Clojurians Log v2

https://jasonbellmusic.com - all the soundscape stuff is there. Working on a solo Stick album for October.

Low D, get a Low D! 🙂

Already did!

I'm not very good with Piper's Grip, and I'm not as happy about this particular Low D as I had hoped (James Dominic). I'm hoping I'll become more happy about it, but it was cheap, so no worries if I don't.

Do you play, @jasonbell?

Not whistles no…. just these.

ooo

what is that

give us a listen!

Is it like a lap steel something?

It’s called the Chapman Stick. Ten or twelve strings, played like a piano.

Got any recordings? Curious to hear how you sound

Midday sun shining here in Denmark. Where are you?

Northern Ireland.

😄 We call 11'o'clockish "formiddag", "pre-midday"

Which county? I might as well learn a bit of geography

https://maps.app.goo.gl/uMNvyhHgo4U2QyeSA

huh, never heard of it 😄 I did learn something!

The problem is not Kubernetes itself, its an ecosystem and culture that forces teams to overengineer; its possible to have a very constrained k8s setup (i.e. ban non-stateless applications, don’t use service meshes etc, don’t use helm, host all o11y externally) and have it reliable. Also Nomad exists and it is very simple to understand and operate in comparison

oh and adopting kubernetes without solving all application observability problems first is a disaster waiting to happen

we use datadog for o11y (had to look that up), but not sure that was a good idea.

can be, there are lots of things to like in datadog, and it is a very complete suite. That is also the main problem, some organizations have gotten massive datadog builds by not constraining their use of it and enabling all kinds of features 🙂 Point is, you need to be 100% sure all information is available ASAP before you get an incident, that applies for everything of course, but the lack of info can be much more painful in microservices application where each container is disposable compared to VMs that retain logs and historical information

I think the safest way to use k8s is not to fall into the “kubernetes native” trap (or any XYZ-native hype, we need more arguments for something than just association 😄 ) and use it for stateless workloads when possible. I know there are operators and volumes and what not, but its okay to use managed cloud services or VMs that are not k8s nodes if it is simpler to run and reason about

it's not just you

aah ok. no wonder I don't get it.

I don't remember where I heard it the first time, but I've stuck to it: don't even consider kubernetes unless you have a million dollar ops problem

I suspect we do spend that much money on it... but mainly because we run waaay too many K8 clusters. (I suspect)

I've heard this sentiment before. I've never touched k8. Where is the complexity? Too much configuration needed?

it's more that you go from deterministic to organic, k8s will restart/move pods etc based on health checks and other signals. Makes sense at scale, but can also go badly wrong. If you can name the different servers you have, you don't need k8s

but yeah configuration itself is also a big thing

We need to change ingress controller... and my colleague explaining what is needed to replace it sounded so complicated. I didn't understand any thing about it.

not to mention "helm charts" which are string templating over yaml... if that sounds like a terrible idea, that's because it is

yeah, lots of sutff....

and things like operators and controllers and loads of over stuff. nodes, pods etc.

maybe I am just too old for stuff like this.

I think Kubernetes is intentional complexity, not accidental. Malicious, too

🤣

sounds very likely

Our work project has enough incidental complexity as it is. I'm very happy we don't have K8s! Or Microservices, someone was keen on that at one point 😮

Microservices can be bad as well... I worked on one project where they had over 150 Microservices.... and doing transactions over multiple services.... how to shoot yourself in the foot with a massive canon.

Isn't microservices akin to shooting yourself in the foot with a thousand tiny cannons?

hmmmmm 🤔

I need to think about that.

a massive cannon shooting a thousand tiny cannons at you

🤣

that must be it!

Reminds me of this

seems about right

Hey... that is our secret architecture picture!

how did you get that?

what do you mean, you don't have one?

ooh... we do... way too many. We have a whole cadre of Solution Architects working here. and they make lovely Archimate(?) pictures.

Yes. While I prefer Podman to Docker, Docker works just fine.

I remember at a certain job we resisted hard to bring in kubernetes, which one new colleague insisted on

we were using docker swarm at the time

There seems to be a lot of CV driven development these days.

A long time client of ours at some point decided to go k8s. They had a potential big client, but it had to run "inside the firewall", and they were using k8s, so they were kinda forced. This is where I learned most of what I know about it. They figured it out, it was mostly fine, but they kept running into stability issues. It almost bankrupted them. In the end they went completely the other way, a bunch of shell scripts running in tmux. AFAIK that's what they still do today.

wow, what a change.

hahahaahahaha

it's not often that I genuinely laugh out loud at Slack stuff

we run so many AKS clusters I think we lost count.

Shell scripts are a different kind of poison, or so I've heard. I think I haven't been bit by overly ambitious shell scripts yet, fortunately.

I'm a big fan of tmux though 😄

babashka

just what I was thinking 😉

Yes, that's what I would do 😄

If it got to that point.

all great minds think alike 😉

I'm also a big fan of babashka. We've got some lovely bb scripts at $work