Has anyone used AWS Fargate? I'm not all-in on AWS tech, and am feeling like it's better to keep things as agnostic as possible for projects. But maybe fancy tech is worth it? Or is this containers-lite?

/@keita @cgiffard
242840
@33MHz It sounds like Kubernetes on AWS, where you just lob Kubelets at them and they worry about the infra for you.
// @cgiffard @keita
242850
@jws it’s a bit different. For Kubernetes (k8s) and AWS’s proprietary ECS, you still need to manage the instances where the containers are running on. Fargate is a service that does that for you — you still need to use k8s or ECS @33MHz @cgiffard
243080
@keita I didn't connect K8S and ECS together like that before, as competing container orchestration options. I think I tried to say what you said. Briefer: Fargate = "Let Amazon worry about where containers run"? @cgiffard @33MHz
243082
@jws “let amazon worry about the instances containers run on”, I think. ECS and k8s choose what instance to put a specific container on — although with Fargate the line is blurred @33MHz @cgiffard
243086
@cgiffard @33MHz @jws also, if you’re wary of vendor lock-in, I would highly recommend against learning ECS and waiting for EKS (amazon’s managed k8s service) to become public. Or use GCP and their GKE service
243081
@jws @33MHz @cgiffard k8s is super complex though. Really powerful, but has a pretty steep learning curve (at least for me. Maybe you’d pick it up really quickly:) )
243083
@keita @cgiffard @33MHz Haha, I am a terrible sysadmin. I count myself lucky my DO droplet hasn't gone up in flames yet. ;) (Or maybe it has and I don't have the monitoring to tell. >.>)
243084
@jws ugh, that happend to my VPS yesterday.
It just hung, so no error monitoring, but it was dead. Need to improve that monitoring... @33MHz @cgiffard
243089
@keita @cgiffard @33MHz Julia Evans has some great blogposts on the hard stuff she's learnt about k8s. I'd start by learning from her mistakes. ;) https://jvns.ca/categories/kubernetes/ and https://jvns.ca/categories/containers/
243087
@jws thanks a lot. Once EKS is available (probably earlier with ECS), I’m going to move as much as I can to containers. We’re managing 70-120 instances by hand and it’s not pretty. @33MHz @cgiffard
243090
@keita Does it give you much orchestration capability over, say, EC2 via terraform? @33MHz @jws
243276
@cgiffard yeah, especially when considering application environments. It’s a lot easier to maintain a Dockerfile than it is to maintain a set of Ansible or Chef scripts @jws @33MHz
243287
@keita TIL: I’ve successfully waited out the need for Ansible. ;)
// @33MHz @cgiffard
243434
@jws hopefully the only thing that's being managed by Ansible after I'm done with this is Elastic search! @33MHz @cgiffard
247771
@keita Hah. We just shove that in a managed service and make it somebody else’s problem! :) @33MHz @jws
247796
@cgiffard I really wish we could do that. Maybe we could, actually.. we're only using a few irregular plugins... @jws @33MHz
247798
@keita The managed service from elastic has kuromoji, but if you need more you can get them to install some stuff and change configuration if you’re on a support plan. @33MHz @jws
247880
@cgiffard hmm. That might just be enough for us @jws @33MHz
247881
@keita @cgiffard @33MHz You’re making me want to burn my VPS down and rebuild. Maybe next year… Though I kinda want to go the FreeBSD route, I definitely am more familiar with Docker than with FreeBSD jail admin at this point.
247832