It is a antipattern to use ansible for provisioning. Ansible is for config managment. Better use the official terraform provider for creating you're talos cluster. I recommend you to take a look into the cilium CNI which also integreated LB Stuff like MetalLB directly into the CNI and uses BPF. But I'm happy to see more about Talos on youtube! Hope more people using it instead of the old OS ways :)
Been meaning to give Talos a proper go, and this will definitely push me over the edge. I'm also curious about Typhoon, not sure if you've heard of it, but looks like an interesting and declarative way to deploy K8s clusters. I'm testing stuff at home but need to find a solid solution to deploy at work too... I guess it truly is a "home lab" haha Cheers and thank you for all your quality content, both the videos and playbooks! :3
I switched a home cluster over to Talos v1.7.6 a couple of days ago, it took a little bit of work - my synology-csi driver needed some reworking and Cilium attempting to replace the Kubeproxy just wasn't a working configuration but after those small speed bumps it's all been really good. It really is minimal and I did my first K8S upgrade today from 1.29.7 to 1.29.8, it pops up with a lot of warnings saying that it's reading 1 vs 2 but this is just waiting for the config to converge before moving onto the next steps. Pretty flawless apart from with this config I have noticed a lot of pods left in Error or Complete on the node restarting (currently only 1 node). This could be related to Cilium terminating ahead of other nodes and blocking clean shutdown of the pods but I'll understand it better soon I'm sure.
1:30 No no no, that's NOT what that badge says. Kubernetes is certified and Talos is a member of CNCF, just like 700 other entities. But that doesn't certify Talos on any way. They're not even an active project on CNCF Projects. Careful with those statements, please!!
Ha, I installed that on four RK1 CMs on my Turing Pi 2 board a week ago. I have yet to put anything else on the cluster though, but it was very painless to get up and running. Caveat, I did use their Terraform provider instead, I'm not a fan of blobs of YAML.
Hi Jim, thanks for the amazing video! It works with one issue, i don't get the talosctl config under .talos. So i'm not able to communicate via the talosctl, kubectl works perfectly. Any hint? Thanks in advance! 🙂
@@Jims-Garage I found it out, the config is always placed under /home/ubuntu/.talos even if you start the script with a different user. I just copied the config over and now it works. Thanks a lot! 🙂
Nice! Have you tried running Longhorn on it? There were some issues in the past because Talos didn't support open-iscsi which is a requirement of Longhorn's V1 storage engine. Their V2 storage engine should work with Talos because it's using the SPDK storage backend instead of iSCSI (if I read that correctly). Haven't tried it yet though, and I'm not sure if you can easily convert your existing V1 volumes to V2.
You are using Kube-VIP with Talos? You dont need to use it because Talos has the VIP function build in. And there is a much easier way to deploy talos, you can boot it with pxe and give it the talosconfig file with pxe. After that it installs it by itself. Thats they way i am deploying my talos cluster. :)
No, I'm not. It's a relic from my RKE2 (explained in the video) and something that I am actively investigating. It's become clear (and thanks for confirming) that it's superfluous.
I'm glad to see a shout out to my cluster-template project, hope your Talos journey has been as fun as mine! Thanks Jim.
@@devinbuhl you're welcome, it's an awesome playbook, I'll definitely be trying it out.
I've been dragging my feet about trying out Talos...this may make me actually try it. Great vid Mr. Garage!
It's well worth a look IMO, but please... Dr. Garage 😜
@@Jims-Garage Ahh sorry, Mr. Dr. Garage
That's more bloody like it!
It is a antipattern to use ansible for provisioning. Ansible is for config managment. Better use the official terraform provider for creating you're talos cluster.
I recommend you to take a look into the cilium CNI which also integreated LB Stuff like MetalLB directly into the CNI and uses BPF.
But I'm happy to see more about Talos on youtube! Hope more people using it instead of the old OS ways :)
sheesh, now I'm going to have to tear down and rebuild my cluster LoL!
It's never a dull day in the homelab
won't be that hard, just replace each node one by one, draining/cordoning/deleting each old one in the process
Thanks for the video. Happy to hear you are enjoying Talos!
You bet! So far, so good. Going to continue playing with it.
Waiting for Part 2 :)
Thanks for video, bro, very interesting.
Have a good day and Ride on!🙂
Thanks, you too!
Been meaning to give Talos a proper go, and this will definitely push me over the edge. I'm also curious about Typhoon, not sure if you've heard of it, but looks like an interesting and declarative way to deploy K8s clusters. I'm testing stuff at home but need to find a solid solution to deploy at work too... I guess it truly is a "home lab" haha
Cheers and thank you for all your quality content, both the videos and playbooks! :3
You're welcome! I haven't yet played with Typhoon but have heard about it - another rabbit hole to go down...
