Red Hat OpenShift Ask an OpenShift Admin | Red Hat Livestreaming, 19 Jul 2023

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: Ask an OpenShift Admin (E106) Next-Level Cluster Provisioning

Description

Andrew and Jonny are joined by Senior Principal Software Engineer's Avishay Traeger and Eran Cohen to talk about next-level cluster provisioning with Hosted Control Planes (aka Hypershift) and the Assisted Installer.

Want to learn more?

Try Red Hat OpenShift: https://red.ht/3uVxkya

Follow us:
Andrew Sullivan https://twitter.com/practicalandrew
Jonathan Rickard https://twitter.com/jrocktx1
OpenShift https://twitter.com/openshift

Subscribe to Red Hat's YouTube channel: https://www.youtube.com/redhat/?sub_c...

About Ask an OpenShift Admin:
Do you have questions? We want to answer them! Every Wednesday at 11am ET, hosts Andrew Sullivan and Jonathan Rickard sit down to answer your questions about Red Hat OpenShift.

About OpenShift:
Red Hat OpenShift is an open source container application platform based on the Kubernetes container orchestrator for enterprise application development and deployment.
OpenShift on redhat.com https://red.ht/3rYz5ZS
OpenShift on GitHub https://github.com/openshift/

#RedHat #OpenShift #Kubernetes #ContainerPlatform

A

A

Let's get into the big topic of Open Source.

B

This is so awesome.

A

C

Hello and welcome to this week's ask an openshift administrator office hours live stream. As always a mouthful and I am co-host. Andrew Sullivan joined by my co-host, Mr Johnny Ricard. How are you today Johnny I'm, good co-host? How are you yeah, you know it's uh it's it's a joint thing right. It's shared responsibility, we're you know either we both get hit by the bus or neither one of us gets hit by the bus. Oh.

A

C

You know, or you know we both go down with the ship I, don't know where I'm going with that you.

A

C

Be a little less morbid, you know yeah, so uh hello, everyone welcome to uh the stream. This is uh episode 106., it's still hard to believe that we've hit that number uh so far and yeah, uh despite a ludicrously hot week here in uh the Eastern or Central part of the us for Johnny, who is letting the dog out I think.

A

C

Go with the dog in like they're yeah how's, all that hot air coming in yeah.

A

C

Yeah, because it is a brutally uncomfortable here, I'm trying to avoid it no distracted because I have uh I have another video of us up and now I'm, seeing you in double so I gotta you're welcome, you're, welcome, I gotta, move that yeah.

A

It's like 90 something Fahrenheit right now with a lot of humidity, so it's pretty miserable outside for us.

C

Yeah, well, it's it's storming here, I'm, not sure which is worse. Oh yeah yeah there's a bit of a pretty good crack of thunder a few minutes ago, but anyways uh so hello to our audience. Thank you for joining us this week. Again this is episode 106. uh and this week we're going to be focusing on something that we've touched on. We've talked about in bits and pieces and that is hosted control planes.

C

So this is something that is um near and dear to my heart, uh so I'm a virtualization person and there's lots of cool stuff, that's going on there and how we can optimize and increase efficiency.

C

But today is special and today is different because we're going to be talking about doing or using hosted control planes with bare metal servers, so I won't steal. The show I won't uh I, won't uh uh uh uh read, read the tagline or gosh. What's the word I'm looking for the headline right out front, instead, I'll, let our guests introduce themselves and talk about today's topic or introduce today's topic. Rather so um I. Don't remember who I'm handing off to uh I? Don't remember if it's Iran.

A

C

D

Take it so hi hi everyone, I'm, avishai, I'm joined with Iran, um we're both singers principal software engineers and the ecosystem engineering team, and we actually joined Red Hat together a bit over three years ago. After working together on the same startup, and that's we, together with our team, we led the development of on-prem openshift deployment methods such as the assistant, installer, zero touch provisioning, and what we'll talk about here? uh Hypershift on-prem uh Iran, also spearheaded single node, openshift and lately I've been working with strategic Partners, particularly in the industrial space.

C

You'll have quite the uh the CV there, quite the set of credentials for going back with a lot of really cool Technologies right, single node openshift is a a fan favorite, oh yeah, so as well as joining together. That's pretty awesome. You know I think we said before Johnny and I worked together a long time ago, and then we both in a very circuitous way, ended up at Red Hat together. So right, it's uh it's always fun to see people in the you know or reconnect with people right that you've worked with before.

C

So that's awesome.

B

Sorry connect we never.

D

We never disconnected yeah. That's.

C

Even better, it's even better. You know. Congratulations on that.

A

You guys have like matching resumes with the other side of the fist bump. You know like just show up together. It's.

D

Actually about 14 of us uh joined together from that company to here nice. So it's pretty nice.

C

I like how that goes, I like it when we can get uh lots of good people, you know coming from other places, so red Hat's a great place to work. uh Let's see what was the guy I've completely derailed and lost track of mine, because there's a dog staring at me through the glass down here so uh anyways. uh So this is one of the office hours series of live streams here on red hat live streaming, which means that regardless or irrespective one of those isn't a word, I think it's irrespective is in word anyways.

C

uh Regardless of today's topic. You're welcome to ask us anything and everything that is on your mind. So for the audience, please don't hesitate to reach out through chat through whatever it is that you're watching us on and send in any questions uh they all get aggregated up and sent over to us and we'll do our best to address those while we're here on the Stream. If we can't answer your questions, which I'd like to think Johnny and I are pretty smart and we generally know what we're talking about.

C

But uh we don't know all of the answers. We're happy to reach back into well like today, into the engineering folks um or into product management or kind of anywhere across the product in order to find those answers, so don't hesitate to reach out and that's whether or not or whether you're watching us live today through chat or if you're not watching us live. You can also contact us via usually emails the best way, so you can reach me at andrew.sullivan redhat.com.

