wanted to set up a k3s cluster with my pi's. took me a fucking whole day to find useful ansible playbooks (which I needed to fix because outdated).
I want to habe metallb and nginx ingress running, so that differs from the default.
and now i spent the whole day trying to install a fucking pi hole and for some reason metallb does not fart out an external ip for the pi hole.

found several issues regarding this matter.

maaaan i am completely new to this whole clusterfuck and i feel a bit overwhelmed atm. i thought this would be easier. am i just an idiot?

so you're building a cluster (modern word for beowulf) from tiny computers running linux and you think you're an idiot for thinking this would be easy ?

ok that doesnt sound quite right.

point is, no probably not the easiest setup my friend if you're trying to do what I think you're trying to do which is share process execution amongst multiple nodes and report results back to one controller sitting at the end of a private network segment, is that about right ?
again ?

ps you are alive right ? this is a repeat post.

repeat post? this is the first one of me on this topic xD

and yeah i thought i could manage this with less effort.

after all it turned out that i just needed some sleep. i reset my cluster today (thanks to ansible this was less than 5 minutes effort)

and until now everything seems to just work. i have an nginx ingress, metallb and pihole running at the moment.

the thing is my background in this whole topic. i am using containers for a while now. but this wednesday i wrote my first own dockerfile from scratch and managed to get it running after several tries. two days later i wanted to setup a cluster without actually knowing how this all works. wanted to learn 'on the go'.

the first day i did not document my steps. second day i began to write everything down.

I now have a vs code devcontainer with everything needed to setup and manage the cluster. and a very simplistic k3s cluster running. still a long way to go but finally i am makinf some progress

@nebula I’ve not yet had an occasion to containerize apps. I can see the deployment benefits but if I didn’t use hyperv vms I always ran things bare metal

well, it is a huge difference in deployment and it changes the way how you manage your infrastructure.

i aim to have everything deployable in a matter of minutes.

I have rewritten the ansible script and with help of a makefile i can now type 'make teardown' and the whole cluster is whiped (except for the persistant storage). and if i type 'make' it will provide me an empty cluster.

as for the applications running: i am working on automatic deployment for that also.

it is a huge time saver and with a healthy backup concept for storage I can now sleep a bit better.

i also like that you are able to specify that if one application goes down the cluster just spawns a new one - or that you spawn a new instance if performance is poor.

I wish someone would publish actual benchmarks to determine performance drops
One thing I don’t like about things like this is if I’m not the one who owns the hardware because let’s take openshift you couldn’t see or handle anything related to t he base system because it was not self hosted and I’d say an out of the box demo program failed to deploy you didn’t know why

@MadMadMadMrMim yeah in managed clusters this is a thing and i totally agree.

on my local cluster i have everything in hand. i also have a vserver and there i am also able to monitor everything.

@nebula when does the world move forward and the employment opportunities shower down on is all again ?