text
stringlengths
0
1.8k
[3455.18 --> 3461.24] In fact, our init system is a new init system that we're building for the purposes of these style of operating systems,
[3461.42 --> 3462.68] API-driven operating systems.
[3463.00 --> 3464.24] And so that is written in Go.
[3464.52 --> 3466.22] Practically everything that we do is in Go.
[3466.22 --> 3473.32] And so I think maybe that can be contributed to the fact that, you know, we are running Musil, but we haven't ran into any issues.
[3473.94 --> 3479.98] And then Kubernetes itself, since it's delivered in containers, those containers have GLibC.
[3480.28 --> 3485.64] So the role that Musil really plays in our ecosystem is very, very small.
[3485.86 --> 3485.96] Yeah.
[3486.32 --> 3487.24] I always had it different.
[3487.24 --> 3492.06] So usually the host would run GLibC, and the container would run Musil.
[3492.44 --> 3497.52] And then that combination from that direction always seems to have seemed to have create problems.
[3497.94 --> 3499.68] This was about two years ago.
[3499.80 --> 3506.56] I remember when we're looking at RabbitMQ, the image, in the context of running it, you know, at performance, at scale,
[3506.86 --> 3507.82] what's the most you can get.
[3507.88 --> 3509.94] And then you have the Erlang VM, so it's slightly different.
[3510.14 --> 3516.10] But I do remember the Alpine-based images had all sorts of weird issues that the Ubuntu ones never had.
[3516.10 --> 3517.62] Now again, this is the container image.
[3517.86 --> 3518.68] This is not the host.
[3519.02 --> 3522.64] So I'm really curious to try it out for myself and see what is different.
[3522.72 --> 3523.08] Who knows?
[3523.26 --> 3524.38] Performance could even be better.
[3524.58 --> 3526.04] Which kernel version are you using, by the way?
[3526.20 --> 3527.76] We are on the latest LTS.
[3527.98 --> 3530.30] I think it's 5.10.62.
[3530.74 --> 3531.14] Nice.
[3531.44 --> 3531.76] Okay.
[3531.98 --> 3532.24] Okay.
[3532.32 --> 3535.30] We used to run, like, latest Linux kernel.
[3535.70 --> 3538.62] And, you know, we still kind of go back and forth on what we should do.
[3538.62 --> 3545.48] And I think we're now leaning more towards LTS because the changes that Linux introduces sometimes just causes us
[3545.48 --> 3548.72] more headaches, especially when you're on the bleeding edge versions of it.
[3548.94 --> 3552.80] But the latest LTS so far has been really, really great for us.
[3552.92 --> 3558.98] We've been playing with the idea of maybe having LTS-style releases of Talos,
[3559.06 --> 3561.62] which are pinned to LTS versions of the Linux kernel,
[3561.62 --> 3566.58] and then having more edge versions, which are running the latest stable, not LTS.
[3566.58 --> 3572.32] But today, you know, we're still playing with the overall strategy that we want to take long-term.
[3572.52 --> 3575.70] And we just kind of settled on LTS for now because that's kind of a safe play.
[3576.24 --> 3582.20] So speaking about LTS and strategy and roadmaps, anything interesting coming in the next six months?
[3582.48 --> 3585.52] So between this KubeCon and the next one for Talos and Cozy?
[3585.52 --> 3590.94] Yeah, I'd say the biggest one is this week we're announcing KubeSpan, which is, I mean,
[3590.98 --> 3594.54] I'm just super excited about this idea and I haven't even explained it yet.
[3594.94 --> 3595.20] Okay.
[3595.20 --> 3595.90] Yes, please.
[3596.56 --> 3597.88] That sounds very interesting.
[3598.02 --> 3598.20] Please.
[3598.36 --> 3598.70] Yes.
[3598.88 --> 3603.10] The idea is that since Talos can run practically anywhere,
[3603.42 --> 3611.14] we're finding people want to bridge, say, bare metal clusters with instances running in the cloud.
[3611.14 --> 3614.60] And so far, there hasn't been any good solutions for this.
[3614.94 --> 3617.70] With Talos, we're kind of uniquely positioned since it's API-driven.