C

If you are watching us, you can see that on the screen there or at practicalandrew across all of the various social media platforms and all those things, I will admit that I am usually terrible at actually looking at things like Twitter, because Twitter, but slack is also a great way to reach out to us and which one of these is the slack one. So if you go on to the kubernetes slack, there is an openshift users.

C

Channel there that Johnny and I and a whole bunch of other folks are on that's another great way to get in touch with us. You can reach Johnny again via email and he is available at Johnny, j-o-n-ny, redhat.com or Jrock TX. One on the various social platforms and last but not least, I'll also make a plug for openshift Commons. So the commons, which is the openshift user Community, is they also have a slack team, yeah yeah, they're teams, right I, don't know: I can never keep these term. This terminology straight workspace workspace.

C

There we go so they have a slack workspace at openshiftcomens.slack.com that anybody can join and it's another great community of folks to participate in yeah.

A

And and I think it was our hope. Nine last week brought it up that you had to have like a Gmail or an Apple ID or something like that to join it. So I brought that up to the uh to the lead and they were looking into it so give it a shot. Let me know if it's still an issue uh and if it is I'll, bring it back up and try and get it resolved for.

C

You, okay yeah, that must have been uh I'm guessing Karina, so I wouldn't have.

A

C

B

C

Many different ways that you can reach out to us, so we try and stay on top of all those things and uh get back to you with questions and or answers to questions and all that other stuff. So again, don't hesitate to reach out if there is anything that we can do um so yeah. So let's talk about some top of Mind topics. Johnny.

A

Yeah, so I are just a you know. Our schedule is looking awesome. We've got a lot of stuff going on, and um so we've got trillio coming out on the 2nd of August.

A

We've got um openshift data science, uh the roads uh coming on the 26th of next week, and then uh we just got confirmation that we're gonna have the uh product manager for Scupper or red hat service, interconnect to come on, uh and that's going to be us at some point in August um and then Andrew and I are going to do like a uh like a split session where we go over all the things that are in 4.13 since we didn't have, we haven't really had an opportunity to like dive into it, so uh we're gonna dive into that and we're gonna do some more hosting control plan, stuff and just kind of walk through the hypershift CLI and just nerd out a little bit uh and then finally, in September, we just got confirmation about this last night as well is we're gonna have ortwin and uh one of his Partners come on and we're going to talk about mesh lab and mesh and production and stuff like that.

A

So that's another exciting topic that we got coming up um and then, finally, for me, okay, I, said finally I actually have two more um last week, istio graduated. It became a graduated project in the cncf. So congratulations to uh to the istio project. That's awesome uh and Cooper 1.0. uh They. uh You know they're big time now. So you know they've gone from um like a small project. They've got their 1.0 release and if, if you're not familiar with kubert, it's it's openership virtualization for the red hat people. So it's it's awesome.

C

Yeah Cooper is the Upstream for openshaford, so.

A

C

It's uh there's a blog post for that, isn't there yeah.

A

I think there is I'm getting the link right now, sir, so.

C

Yeah they um there's lots of really cool stuff, that's going on there and I'm, seeing more and more other organizations uh or other providers, vendors, adopting and integrating covert into their various offerings as well. I want to say, Platform 9 offers something now so yeah. Just lots of really cool features. Lots of really cool stuff happening in there. uh I think it was episode, 96 Johnny, it was when we had Ian on or 90 years in the late 90s we had Ian Pilcher on and he talked through a whole bunch of openshift virtualization stuff.

C

So uh it just yeah really interesting stuff lots of lots of fun stuff, that's happening. There definitely keep an eye on it. I.

A

Haven't dug into it, that's another one that we should do just where me and you just nerd out on opposite virtualization, because I haven't really messed with it. A whole lot, so it'd be kind of fun to dig into it. I have messed.

C

With it a whole lot.

A

um Oh yeah, so ever yeah.

D

A

D

Play with hypershift and you play with uh openshift virtualization, you can actually do both together.

D

C

That's something that we've been um I've been talking with the product management team about in the background, if we should definitely whenever their kubeferts or uh the Cooper provider for hosted control, planes becomes available, which I think it's Tech preview. Now we should definitely have them on the Stream because yeah it you know it's openshift hosting openshift and it's in an integrated manner. So it's it's really cool stuff.

C

That's awesome! um Yeah Martin! It is a lot of fun. I was actually talking with uh Eric Hendricks Eric reached out to me um so he's the host of the Rel live stream over on the red hat Enterprise Linux YouTube channel- and he was asking me questions he's like hey I, replaced my entire home lab with openshift virtualization like all right, let's go so he was asking me questions about.

C

You know kubernetes stuff, and you know how to learn as a Rel admin, how to learn about kubernetes and openshift, and what all this stuff means and how to deploy. Vms and stuff like that, so yeah lots of lots of cool stuff in there uh Duane I'm, not sure I'm gonna find it right now, yeah. So there's a short link that you can use to go and see all of the um the the it goes directly to the playlist um for ask an openshift admin and it's red, Red, Dot HT, slash AOA,.

C

So 99 was VMware.

C

Has it been that long since we had Ian on man, yeah episode, 89 wow time flies when you're having fun and getting old, so There You Go episode, 89 was when we talked about openshift virtualization, um so I had two things on the list here, um so one uh so an account team reached out to me and then I ended up communicating with one of our customers over some confusion about how or or when a cluster is provisioned using assisted installer, and when you provision that cluster, when you select cluster managed networking.

C

So let me see if I can bring this up. uh Let's see, I want to share my.

A

C

The giant red one.

C

So if I go to console.redhat.com openshift.

