I love lens for my work and also for my homelab! It reduces me the stress of trying when troubleshooting something, and you don't remember the exact command you need. I've also recommended it to new k8s admins. I've been always thought about using portainer as I've used it only with docker, and it's also a great tool. Will try in my homelab for sure!
We have rancher installed at work , it's quite a complex system the only thing that i don't like it's how bloated the cluster becomes once installed but that's probably because we are not using it as they recommend it, and that is to have your management cluster where rancher is installed and have separate workload clusters :)) , we also have it integrated with our Google Workspaces account but will probably switch it to Active Directory
Be prepared for a wild ride with AD. We had major problems with LDAP auth, some users just weren't able to login for no apparent reason... I opened a bug request but everyone seemed to ignored it. Maybe it works now but we settled for Keycloak that was connected to AD. That worked well!
I've used portainer and lens for some reasons, yes lens is good for home lab server with limited resources like raspberry pi or may be server with small resources like my home lab server and I like portainer since I known docker, thanks for always making content like this sir. sorry my english bad
I have hyped Portainer for the reasons you mentioned with Rancher. Thanks for drawing my attention to Lens as that look like a great balance. I also want to give K9S a go just because. This was an excellent rundown. As always thanks another great video.
Great video and I agree on all points you have mentioned. One thing I would like to ask is what shell you are using on mac? How do you get the auto-complete to show up and the colour scheme. Would appreciate if you can reply.
Using Rancher from some time and it's quite good to manage Multiple Kubernetes Clusters very smoothly & Rancher also has built in central GitOps with project Fleet
I’ve used rancher in the past but currently I’m using lens and of course kubectl Longhorn is amazing BTW, I use it exclusively for my cluster PVC’s at the moment
Longhorn has a very major flaw … if you have a power outage. When the pods come back up they don’t auto fsck. You might need to fix every single volume manually with an fsck -y, they don’t do this for you automagically. I wrote a cronjob to just look at syslog and auto repair any volumes that show up requiring an fsck on mount (as long as they are longhorn volumes)
Another tool I like is stern. It is for tailing logs. What I like about it is that I can use selectors to display the logs for all pods in a deployment and use since so get just the most recent logs while tailing. Don’t know about newer version of kubectl but it wasn’t able to do all three at the same time the last time I tried and now use stern exclusively.
@@ThePapanoob You may be able to now but in the past I was not able to use ‘since’ and ‘follow’ together which meant I had to wait as it scrolled through all the logs to get to the end. It wasn’t just the time but it also cleared the scroll back buffer on my terminal. Another feature I like is that it prefixes each line with the pod name when displaying logs from multiple pods. If you like kubectl, stay with it since sticking with one tool is a lot easier. I wasn’t. Also stern is not the only third party tool for looking at Kubernetes logs. Just the fact that there are alternatives implies that kubectl is less than optimal for many people.
OpenLens > Lens None of that Mirantis commercial crap packaged into it. No forced account creation, etc. We use OpenLens at work, it just hooks kube config, and AWS EKS cli just adds cluster and user profiles to kube config, so it picks it up just fine. Not sure why you would need AD integration with OpenLens when OpenLens is not an auth mechanism itself, it just hooks existing mechanisms through kube config. Unfortunately, there are no "official" builds of OpenLens since Mirantis bought/funded the OpenLens team, as Mirantis obviously doesn't want to undercut their commercial product. You have to do a little bit of searching for how to build or install 3rd party builds of OpenLens.
Hey.. Christian... Do you have some experience using jsonnet? ( data templating language) in kubernetes... If yes Can do a video about it..? I heard it could be a good alternativ to kustomize or helm
would Harvester HCI be applicable here? I just found out about the platform recently. I'm interested in how it differs from setting up proxmox > vms > portainer / rancher > kubernetes
There is also fish shell and other fish like extensions for other terminals (zsh etc) that arent dependent on specific terminal emulator, but not quite as fancy either. It will inline hint out suggestions or history, depending on how you have it configured. if you are using kubectl, make sure you setup kubectl autocomplete for the shell you use. zsh/oh-my-zsh has autocomplete plugins for kubectl that will query the k8s api and autosuggest things relevant to your current cli query. ergo, if you are typing "kubectl exec -it " and double tab, it will discover the pods in the current contexts namespace, and suggest them for you for autocompletion.
I have been using K9S at work were I don’t have WUI or GUI and really like it. I don’t think that the lack of ability to create resources as a drawback. I have been frustrated with tools like Portainer that “own” your resource definitions. Once you create the resource using their tool, you can’t export it and use it outside of their tool. I use Portainer at home but never to create or edit my resources so I own my configuration.
