i believe the .sprite dir has some stuff to help claude answer those questions. haven’t done it myself but my friend said he was able to get claude to set it all up for him (yolo mode helps) including connecting to github.
I might just be too dumb to read even an abstract, but it sounds like they said "meh idk, no strong evidence"
(also curious if they'd call going hard on crypto / options "gambling" or not :P, which might be more about how hard it is to get ahead in todays economy than playing games as a kid, but that point is a very hard digression from what we're talking about here).
I suppose that's true in one sense - in that I'm using EKS heavily, and don't maintain cluster health myself (other than all the creative ways I find to fuck up a node). And perhaps in another sense: It'll try its hardest to run some containers so matter how many times I make it OOMkill itself.
Buttttttttt Kubernetes is almost pure maintenance in reality. Don't get me wrong, it's amazing to just submit some yaml and get my software out into the world. But the trade off is pure maintenance.
The workflows to setup a cluster, decide which chicken-egg trade-off you want to get ArgoCD running, register other clusters if you're doing a hub-and-spoke model ... is just, like, one single act in the circus.
Then there's installing all the operators of choice from https://landscape.cncf.io/. I mean that page is a meme, but how many of us run k8s clusters without at least 30 pods running "ancillary" tooling? (Is "ancillary" the right word? It's stuff we need, but it's not our primary workloads).
A repeat circus is spending hours figuring out just the right values.yaml (or, more likely, hours templating it, since we're ArgoCD'ing it all, right?)
> As an side, I once spent HORUS figuring out to (incorrectly) pass boolean values around from a Secrets Manager Secret, to a k8s secret - via External Secrets, another operator! - to an ArgoCD ApplicationSet definition, to another values.yaml file.
And then you have to operationalize updating your clusters - and all the operators you installed/painstakingly configured. Given the pace of releases, this is literally, pure maintenance that is always present.
Finally, if you're autoscaling (Karpenter in our case), there's a whole other act in the circus (wait, am I still using that analogy?) of replacing your nodes "often" without downtime, which gets fun in a myriad of interesting ways (running apps with state is fun in kubernetes!)
So anyway, there's my rant. Low fucking maintenance!
I use it to power CI builds (a lot of them!) and have extremely little issue with it.
Basically I’m just using the API to spin up machines, which do some work and shut down. There’s some extra machines per build job, like database containers or a headless browser for testing. Pretty smooth in my experience.
I think the only occasional issue I hit is the internal DNS being a few seconds behind reality.
Our kids were too young for school during the pandemic (thank god, it seems like it really messed up kids lives).
Now they're both in early-years of school. Attendance is kinda interesting!
We parents EXTREMELY want the kids to go to school as it's the much-needed break to get our own work done - both in basic upkeep of our lives/household, and in our careers. We need the school hours to be long. Longer than they are, in fact. (As a side note, my kindergartener only gets like 20 minutes of recess? wtf? Have you met children before?)
On the other hand, life is expensive. Being restricted to only travel during the most expensive times to travel (around school holidays) isn't ideal. We can work with our kids to make up lost school time.
I also just don't like this third party entity whose value seems to go down every year to control our lives!?
Teaching (from the teachers point of view) increasingly is geared towards meeting metrics that are divorced from the needs of the kids. The teachers incentives are being misaligned with ours.
Additionally, I think we've (royal we) grown distrustful of public school in general. Not in a "big government" sort of way, just that we need to acknowledge that US public schools are designed for conformity. Being different (e.g. having ADHD, or being "on the spectrum") is not tolerated well - you might find your kid in a special needs bucket that effectively segregates them into programs that might not fit their needs at all.
At the same time, private school costs are huge and often the ones closest to you come not only in an extreme monetary cost a culture cost - being overly religious, or not religious enough (YMMV).
So, yeah, it's hard to really want or care to "be a model citizen" to the public schools that are increasingly putting up the pressure on parents (that's a whole other topic, why aren't grandparents capable of being helpful any more? where did our support networks go?) while standards that might be outside of the school's control are lowering their ability to give quality education.
gpu-friendly base images tend to be larger (1-3g+) so that takes time (30s - 2m range) to create a new Machine (vm).
Then there’s “spin up time” of your software - downloading model files adds as long as it takes to download GB of model files.
Models (and pip dependencies!) can generally be “cached” if you (re)use volumes.
Attaching volumes to gpu machines dynamically created via the API takes a bit of management on your end (in that you’d need to keep track of your volumes, what region they’re in, and what to do if you need more volumes than you have)