C

And wait for it to load. Come on internet, so I'll create a cluster I'm going to create the cluster using assisted installer. So we'll just call this one AOA dot newa.com, because I don't need it to actually be valid. So if I select this DHCP only and then I go to next and next so I don't think I'll actually be able to get to the to.

A

The uh yeah it's.

C

It's on the networking option um with so without joining hosts I won't be able to show it, but on the networking tab, there's an option for either uh cluster managed, networking or user managed networking.

C

So if you select user managed networking it's effectively exactly like, you would expect, with any openshift cluster right, you're responsible for provisioning, the load balancer for managing the DNS entries for startout, apps and API, and all that other stuff. If you do, cluster based networking or cluster managed networking, even though it's a UPI install or a non-integrated install, not an IPI install, it will use the integrated keep alive. D virtual IP addresses for those quote-unquote load, balancer endpoints, so startout, apps and API.

C

So the problem or The Quirk that happens is when you do that, the nodes, and again it doesn't matter if they're, actually physical, if they're virtual machines, it doesn't matter where they come from or where they're hosted the nodes when they join the cluster, show up as bare metal nodes, and it's a quirk of how you know, because it's using keep alived and it's using those VIPs that it just assumes.

C

It's like a fall through thing, where it assumes that they're bare nettle nodes that are, you know, part of a a platform bare metal type of cluster. So, even though it's platform, none non-integrated, they show up as bare metal, so.

B

This is important, go.

C

B

Want to maybe uh add to that, so you can actually so whenever you say, I want to get user provision. Networking you're essentially saying I want to get non-platform and if you're, if you're, choosing the cluster management working, then it's going to be somewhere in between. As you said, bare metal and now like, we are going to use the people ID and we're going to create VIPs, and if we notice that it's running on, let's say openstack, we'll say: okay, you can't really do that.

B

But if we see that okay, it's it's libert or something that we know that should work then we'll allow it. And if we actually notice that it's this year, then we'll offer you. Okay, you can do the vsphere integration, you don't have to use bare metal.

C

Yeah yeah there's a toggle I think it shows up on the next either on this.

D

One or the next page.

C

D

A great degree with virtualization.

C

I thought that one was to deploy the uh openshift virtualization operator, which is actually on the previous screen right.

D

So now we can integrate with vsphere nutanix and we just added uh Oracle.

D

C

That's cool kind.

D

Of if you choose cluster managed it's the bare metal platform and that's why we're seeing yeah bare metal artifacts there.

C

Yeah, so it's slightly confusing but harmless yeah anyway, I was talking with the customer and they they had wanted to add. They deployed their cluster, using assisted installer to using virtual machines and the that UPI cluster managed networking scenario, so they showed up as bare metal and then they had wanted to use the bare metal IPI experience to add physical nodes into the cluster and it just because it's a platform, none cluster, even though the nodes are showing up as bare metal virtual nodes are showing up a spare metal. It just doesn't work like that.

C

So slightly confusing slightly. um You know not not perfect, but.

B

Maybe something to add today, so you can actually uh that they do work on those and make Masters a bit more complicated, but you can do it from SAS. The only Gap is that you'll need to approve the the node CSR. The certificate setup question yourself: uh yeah.

C

You just can't use the metal, cubes or metal three integration to automatically like the you know, with machine sets provision and integrate those nodes. um Yeah, you can absolutely do the standards. You know IPI method of manually, adding nodes or use the assisted, installer interface to to provision those nodes join those nodes both of those work exactly as expected.

C

um What was the other thing? Oh the other thing, real quick, um so somebody had asked on the uh internal answers platform about the size of the scg database, um so the NCD database is so one I think they test up to something like eight gigabytes total size for the database and I. Think it's up to one megabyte and I, don't know if that's an NCD limitation or just what we test, but up to one megabyte per record size.

C

So this is important for a couple of reasons, mostly because people want to plan around capacity and stuff like that, so A lot of times, we will pay attention to you know, especially when we're provisioning clusters we'll pay attention to the disks that the control plane nodes are using. uh Generally speaking, SCD isn't a a huge concern when it comes to the the capacity it's the latency, that we want to pay attention to I always reinforce with folks I always tell Folks at CD.

C

Latency is by far the most important metric when it comes to etcd secondary by a long shot is iops because, again the largest record size is going to be a a single megabyte. So it's usually only a few iops to do each one of the operations, but it needs to be low latency, both at the disk level, as well as at the network level, and we've talked about that in depth.

C

In other other episodes, including I, think when we had the NCD folks on back in I, don't even remember what a stream, what episode that was Johnny, I removed, see I could probably find it here, real quick that was episode, 21.

C

wow, that was a long time ago. We should probably revisit that topic, since it's been that long 21 would have been back uh almost two years ago.

A

Yeah before my time, yeah.

C

Yeah, you came on in the 40s, so right, old timer. Now that's.

A

I know I know, Season bet.

C

All right, that's all I got uh I got nothing else. So, aside from I forgot to mute all my notifications, so I'm listening to a bunch of dings and Bings and clicks and Clacks and stuff happening in the background. So now I got to do that. So you want to kick off today's uh topic. Yeah.

A

All right Ron have a shy, let's get after it. This is this is an awesome subject. Right, like we've, been excited about this I I am I'm a huge fan of hypershift I I really started kicking the tires with it a couple weeks ago and um yeah I'm I'm like super impressed with it there's one thing: I have one grumpy thing, but I'll take that offline, but other than that, like it's, it's an awesome project and like uh I'm I'm super excited about where it's go and I love. This topic today get after it.

D

Run you're, muted, yep.

C

B

Was beautiful, muted,.

C

In the platform, I admitted you.

B

Okay, cool I wanted to say that we're responsible, mainly for the integration with hypershift or between hypershift.

A

And the system installer.

