Playing the first time with #microk8s for some CI tests. Honestly, opposed to other solutions it doesn't feel so micro... #kubernetes #k8s
Accidentally downgraded my home #microk8s cluster to some version from 2022 instead of upgrading it. The following hour was... interesting
A Comparative Analysis of K3s, MicroK8s, and Alternatives
#microk8s cluster rebuilt with #talos and #raspberrypi nodes swapped out for used hp elitedesk minis with i5s and 16GB…
I always forget how much work moving things around in the rack is but the #homelab is chugging away again happily now.
hey folks who know their postgres and kubernetes, can you help me with this mystery?
i have a bare-metal kubes setup and I'm benchmarking options for cloudnative-pg storage, namely those two
1. longhorn (replica bound) clocks at about 3000 tps
2. microk8s-hostpath storage scored only about 480 tps
that's 6 times slower. why is that? i thought hostpath would provision local PV and it should be faster than longhorn
OK, so I've found fixes for both the problems I ran into yesterday trying to run microk8s on Fedora 41.
I've updated the blog post with the details
https://blog.hardill.me.uk/2024/12/06/microk8s-on-fedora-41/
OK, that was a fight
microk8s on fedora 41 needs the following
- totally disable SELinux
- turn off firewalld (or ingress doesn't work)
Been continuing my adventures of building an #ansible role for #microk8s which worked well and got me to a working thing with a giant but involved.
Apparently one can not properly run molecule test scenarios involving snap packages when using containers for testing.
I found https://bugs.launchpad.net/snapd/+bug/1712808 and wonder if testing in containers might simply be impossible.
PPS it does looks like disabling the #microk8s rook-ceph addon would destroy the rook-ceph namespace even though the addon was not installed with microk8s
Interesting… it looks like #microk8s addins will show as enabled even if they weren’t installed using microk8s enable… so now I will need to remember which truly are an addon and which are not. Hopefully have this in my install notes!
PS not sure how much this matters other than the microk8 list being accurate. Luckily it won’t try to upgrade addons as part of a kubernetes upgrade… I was just curious why rook-ceph (and others) showed up as enabled…
Spent two days trying to make OpenEBS/mayastor work with microk8s and kept running into issues after issues
I gave up and it took 10 mins to boot longhorn
Mal angenommen, ich wollte als Autodidakt Kubernetes lernen, welche #kubernetes Distro würdet Ihr mir empfehlen und was würdet Ihr als hellow world nutzen?
Kenntnisse in vmware, hyper-v, virtualbox, docker, dns, netzwerke usw. vorhanden. Es geht mir darum, hands-on ein Gefühl für kubernetes zu bekommen.
Nach eigener Recherche würde ich mir jetzt nen #microk8s cluster aufsetzen. #talon soll aber auch gut sein?
#Canonical Observability Stack(COS)을 이용한 모니터링 시스템 구축을 소개합니다! #조동원 #김세영 님의 워크샵에서는 #Juju와 #microk8s, #COSLite를 활용하여 모니터링 시스템을 구축하는 방법을 다룹니다. Juju의 개요와 함께 Juju 소개, 부트스트랩, 그리고 다양한 기능을 살펴보고, microk8s의 개요와 배포 방법을 설명합니다. 마지막으로 COS Lite를 배포하고, 다른 컴포넌트를 모니터링하는 방법과 구성 요소를 파헤쳐 봅니다. #모니터링 #시스템구축 #CanonicalObservabilityStack #오픈소스 #클라우드 #UbuCon Korea
키사이닝 파티도 많은 참여 부탁드립니다!
https://discourse.ubuntu-kr.org/t/ubucon-korea-2024-openpgp/49569
From this week's ADMIN Update: Andreas Stolzenberger looks at three simple, compact Kubernetes distributions for operation on edge devices or in small branch office environments. https://www.admin-magazine.com/Archive/2024/80/Simple-small-scale-Kubernetes-distributions-for-the-edge
#Kubernetes #cluster #K3s #MicroShift #RHDE #MicroK8S
Canonical Launches Data Science Stack for Data Science and Machine Learning Beginners #DataScienceStack #DSS #Datascience #MachineLearning #Ubuntu #Pytorch #Tensorflow #Microk8s #Jupyterlab #MLflow #Linux
https://ostechnix.com/canonical-data-science-stack-dss/
Looking to extend my raspberry pi kubernetes cluster into storage…
Does anyone have #homelab performance results for a raspberry pi ceph cluster? Considering making the leap! Will three pi5 nodes be sufficient for managers monitors and osds for a basic HA cluster or do I need more?