things are and always have been absolute dogshit once they've matured past the complexity of optiplex under greg's desk
|
# ? Apr 8, 2024 04:59 |
|
|
# ? Apr 8, 2024 05:00 |
|
well-read undead posted:aws has autoscalers for ec2 instance groups that you can connect to various metrics in various ways and i'm sure there are similar things in other public cloud providers. they're actually what the cluster autoscaler uses under the hood if you're running eks (or otherwise running k8s in aws) TIL
|
# ? Apr 8, 2024 05:35 |
|
let me just pedant myself though: i think it might just be the target groups that the cluster autoscaler uses, not asgs themselves, idk, it's sunday
|
# ? Apr 8, 2024 05:56 |
|
well-read undead posted:aws has autoscalers for ec2 instance groups that you can connect to various metrics in various ways and i'm sure there are similar things in other public cloud providers. they're actually what the cluster autoscaler uses under the hood if you're running eks (or otherwise running k8s in aws) I've used ECS for a few different services and it's fine, I never used any of the advanced features in kubernetes, ECS basically just has the stuff I care about which is deploying and auto scaling containers
|
# ? Apr 8, 2024 06:06 |
|
i used ecs for a very short period and let me tell you friend i found it to be excruciatingly bad
|
# ? Apr 8, 2024 07:41 |
|
well-read undead posted:aws has autoscalers for ec2 instance groups that you can connect to various metrics in various ways and i'm sure there are similar things in other public cloud providers…. Azure Container Apps is opinionated Azure Kubernetes Service that sets up ingress with properly signed certs for each app, autoscaling with good defaults but it takes all the fun of fiddling around with K8s away so I will never hear the end of the bellyaching about not using aks
|
# ? Apr 8, 2024 13:47 |
|
FormatAmerica posted:then the same in-house bozos will come up with any excuse to use k8s (or eks, or aks) instead of other managed services by aws/azure that offer even higher levels of abstraction that are better than whatever your in-house ops bozos would cook up. This is the truth. They'll run k8s on ec2 nodes, then they'll run AKS, then they'll get to wondering why they don't just run everything in ecs. Every step of the way OPS guy is blocking any attempt to go straight to the conclusion with "BUT WHAT IF WE NEED THE POWAR?!?" Then once it's all finally peaceful they'll all move on to more senior jobs on this experience and get replaced by a bunch of wide eye'd noobs once again asking the unholy question and the cycle continues through yet more crap tech stacks. FormatAmerica posted:Azure Container Apps is opinionated Azure Kubernetes Service that sets up ingress with properly signed certs for each app, autoscaling with good defaults but it takes all the fun of fiddling around with K8s away so I will never hear the end of the bellyaching about not using aks Internet Old One fucked around with this message at 18:06 on Apr 8, 2024 |
# ? Apr 8, 2024 18:03 |
|
well-read undead posted:my take is that adding abstractions to k8s makes it harder, not easier. a good multi-purpose helm chart or two is all you really need helm is a horrid abomination of hasty design decisions and aimless open sores development gotemplate should have never been let near yaml and is complete rear end to do anything complex with, which you'll inevitably need to do for a chart not targeted at a single environment because there's gently caress all way for end users to make their own adjustments i also love the people that clamor for the ability to write their own templates inside values.yaml, like what the gently caress, why
|
# ? Apr 8, 2024 18:08 |
|
look, it's not pleasant to write go templates, but it's completely manageable, and far better than running a whole additional stack of machinery that does god knows whatQtotonibudinibudet posted:i also love the people that clamor for the ability to write their own templates inside values.yaml, like what the gently caress, why yeah that's cursed
|
# ? Apr 8, 2024 18:19 |
|
|
# ? May 3, 2024 01:19 |
|
what if someone made an OS that you can just run your app in? and it was standardized so you could run your app in it everywhere. and you wouldn’t have to ship a whole userspace with your dumb service
|
# ? Apr 9, 2024 03:51 |