B

D

Yes, I'll just give a quick introduction to hostel control planes for whoever is not familiar. uh You might have also heard of it called hypershift, that's the Upstream name. So so, let's say we have a standard openshift cluster. We call it the management cluster and then we want to add additional clusters in our environment.

D

So what hosted control points does is actually runs the control planes for those additional clusters as pods as workloads on this existing management cluster, um and so why would you do that? So, first of all, bringing up pods is much quicker than bringing up actual machines, so you can get your control plane up in, let's say 10 minutes instead of 30 40 minutes, um it's more cost effective.

D

So instead, especially with.

A

The bare metal so.

D

With standard openshift, Standalone open shift, you're using three dedicated machines for each control, plane and here, you're kind of packing them together and there's a whole lot of other benefits that you probably heard about in the posted control. Planes episode with ease of management, multi-tenancy and so on. Not gonna really go into that, but each of these control planes is running in its own namespace on the management cluster.

D

We'll see that in action in the demo in a minute, um then we have a resource called the node pool, which is managing the worker nodes and those worker nodes are external to the management cluster. They can even be on a different network than the management cluster.

D

um This thing is called the hosted cluster, the hosted control, plane, plus the workers, foreign and that's basically hosted control planes in 30 seconds.

C

Yeah and I want to point out and kind of reinforce a couple of things here of so the the hosting cluster or the managed cluster I. Think is what which you will.

D

Management, cluster management, cluster.

C

So it can be physical or virtual; it doesn't matter whether and it can manage or or host control planes for physical or virtual clusters right, it doesn't matter and those control plane services are deployed as pods into a namespace in that cluster. So I I think this is important for a couple of reasons. One.

C

It means that it's really fast to deploy those new control planes, we're not having to wait for a virtual machine or physical server to get deployed. You know not having to wait for core OS to be, you know, deployed and configured and everything to stand up and find each other and do all those things right it. It just deploys some pods and is pretty quick to stand things up, I think the last time I used it. We did a whole cluster provisioning and you guys probably know far better.

A

C

What this provisioning times are, but it's like way less than 10 minutes most of the time, uh so it's deployed as pods, which is really cool for a couple of reasons. uh In my opinion, one.

C

It means that we can decouple control, plane, scaling from the node sizes, so as that hosted cluster as it's putting more Demand on the control plane, it's consuming resources in the management cluster I, don't have to go through this rigmarole of you know: I need to remove a control, plane, node and then resize it or deploy a new one, that's of the new size and then rejoin it and go through the whole process.

C

Rather it's just you know at worst I'm having to to add a new or a larger node onto the uh the management cluster, particularly if you have a lot of clusters, a lot of cluster churn right that can be kind of a big deal. That's.

D

Right, your control points are now kubernetes workloads, so you can scale them up scale them down. Yeah.

B

It also decouples the platform. You can run the control plane on x86 and you can add workers to it on our machine on dpus.

C

Yeah yeah I, don't know whether or not it's supported but I remember when I was uh using openshift or uh hosting control planes. This has been like six months ago. I had my management cluster deployed in like U.S east and then I had other clusters deployed in U.S West, like multiple regions in in US West, and they were deploying managed clusters from there and like it was this whole tearing out thing and I don't know if AWS is is a target platform. I know it's used by Rosa yeah.

D

But so right now, HOSA control points. This Tech preview, it's shipped as part of multi-cluster engine and ACM, uh aiming to be GA in the next version, and it's available as Tech preview on AWS self-managed, so running it yourself or as part of Rosa, as you just mentioned, Tech preview as well.

D

um Ibm cloud is also using it for their managed service and you can also use bare metal, which is what we'll show and Cube verb. So those are the very kind of plugins that are available.

C

Very cool, thank you.

B

C

B

C

B

Sure screen and.

B

Okay cool, um so here you can see the management cluster. Okay, simple, open shift. As you know, I installed multi-cluster engine on it, and that gives you this cute button that allows you to go to this amazing console, and here we can see the multicaster engine UI, and so first thing I want to do is I want to carry the host inventory.

B

Okay. So for that I'm going to create an infrastructure environment. This is where we were going to say how our infrastructure look like I'm gonna call it my yes, my agents.

A

B

Local, we can add labels to it. If we add a label here, it will be added to all the hosts that are booted with this infrastructure environment. The point of infrastructure environment is essentially to create a.

A

B

For us to put the host with um I'm gonna need my post circuit for a second. So sorry, can you please thank you.

C

Yeah, we don't, we don't want to have another one of those incidents.

A

That's right because then it's not easy. I.

C

Think both Johnny and I have have accidentally exposed or pull Secrets before so.

D

For one person, so if you guys know assistant installers, so it's that same we're now creating that same ISO that you would create over there um and we call that the infrastructure environment.

C

B

So here, I added my SSH key just in case I'm gonna have to SSH into the node and understand what's going on so here we got the infrastructure environment and within a minute this should go away now. I, like.

A

B

Add a host to it, so there.

A

Are a few options you can even download the.

B

Iso: okay,.

B

C

And so, if I remember correctly and I admit that it's been a minute since I used the infrastructure environments inside of ACM, you can so you can create that ISO and boot the nodes to it, and they just kind of go into a waiting pool to be assigned out to clusters.

D

C

I I think that's really cool for and I see all the different methods for booting there I think that's really cool for a number of different reasons. uh You know I, remember: I was a VMware administrator for a long time and we would go in when we were doing provisioning and basically boot every one of the servers and just let them pixie boot off of the network and boot into you, know: uh esxi, Boot and install into esxi, and this is kind of a very similar capability right I can boot.

C

My host, however I choose to they all go in and they register into the interface and then, as the administrator I can go in and create clusters. However, I want right, I want this node, you know nodes one three and five to go in and be control plane, nodes for cluster, a I want nodes, two and four to be worker nodes for cluster Z and all this other stuff.