I switched a home cluster over to Talos v1.7.6 a couple of days ago, it took a little bit of work - my synology-csi driver needed some reworking and Cilium attempting to replace the Kubeproxy just wasn't a working configuration but after those small speed bumps it's all been really good. It really is minimal and I did my first K8S upgrade today from 1.29.7 to 1.29.8, it pops up with a lot of warnings saying that it's reading 1 vs 2 but this is just waiting for the config to converge before moving onto the next steps. Pretty flawless apart from with this config I have noticed a lot of pods left in Error or Complete on the node restarting (currently only 1 node). This could be related to Cilium terminating ahead of other nodes and blocking clean shutdown of the pods but I'll understand it better soon I'm sure.
1:30 No no no, that's NOT what that badge says. Kubernetes is certified and Talos is a member of CNCF, just like 700 other entities. But that doesn't certify Talos on any way. They're not even an active project on CNCF Projects.
Careful with those statements, please!!
Thanks for the clarification, I'll be sure to double check in future!
@@Jims-Garage No problem, thank you for the video! I will definitely give Talos a go! Your playbook looks super useful for that. ❤
Ha, I installed that on four RK1 CMs on my Turing Pi 2 board a week ago. I have yet to put anything else on the cluster though, but it was very painless to get up and running. Caveat, I did use their Terraform provider instead, I'm not a fan of blobs of YAML.
Awesome and good to know it works on lower end hardware
Welcome to today's episode of migrating current cluster to new cluster!
Everyday is a school day!
Got half way into RKE2 but wanna try this as it looks like less moving parts
Jim "here's my get out of jail free card..." YT load Ryan Reynolds Mint Mobile Ad... Ryan Reynolds is Jim's get out of jail free card.... nice
Great ! Thanks
@@chikosan99 you're welcome
what's the license of the ansible roles?
Why are the IP’s of the servers & agents different from Ctrl planes and Wrkrs?
@@yokomodo1144 nodes have internal and external IP addresses
Hi Jim, thanks for the amazing video! It works with one issue, i don't get the talosctl config under .talos. So i'm not able to communicate via the talosctl, kubectl works perfectly. Any hint? Thanks in advance! 🙂
@@LarsDanzer odd, it must create it as the talosctl commands worked. I'm not sure, check you have hidden folders enabled?
@@Jims-Garage I found it out, the config is always placed under /home/ubuntu/.talos even if you start the script with a different user. I just copied the config over and now it works. Thanks a lot! 🙂
@@LarsDanzer you'll have to change the Ansible user variable. Glad it's sorted
@@Jims-Garage Thanks! I overlooked that one!
Have you played with flatcar linux?
I haven't but it looks like a great open source alternative. I will be looking into it
Nice! Have you tried running Longhorn on it? There were some issues in the past because Talos didn't support open-iscsi which is a requirement of Longhorn's V1 storage engine. Their V2 storage engine should work with Talos because it's using the SPDK storage backend instead of iSCSI (if I read that correctly). Haven't tried it yet though, and I'm not sure if you can easily convert your existing V1 volumes to V2.
@@cheebadigga4092 I haven't yet, this is the extent of my foray into Talos. I'm definitely going to be trying it though.
@@Jims-Garage I'd love a video of that journey! :D
niceeee thanks!
Welcome!
You are using Kube-VIP with Talos? You dont need to use it because Talos has the VIP function build in. And there is a much easier way to deploy talos, you can boot it with pxe and give it the talosconfig file with pxe. After that it installs it by itself. Thats they way i am deploying my talos cluster. :)
No, I'm not. It's a relic from my RKE2 (explained in the video) and something that I am actively investigating. It's become clear (and thanks for confirming) that it's superfluous.
Please make a video on Kairos with RKE2. This isn't any good documentation or video on it. 😅
@@ayushchothe8785 thanks, I'll take a look. Bad documentation isn't usually a good sign though...
@@Jims-Garage They have an open issue in the GitHub repo regarding docs.
Are you working on a video on this topic?
Star trek tos ref in the name talos?
Ha, it could be. I wouldn't be surprised if they're Trek fans!
Not Opensource , be at the mercy of the company to not sell it someone who makes it as a paid product.
❤🔥❤🔥❤🔥
Thank you for this video. Great stuff, as always. (commenting because TH-cam prefers long-form.) 🫡
Much appreciated, thanks!