@@borntoexcel14 I don’t create resources with K9S or Portainer. At home I don’t worry about supporting multiple environments so a simple YAML file with related resources works well. I also experimenting with Ansible for Docker. At work I am working with deployed resources. The biggest differences are in configuration/secrets and scale. I have tried Terraform and CloudFormation. Currently we are using Jenkins to provide environment specific properties. As a developer I find it very enjoyable to come up with complex solutions but in the end, they don’t add value. They just add complexity that you need to deal with in the middle of the night. The other issue with complex solutions is that no one else wants to touch it so you own it and can’t move on to new projects.
IMO, Rancher is just too buggy for long term usage. I use it for years and every year ran into major problem and have to rebuild my cluster from time to time. I ended up dropped it for my homelab setup
@@LampJustin that's why I moved to rke2+fluxcd v2+lens. I am start seeing rke2 not that stable either.... so might ditch it for Talos or something. But I need a way to run nebula overlay VPN and yet find a way to do it right.
@@admun yeah that's why I just use kubeadm based systems like kubespray, clusterapi and Talos. That's the most stable you can get. It's what almost everybody uses. Nebula is pretty cool. If you don't rely on it too heavily, you can use Talos and it's WireGuard based mesh.
Pity Lens is not being offered with a tablet/smartphone platform in mind as well, the download section only lists MS exes, linux rpm/deb/appimage/snap and macos dmg... while dropping the graphical UI from the cluster to a desktop app releases the resources on the nodes for actual work, users typically want to be able to use/monitor/manage the cluster from any device. If the UI is baked in into the kubernetes distribution or the node OS, you only need a device with a browser to work with the cluster.
a little too late to the party here. I want to practice Kubernetes on my WSL2 setup, which i will ofcourse push to a VPS in the cloud for production. any tips or video you can direct me to @Christian?
* Better lifecycle management than just docker alone. Deployment/Statefulset upgrades are "built-in" to some extent and can roll-back if not successful. * Helm charts are available for most software that is already containerized * Much better ingress handling for multiple services vs just a self-managed reverse proxy * Cert-manager automatically will generate and renew certs (incl LetsEncrypt) for your web services, just "enable" SSL on the ingress * external-dns can automatically create/update/delete DNS entries for your web services, either locally (such as a LAN DNS server), or externally (such as CloudFlare or some other service). * Can ingress, secure, and proxy traffic to other web services outside of the cluster, but using the cluster to manage access to those web services centrally. Ergo, you have another web server thats not running in kubernetes, but can use an ingress, service, and manually created endpoint, to point to that and have it all SSL and DNS automated via aforementioned tools. * Kubernetes at its core, is a state engine, so it will do everything in its power to maintain that state. Liveness probes, readiness probes (readiness determines availability for receiving traffic, vs liveness means its running, but not dead), restart containers when things aren't responding or crashing. * Could setup HA cluster, even on a single bare metal machines using VMs, so that you can have an "always up" path for updating the cluster. Not redundant, but hey, something. * Could setup HA cluster on things such as raspberry pi quite cheaply. * Can define your home environment and services more agnostically to what is actually running kubernetes. Setting up a kubernetes cluster can be a fairly straight forward process and runs on just about anything. You could back everything in a git repo, then redeploy your entire setup from scratch this way within just a few steps if you had to. * More storage options and redundancy out of the box with kubernetes storage providers than with something like docker alone. * You learn how to properly setup, manage, and upgrade a Kubernetes cluster, which can get you jobs paying $180k+ with no degree, as even the concepts in a home lab scale on Kubernetes. * I am sure there are several more points that could be made, but these were just the ones off the top of my head.
@@jslay88 For the ingress part I just go with Traefik, a compose file start with Traefik seems to be able to do them all simple and easy. I don't know other benefits though. But I really could switch to k8s if it's much more resource efficient like less storage or cpu or memory, although as I can see it always drains more because you have to pair it with a lot of stuff like tools OP suggests or tons of clusters to help maintain visualizable and appreciable. Just my two cents though, I don't know nothing about k8s, I'm sorry if I'm wrong.
Portainer has gone completely crazy, they monetize in such a vulgar way that the free version is starting to feel like a cancer, it is over run by stupid things reminding you that you are getting it for free, well with all their annoying "BEGGING" for money I will much rather eat my own foot that support their stupidity. I feel like taking the code, fork it and remove all their CARP and give it ack to the community, but not stop there and get a full time developer to push hard to make the free version better than the Begged for version.
I don't think that's true. Developers cost a lot of money. And somehow there needs to be money made to pay the people. They also give a lot of stuff for free back to the community.
Half of my comments keep disappearing/being deleted. Worthless. What's the point in trying to provide useful information here if its just gunna disappear. I spend quality time trying to write something useful up and help but nope! Later.