C

B

C

Like that experience.

D

This is actually the assistant server, the assistant installer, that's running on Console red hat.com. It's actually running here as part of multi-cluster engine as an operator, so it's 99 the same code. The workflow is a little different because in the in console reddit.com we have a simplified flow where you create a cluster. You add hosts to that cluster specifically, and here we give them more uh long-term view where you can create a pool of hosts and kind of attach them to different clusters.

D

These hosts that are going to come up. We call them agents here, you can attribute them or we call it binding. You combine them to a regular openshift cluster or it's a hosted cluster.

B

Right so here, I just show you how you can create one with uh with the BMC credentials. You can also just apply uh yamos, so here I have uh yes there we have the management cluster, keep config and I. Have this.

B

Which has all the dmh yeah.

D

A

Bare metal hose.

D

Is part of the bare metal operator using metal, cubed and ironic to put these hosts okay awesome.

B

D

And then as I create them.

B

They, oh it's fun,.

B

Okay, I guess it needs to copy them.

D

And we have some really deep integration with metal, cubes and ironic, so it's actually putting the iso automatically.

A

And ironic is the project that uh that's what openstack uses that for their provisioning as well right.

C

All right, okay, yeah and I was just going to ask. If so my understanding of metal, cubes metal, three I, think it's metal cubed. Is it effectively uses the same mechanisms as ironic, so you give it information to connect to the BMC or I, think it'll use redfish or a couple of other different methods. You know ipmi um and it just it programmatically through the API, interacts with and configures that host.

C

So things like, like you, said it attaches the iso and it Boots the node, the physical node to that, so that it can register and join or with IPI we're doing things like installing core OS and configuring other aspects of that node. In order to join um see what is it? Is it the? uh What do we call the poison pill operator now, um self-node remediation, node self, remediation operator, I.

C

Think uh where it'll use that same thing to you know: it'll it'll talk to the road to the node via metal, cube via The BMC and forcefully reboot, the node. So that way we can authoritatively determine that it has been disconnected. There's no PVCs connected and restart workload. So really.

B

Cool stuff: okay, it's using it's using redfish and apparently what it's doing ironic booted, the machines, it's running its agent on it and it's inspecting the hardware. uh This is actually new to 413. So far, it was booting the machines with an ISO we generate and it was bullying and it was running the assisted service agent, as it says, dollar agent, and that was the the the code that was inspecting and doing all the inventory collecting all the inventory from the host and right now we have a new thing. It's called.

B

We call the convert flow where both ironic agent and the assistant agent can run on the same on the same host. They do some handoff between them. Ironic is doing the inspection and later on, it runs the assisted, installer or the agent to perform the actual installation.

B

This actually enables a new cool feature like the pixie option we saw earlier and also enable enable the use the customers to do firmware, setting bios upgrade all that you can now. We can now support that um I.

C

Didn't know that, how do you how so I'm going to tangent for a moment? How can you configure those that type of stuff done through machine config.

B

No, so if you're like not in this, like the the way you do it is, when you apply the bare mineral host yum, you can add uh you, you need to add uh how's it all uh firmware settings I, don't remember uh something for more settings. Okay,.

C

We don't have to dig into it, I would just that's really. Cool I did not know that we could do that, and that's that's exciting to me.

B

Yeah, it's just one, one more uh crd that is available to you and you can. You can choose to configure or operate the firmware with that.

B

uh So right now it's provisioning and soon enough it should be available for.

D

B

uh We can go ahead and start create.

D

B

And we'll see them as available once we once we'll meet them. Oh okay,.

A

Are we able to do this in, like disconnected environments as well like is.

D

A

All this supported completely disconnected yep.

B

A

B

Hypershift there are some gaps like what we see here. Works in this connected. There are some gaps in hypershift, but we're we are. People are nailing it down. uh These days like this should get results. Okay,.

A

Awesome I got asked about it last night. That's why I was brought it up so.

D

So here you can see in the UI. It should remind you a bit of what the assistant store looks like in the SAS yeah.

B

D

Coming up in the inventory.

B

And you notice that the host name got got updated because after the I've already did the inspection and the hardware information the and checked all the hardware I noticed: okay, it's it got a different host name than what we thought it's going to be, and it just updated here. uh You can actually change it if you like, and it will set that on the host. And now we see that those horse are available to us. It has the hardware chat, the networking checks and we can go ahead and create create a hosted cluster.

B

So here we see the local buffer, the one we're running from the management one and if I click here I got multiple options, uh so we can create one and a level.

A

B

Gcp we could use red virtualization that you mentioned, but we are going to use the host inventory and I think this is something that will look very similar to you if I choose, standalone and I want to use existing Host. This is a very close to the same UI you showed earlier of the assistant, uh but we're not going to do that. uh We will use okay, we will use the control plane.

B

Okay, so let's carry the cluster in my cluster I'll choose to use the default password set. This essentially determine what OS we're going to put the host with uh here. I'm gonna use redcat.com we're gonna, install 4 13.4 again. Can I ask it please hi it's great.

A

So, while you're doing the copy pasta, um the 4.13 is that, based off of the um the the hosted clusters environment that it can deploy or like ours, you know what I mean like is it? Is it kind of like the Cincinnati operator, where you've got like the graph, where you can kind of go through like a stable Channel, or is it like, basically like you, have to do hosted cluster level and greater like what like? What's the it says, ocp version there like I?

A

Imagine that's a selectable field so like where what are the fields that I can select.

B

So assistant service gets its own configuration for what openshift we we support or what images what release images it knows in this. In this, let's say: self-hosted, environment, um yeah, sure, yeah,.

C

B

Click next from the previous screen, I added my cool secret and I click. Next, and so there are a few constructs for to host the control plane. There's the hosted, control plane and there's the the node pool and both of them together uh composed a hosted cluster. So earlier, I created the hosted control plan and now we're looking at the notebook here we can choose the namespace that we want to consume the pages from we have only one so we're gonna miss that one we can filter which agent we want by labeling.

