text stringlengths 0 2.29k |
|---|
**Gerhard Lazu:** Just go for it. That's very nice. Okay. |
**Andrew Rynhard:** Yeah. |
**Steve Francis:** Talos delivers vanilla Kubernetes at the end of the day, so you can run whatever your choice is of any of these capabilities. We will probably have easy defaults, so it's like the default install; unless you say otherwise, we'll install MetalLB on bare metal, and maybe let you configure a different C... |
**Gerhard Lazu:** That is a great starting point. And again, everything that I've tried so far, it worked really well. And I wasn't expecting it to be that simple and straightforward. Now, there's a lot of blanks to be filled... And that's on purpose, right? Because you can't know what CNIs people will choose. And if a... |
**Andrew Rynhard:** \[44:14\] Yeah, definitely. We've kind of touched on them already. MetalLB is definitely common. Rook CEPH is definitely common. What are some other ones? Definitely ingress controllers, obviously. The NGINX one is great. That's the one I've used. It works very, very well. |
**Steve Francis:** The usual monitoring and logging... |
**Andrew Rynhard:** Yeah. Prometheus, Loki, Grafana... Sort of the cool crowd of all those day two operations tools. Those work just fine on Talos, and they're definitely popular with Talos. |
**Gerhard Lazu:** Okay. So again, continuing on this high-level trend... What use cases is Talos known to work very well for? And if you have a couple of specifics, go for it. Steve, maybe this is something that you can share with us, some use cases that you know, that are okay to be public, maybe... |
**Steve Francis:** Yeah. Bare metal... So people like bare metal often for either latency of performance, or geographic latency. So Talos is extremely low \[unintelligible 00:45:14.17\] because it's such a small operating system, it itself uses very minimal resources, and leaves the rest for the workload... So we have ... |
**Gerhard Lazu:** Okay. Okay. We've mentioned Omni a couple of times... I think now it's the moment to talk about it. Steve, do you want to continue? |
**Steve Francis:** Yeah. Well, so Talos does make setting up a Kubernetes cluster really simple. Omni makes it next-level simple, where it's really -- so Omni is our SaaS service for the installation of Kubernetes. So the way it works is you log into your Omni account, your SaaS portal, you download the installation me... |
And then, all you do is you go into your web portal, if you want to make a new cluster, you go to your unallocated machines and say "This one's a control plane. This one's a control plane. This one's a control plane. Worker, worker, worker, worker, worker. Go. Create cluster", and that's it. |
And WireGuard gets deployed, KubeSpan is configured, Talos is installed, Kubernetes is installed, the cluster is bootstrapped... But that all happens automatically; you get a nice management GUI where you can see performance, and nodes, you can run upgrades... It's really simple, really slick. |
**Gerhard Lazu:** \[48:01\] Yeah. That is the one thing which, again, I was expecting... Because I started with the open source one, and I was expecting there to be more things to do... But once I realized that "Hang on, the image which I downloaded, that is mine, was generated for my own account", and as soon as that ... |
**Steve Francis:** No, because all the authentication is done through the SaaS account. So it ties into your authentication provider, Google, or GitHub, or whatever. So you can have multiple users going in. And if someone leaves your company, you don't have to lock down all your Kubernetes clusters and take away their ... |
**Gerhard Lazu:** Yeah. That was the first thing which got me. I was thinking "Why does my Kube CTL not work? Why does my Talos CTL not work?" And for the Kube CTL, it was like really simple. Just a matter of \[unintelligible 00:49:09.18\] install the plugin, and off you go. Because it needs to authenticate with Omni. ... |
**Steve Francis:** In general, no. Because Omni is going to be the source of truth, so it's going to reconcile the state of machines to the state that it knows about. But if you do something kind of out of band using Talos CTL and change the configuration of the machine, Omni is going to reconcile it differently, and o... |
**Gerhard Lazu:** So you're the second person telling me that. It is definitely true, because Andre told me exactly the same thing. I was like "Hey, Andre, what's going on here?" And he said, "Yeah, I mean... Omni manages that." So yeah... Even when you specify a node, like -n, you need to figure out which one to speci... |
**Steve Francis:** Yeah. |
**Gerhard Lazu:** Okay. |
**Steve Francis:** But you can get all the information from your nodes, for sure. |
**Andrew Rynhard:** Yeah, we wanted to make it -- you can still debug, of course, with Talos CTL; that's still really, really important. But when you're managing Talos nodes with just Talos CTL, you're forced to think about them as individual nodes still, where with Omni, it gets us