[3617.88 --> 3618.90] We own the whole stack.
[3619.26 --> 3621.12] We know we got Cozy managing the network.
[3621.12 --> 3627.08] And so what we did is we went ahead and we actually wrote a tooling to basically automate
[3627.08 --> 3633.22] the key distribution and the peer discovery of WireGuard VPN.
[3633.22 --> 3640.66] So I can spin up a cluster right here in my closet that's running on Raspberry Pis and extend
[3640.66 --> 3644.94] that out to AWS really, really simply, really, really easily.
[3645.18 --> 3651.86] And the latency is somewhere like, I think the latency that WireGuard adds is somewhere around
[3651.86 --> 3652.52] a millisecond.
[3652.52 --> 3655.12] And so it's, you know, it's negligible.
[3655.12 --> 3661.74] But you get this consistent experience network-wise regardless of where you're running that particular
[3661.74 --> 3662.12] node.
[3662.68 --> 3665.22] Even the pod traffic can be routed over it.
[3665.42 --> 3669.92] Kubernetes can actually be configured to purely talk over the WireGuard network.
[3669.92 --> 3676.72] And so the idea with this long-term, the vision is that we're going to have users, customers
[3676.72 --> 3681.60] that running in the data center, bare metal, which is a large part of our user base.
[3681.78 --> 3687.58] All of a sudden, they have an influx in traffic and they need to expand the cluster, but they
[3687.58 --> 3688.44] don't have the resources.
[3688.96 --> 3689.50] Okay, fine.
[3689.58 --> 3692.32] Let's just expand out to AWS momentarily.
[3692.48 --> 3696.46] And when things calm down, we'll scale it back down to our core infrastructure or even
[3696.46 --> 3698.80] another data center, spill it over to another data center.
[3698.80 --> 3702.34] Now, a completely different use case, but very similar is maybe the Edge.
[3702.42 --> 3708.10] I have some Raspberry Pis that I want to actually join up to a cluster at the core, which is
[3708.10 --> 3710.26] hosted in AWS.
[3711.12 --> 3717.56] But maybe these Raspberry Pis are running in shipping trucks and they have, you know, intermittent
[3717.56 --> 3719.00] network connectivity.
[3719.24 --> 3722.86] That's kind of troublesome when you're talking about running the Kubernetes control plane.
[3722.96 --> 3725.38] But a worker, you know, it can kind of go in and out.
[3725.38 --> 3730.92] And, you know, I think the story there could be better on the Kubernetes side, but at least
[3730.92 --> 3735.34] using WireGuard, as soon as they get any kind of networking, whether it's, you know, some
[3735.34 --> 3741.60] Wi-Fi when they pull up to a store or mobile data, they can join the cluster with WireGuard
[3741.60 --> 3745.04] and everything just seems as if they're right there on the same network.
[3745.38 --> 3746.14] That's really interesting.
[3746.14 --> 3748.62] So let me see if I understood this correctly.
[3748.96 --> 3755.54] You're saying that you can scale out your Kubernetes clusters on demand, wherever, whether it's your
[3755.54 --> 3760.32] closet or whether it's not the data center or the cloud, you can maintain the same privacy
[3760.32 --> 3761.10] of the network.
[3761.62 --> 3762.34] Everything is encrypted.
[3762.64 --> 3766.72] The data on those workers, you would think it's ephemeral data so that you don't store any
[3766.72 --> 3768.88] state there so that you can scale back in.
[3769.24 --> 3772.10] And KubeSpan makes this seamless.
[3772.52 --> 3773.30] Is that what you're saying?
[3773.30 --> 3774.68] That's exactly what I'm saying.
[3775.12 --> 3778.82] Of course, there's, you know, little caveats, like the way WireGuard roughly works is you
[3778.82 --> 3781.30] need at least one direction of communication.
[3781.66 --> 3786.74] So in the case of, say, my private cluster running right here in my closet, it needs to
[3786.74 --> 3789.24] be able to at least reach the workers.
[3789.38 --> 3791.32] The workers don't necessarily need to reach it.
[3791.44 --> 3794.06] It can establish the channel that way.
[3794.62 --> 3799.32] And so there are some, you know, limitations within the system you can find in the documentation