B

If we choose this one, we're gonna only get the agents that have their specific Library there's only going to be one. We can filter them by location. All those labels are automatically added. I think this is something I wish. I did a while back um by the hosted inventory. Once we see the inventory, we just add labels that reflect uh what this host looks like to us and then later users can use it to filter uh the host, but I'm not gonna. Do that.

B

D

You can either use the pre-generated labels that we have on agents, you can add them manually yourself or you can even create rules based on the host's inventory as to what labels will be applied, so you could say like if it has you know, between this much and this much CPU and RAM, then it's a medium label right and that's, then you can, in your node pool, say I want workers that are medium or you can say I want workers that are in this availability Zone.

D

So it's filtering uh option is really useful. For you know you could have you can even have multiple node pools for the same hosted cluster yeah. So.

A

As you can say, I.

D

I want some that are medium and I want some that are large or I want some in this region, some in that region. Okay and that's.

A

Really, how you set up the multi-tenancy aspect of it like where someone wants to come in and say like.

C

D

A

I just need a node really quick over and what the you know, this type of CPU or this type of Ram or whatever, or this much RAM, and then they can just select that node pool and then go. Is that kind of the idea.

D

In terms of multi-tenancy I'm,.

A

Thinking like you've got this cluster right. You've got you've got you know a team right and like every like, hey I, just need to spin up a cluster really quick to to do something. That's that's my idea of playing multiple sentences here, just more of like the self-service.

D

Yeah, so you can just spin up a posted cluster for them with that one worker, okay,.

A

D

We'll get in a few minutes.

B

I think this is what we'll do now we'll spend obviously faster for them with just one worker.

B

Okay. Here we got the.

D

B

uh Part of this, uh and so I'm gonna use a note port to expose this, uh the Jose cluster API terrorists.

B

um We can configure address networking options, but we can also skip that I'll. Add my SSH key again. This is in case. We want to be able to SSH to the nodes of the resulting cluster. The worker now click. Next, you can, if you want to create some automation, that multiple searching will apply on your resulted cluster and you can you can create an Automation and and refer to it here and it will you.

A

Know deploy you different operators.

B

You want whatever you're configure in your own information, and here we can review everything and if we're happy with it, we can go ahead and click create and cool okay. So now uh the whole system properly operator is going to create our hosted control plane. So.

A

Don't be scared.

B

As all of these will turn on green soon enough, uh but you can see all the status condition of the hostel control plane, reflected here and essentially.

A

At this point,.

B

You can actually, you can see that the notebook is still pending on the assignment. This is because it doesn't have an English available ignition anyway. What we can do now, while this is, let's call it bootstrapping, we can go and look at the controlling pods. So if I click here, I go back to the manager cluster, uh regular openshift console UI.

B

Hopefully it is a load faster, okay, yeah.

C

It's taking its time, it's okay!

C

So so far we haven't had a Johnny level event happen in inside of the environment, so yeah.

A

Everything's working.

B

A

Far so far, okay.

D

B

You want to walk us through on, what's running what it's doing, because they think.

D

So here we're looking at the namespace, that's hosting those pods that make up the hosted control plane.

D

um We can see the the first two pods are for copy or cluster API. So that's a kubernetes sub project. That's dealing with cluster lifecycle and mostly control planes uses that, in the background kind of internally to manage the worker nodes, so Iran.

A

And I created a.

D

Cluster API provider that links hypershift to uh the assistant, installer and the cluster API providers are also used for the AWS integration and cube root, integration and others.

D

um So so that's really cool like utilizing this open source Upstream for kind of dealing with our workers and managing them, bringing them up taking them down. Auto scaling, yeah.

B

Start up: uh what's the control plan with fighter Bill, yes, this Cappy provider will be AWS can be, polite uh can be provided. Yes,.

C

You all probably know this better than I do, but if I remember correctly, wasn't machine API, uh more or less a fork of cluster API.

D

B

Okay and I think openshift wants to let's think at some point: transition back to cluster.

C

Api I mean we kind of are right, I mean you've, got cluster API here for hosted control planes, so we're kind of getting the Best of Both Worlds. Here.

B

Yeah uh other control plan parts that run here is a CD that we've seen earlier. They keep scheduler the cube, API server. uh It's all running here. The controller.

A

Manager so essentially.

B

The control plane plus, let's say connectivity server, which is the proxy that allows the control plane to talk with the worker nodes and everything gets resolved as if they're on the same network.

D

um Another thing we can see is that here we have, for example, one copy of etcd, but you also have the option of running it highly available with three copies and um and you you would have anti-affinity between them to run on different nodes. So you have that flexibility.

C

And that's defined through the CR, the custom resource that's created.

A

B

Yes, you can ask for an h, a highly available, uh hosted, control, plane and you'll just get.

A

Three replicas two replicas from whatever.

B

You need um okay, let's go back to MC and now we can see that the node pool is not ready. Oh it's already. Okay, oh I thought I thought this will take longer. Okay,.

D

We talked for two minutes.

B

Okay, it's installed okay, so essentially, once it has an available Edition endpoint, uh the copy provider is so cluster API will generate machine deployment and machine and a machine, and then the Cappy provider will try to reconcile that machine to make the machine Happy by binding an Asian machine. So all of these are different crds that are trying to make cluster API Happy by essentially making one of our available agents to join this cluster and sold. This are linked together and the agent machine finds an available agent.

B

Then it binds it to the cluster and and once it's bound to a cluster. The assistant installer will start installing this node actually to First perform few validations like it will make sure it can reach the ignition endpoint so make sure it can actually download the the few images of this release just to make sure that the pull secret that we added is is valid for this uh kind of installation.