a centralized place to think about t... |
**Break:** \[50:55\] |
**Gerhard Lazu:** So there was KubeSpan in 2021, there was Omni in 2022... By the time this comes out, it'll be 2023; the first episode for 2023. What can you tell us about the things that you're thinking about for 2023? |
**Andrew Rynhard:** Wow. We literally just had a meeting about this this morning; like, that was literally 20 minutes before this. |
**Gerhard Lazu:** So timely! So timely. Great timing. It's meant to happen. |
**Andrew Rynhard:** I don't know if I've had enough time to really digest what I should say publicly... But I definitely think things like making Talos even more secure is high on our priority list, secure boot being one of them; looking at things like integrity measurement architecture, where we can actually remotely ... |
So I would say just security in general is always a thing that's on our list. Steve, I don't know if you have anything that stuck out to you in that list? |
**Steve Francis:** No. I mean, security is the main thing, including \[unintelligible 00:52:38.06\] now, but it's not as smooth as we want it to be, or some of our customers want it to be... But in general, Omni will certainly be our focus for the next six months. It's still in beta right now. It'll be GA hopefully by ... |
**Andrew Rynhard:** Yeah, Omni is definitely in the 2023 list. I'll just add one more thing that I think is exciting for our users, for our old-school users... And that is that we're going to be looking at breaking up the Talos config into multi-doc YAML. So if you need to configure an interface, it will be of kind int... |
**Gerhard Lazu:** Wow, that is super exciting. That's the one thing which I was wishing for, but there's like a couple of other things... Because you're right, they're like individual components; it's starting with desks. So you don't always have to apply the whole machine config just to modify the disk. That makes so ... |
**Andrew Rynhard:** Exactly. |
**Gerhard Lazu:** If there's a failure in here, and was it applied... So yeah, it's a lot more atomic, and that makes a lot of sense. |
**Steve Francis:** Yeah, you said it's great to work in YAML with a straight face. So that was -- \[laughter\] |
**Gerhard Lazu:** Yes. Now, now, now... I have a love/hate relationship going back about 10 years... So I've been through all the cycles. So I've been down there, I've been like in a ditch, in the hole, and I'm back up again, and like through the plateau of disillusionment, and all of that... So I'm where I need to be;... |
**Andrew Rynhard:** Yeah. Right, exactly. I was just gonna add that it does open up some interesting opportunities as well for our users, where you could build controllers that Talos could load up very early on in the boot process, and they contain business-specific logic, and almost like CRDs, you can have a configura... |
**Gerhard Lazu:** \[55:52\] Yeah. Okay. Now, the one thing which I should say is that I did manage to upgrade from 1.27 to 1.30. And again, there's like a theme here, because I didn't realize just how simple the whole process was going to be. The only gotcha was that I had a single node. And there's a safety feature to... |
So upgrades, which tends to be a very complicated thing, was fairly simple now. I didn't have many workloads, so maybe when there's more workloads... But you have like a nice, graceful shutdown, I could see all the steps it was going through... It's really well thought through; it's as if you've been doing this for mor... |
**Steve Francis:** Andrew's been working with Kubernetes for a long time. At LogicMonitor, he was the one that spearheaded our move onto Kubernetes... |
**Gerhard Lazu:** Really? |
**Steve Francis:** ...so that was probably what Kubernetes 1.10, or something like that...? |
**Andrew Rynhard:** That was 1.7... And I think that was like my first official job in software, even though I was like studying software on my own for 10 years before that... |
**Gerhard Lazu:** Wow... |
**Andrew Rynhard:** I just loved Linux, and I think I was like six months into my journey there... And so for better or worse, I was put in charge of Kubernetes there. But it ended up actually working out really well. |
**Steve Francis:** And you got hooked. |
**Andrew Rynhard:** Yeah. |
**Gerhard Lazu:** Wow. Can you imagine not using it? Can you imagine not using Kubernetes, using something else? |
**Andrew Rynhard:** I can, yeah. Absolutely. I think it's dangerous when you start to put anything in life as the ultimate answer to everything. I think Kubernetes certainly has its pitfalls and downsides to it... I do think it's the best thing we have today. I also don't think it's for every use case out there. That b... |
**Gerhard Lazu:** Yes, that's exactly where I was going with this. Okay. Okay. Okay. So let's talk about this after we stop recording. And listeners go "Noooo...!! Keep that in!" |
**Andrew Rynhard:** That's how you're gonna keep them hooked for the next episode. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.