B

It also performs some checks about like the disk space and other networking chats and once all that passed and we're happy, then it's gonna Kickstart, the installation.

B

uh It will write Arcos energy to the disk with what we call a point, recognition that will point to the ignition endpoint and then I'm gonna, tell ironic uh okay, I'm done! You can go ahead and reboot and at this point ironic will reboot uh I think soon enough. We can go back to our host inventory and see it here that we have four hosts and one of them is bound to this cluster and it's installing one of the Opera are available.

B

uh But uh if we we don't actually need to wait for the host to join. We can already talk to this cluster. So let me download the cubeconfig and.

B

A

B

Everything is already running, we can actually talk to it and the worker will know once it wants its original. So here I'm just gonna edit, the sage here perfect. With this it gets downloaded so on the lower experience. I'm just gonna use the.

B

So this is talking to the to the hostess control, plane and.

A

You can see that okay, it has like one pod.

B

Running but there are no notes yet, um which is kind of cool. uh What else can we ask it.

B

So we're talking to the API and we're still waiting for I, don't know to start.

C

So one of the things I wanted to ask about and I don't know. I I seem to recall that in the no I'm not gonna now I'm not gonna. Remember the name of the custom of the crd for provisioning, the cluster not node pool. um Is it manage cluster so.

B

It's actually cluster or hosted control, plane. Okay,.

C

Manage cluster is the ACM object. I think so. I seem to recall that in in that crd for the hosted cluster, you can configure some properties of the deployed cluster as well, so things like authentication and some other basic properties which to me one I, think it's interesting that you can completely programmatically it's just the kubernetes API deploy and manage clusters as sort of that cluster administrator or super administrator type of role. Two and Johnny will like this.

C

You can also do it all entirely get UPS based right, so you can use get Ops Paradigm in order to go in, and you know not only spin up a new cluster, the hostage you know, control, plane, aspect, the node pool aspect, the nodes that are joining the configuration of those nodes, all from a single cluster and then, if you're, using ACM as well, because from again from what I recall, MCE multi-cluster engine doesn't require ACM, but ACM requires MCE.

C

I think so yeah, if you're tying in with ACM, then you can use ACM to further. You know configure and manage those clusters in all of the different ways. We start to see this holistic view of doing end-to-end cluster management. You know life cycle management through all of these different tools and again all programmatically, so you can adapt it and adopt it into whatever systems. Your you know, processes you're already using um I, did have a question for you, though, um what is like? How does node Auto scaling work with a hosted cluster.

B

Cool so I'm going to show that and I don't know, I guess four minutes we'll get there.

C

Okay, Alan Alan asked a question: uh can you deploy with bonded Nicks using mode four or lacp.

B

So I think you can so first the when I created the infrastructure environment, you can always add a static network configuration to that and it will configure the next. However, you like, uh however, than that, once it has that network configuration usually I'm, pretty sure of that the assistant installer will, when it installs the core OS. It also add the dash dash copy Network option so that should preserve the same networking configuration you added there throughout, like after the node is booted with the ignition endpoint. So.

A

It should stay the same.

B

uh I haven't tried, uh but exactly yes, but this is the.

D

Short answer is, if it doesn't work open, a bug.

B

C

It should work yeah and if you don't need, if you don't need that, for the first or the primary Network right, the one that has an IP address on the machine, Network dot, cider and uh install config, you can always do it day. Two as well. You know using uh NM state or something.

B

Yeah, so you see, we have one node, it's ready. It's happy now. One thing I'm gonna, I'm gonna do now is I'm gonna configure the load balancer for the increase traffic? Okay. So, sadly, as of now uh like currently, we do not have uh like the platform integrated load balancer. So we need to do it uh some other way. So what I'm doing is I'm going to deploy metal bit on this uh on the outside roster, okay and I'm gonna use a subscription and deploy it.

B

Then you want to create one I'll create an IP address, pool uh it will just have this one single IP and then I'm going to create an L2 advertisement that will tell everyone. Okay. This is the IP.

A

That we're using.

B

If you have hosts on different uh layer, 2 Network, then you should use bgp advertisement or something else, um but that's also supporting you. Don't have to have all worker nodes on the same subnet. You can bring them with different places and then at the end, I'm gonna expose the inverse via this load, balancer type service. So I'm just gonna start that.

C

And if you have any questions, go ahead: I I do have questions um and for.

B

C

In the audience, don't hesitate to ask questions too um so so that was one of the questions that had popped into my head before is hosted control planes.

C

You know we're deploying all the control plane Services into an existing cluster, which I can only assume means that we're using a route to expose access to the API endpoint, um which now that I think about it, has its own set of questions, because routes normally are only Port, 80 and 443, and the API is normally six four four three anyways I'll we'll we can get to that later. So for the cluster the hosted cluster, the manage cluster hosted cluster.

C

um Where does its? You know start out apps, and all of that, where does that get hosted at? Is that deployed in the normal way, using the Ingress controller and then I I I think? Maybe that's what you're doing now is deploying metal lb to then expose those Ingress controllers so that you can? You know Point start out apps at that exactly.

B

Okay, that's exactly what I'm doing um so. It took a bit longer than I thought uh so I'm just gonna rerun it. So the minute will be will get uh we'll get deployed, apply the crd and we can create one um and about what the other question. So you can see it here. So if um so you can see, this is how we get to the to the Jose cluster API right now, so um I use node Port. So that's why it's actually using the Note 4 for to get into that.

B

If we were to use, if we were having metal will.

A

Be on the management cluster or some other I.

B

Don't know load balancer running on advisement cluster, then we could use a hello bounce or.

C

Okay, I mean I would say you know that that might require some additional config but I assume, not because you know, if you're interacting with an OC, it's probably using the kubeconfig file to interact with it, and it has all of that Port information, everything in it so I, don't know for some reason: there's a something sticking out in my head of it was for a while or I. Don't know it didn't want to change the API from 6443 or we didn't support changing the API Port from 6443.

C

So that's really cool to see that being more uh more flexible.

B

Thanks so after we configure the balance there, you should be able to get to the hosted cluster console cool. Okay, um let's see, let's see that we can get there.

B

Let me just go: grab the credentials.

A

And the cool thing, too, is when you have the hosted cluster. If you were to do like an OC get nodes, you only see your workload right. You don't see the control, plane, nodes and stuff like that. So it's it's got a very similar feel to like eks and AKs right like where you just you get your cluster. You go into your thing and like, as far as you know, you're you're, just in your bubble, doing your work right.

B

Yeah so Alan. This is what we're showing this is on-prem ACN and it's Tech preview, but uh yeah so cool. Here we.

D

Can if you check the docs, there is a step to enable uh Jose control plans.

B

D

Yeah another step to enable the we call it central infrastructure management, that's kind of the code name for assisted installer, on-prem, so you'll.

A

D

Enable those, but once you do that then you'll be able to see everything just like this.

B

Okay, so, as you can see, this is I guess uh somewhat special, like uh it's an open shift. Cluster is also at the front leg, openshift cluster, but they just have one worker and nothing else. uh So pretty cool and I think now we can get to the question you asked earlier about. Okay, how does Auto scaling work so I want to say it works real nice. uh So let's create a deployment. Okay, just make sure this is the hosted, control, plane, yeah, cool, okay and just to make the auto scaler work.

B

I'm gonna be fine, so minimum requirements, okay,.

B

Console we should be able to see.

B

Okay, appointments.

B

A

B

It is so I asked for like five gigs for each pod in this deployment, so let's just scale it up to five and see what it's doing.

B

Okay cool, so it's it's running, workloads which is nice, but not all of them, because some parts can start because I do not have sufficient memory. Okay, so here's the cool part. uh We do not have a UI for this yet, but.

B

This is second figure if you want to do on the management cluster, so I get a notebook.

B

And I'm gonna patch it this okay I need to pass. This is the name and the patch goes like this. It's going to remove the replicas and instead because currently it has replica well uh I'm, gonna, replace it with auto scaling and I'm going to say. Okay, we want minimum of one host and, if you need more, then it can take up to four and the minute we do that. If we go here, we see okay, it's immediately the other scale indicated and machine the can be provider traded, the Aged machine.

B

It found uh an available agent, it's binding it to the cluster and soon enough it will start the installation all the part that we missed earlier, the validation that it's running and once it's happy with the host, then it's going to start the installation. So that's how Auto scaling works and once they're not join. We can see that okay, all parts are running. We can scale the application down again and a few minutes after the auto scaler will say: okay, I, don't really need to hosts here.

B

So it will scale it back down again and, and this host should go back and as we'll see it soon enough uh to go back or one of those will go back and become available here. So now you can see it's insufficient, that it's bound to this cluster, and if we look okay, so we say so. These are validations that are still in flight, and so it knows that it's bound to Cluster. It needs to pass all these validations.

B

They did not pass yet so we're getting this warning, but.

D

It says they're pending input, so it's waiting to get more information.

B

Yes, right now, the agent that's running on the host is running those validations, making sure that the what that, if we kick start the installation of this host, adding it to the hosted control plane that this should succeed.

B

uh This is how I guess this is the magic behind the system installer and how it gets such great installation success rate, uh because, before you start anything when you try to make sure that this should work.

C

Yeah I think we talked about that with ROM when he was on to talk about assisted installer and how you know if you choose the node to be a control plane node, it does things like run a disk latency test to check and see if the discs are adequate for SED.

C

um You know a whole bunch of other validations that you know we were just looking at I, always thought that was really cool and I really like that. It does that.

B

All right and now you can see it's starting the installation and straight enough we'll see. Okay, it's gonna write to this. It's gonna reboot and sometime later that.

A

Host uh should show.

B

Up here this is the establish control place, the waiting to get more computers.

B

B

Got anything else we want to show you guys.

B

A

Questions Maybe.

A

What's next I'm.

C

Sorry, no no I'm I'm, just watching I'm, it's funny, because I remember when we first started exploring this and looking at it, um you know way back uh gosh now uh it is completely escaping my mind.

C

So anyway, the ACM team also has a live stream and their very first stream, which was the better part of a year ago, was on hosted control plans and it's it's so much fun to me to see the maturation and everything that's happened since then, and just this overall experience you know again, I I think in perspective of, if I'm a super administrator right, I'm, the administrator of all of these open shifts.

C

This is hugely powerful and not just because it increases efficiency, but because of the manageability aspects, you know the the centralized control aspects, even more so than using something like ACM or get Ops or any of the other tools. I.

B

Want to say this is like we're showing it from daylight, because it's super cool showing it from the UI, but you can do everything with with pianos. You can do everything with CLI. uh This is really really something that is comfortable to work with.

D

And for a SRE that has to manage a bunch of clusters just having them all, you know as workloads on a single cluster not having to you know, find where each one is and and debug them separately, you're, basically debugging, openshift workloads, so it's much easier and for metal like we really created this experimental Cloud, you have a pool of hosts and you can just take them. Auto scale them as necessary when you're done with them. You just return them to the pool yeah.

B

I do want to say maybe because Ellen just asked what's happening at five out of nine So. Currently, it's waiting for ironic to put the note, but.

A

B

Say this is something like this some delay that was just uh he just was just added, because we, as I mentioned earlier, we had this convert flow happening in uh 4.13, and apparently it added like a minute here. Two minutes there uh to to do this, handoff that the assistant agent and the an ironic agent won't step on each other toes and maybe there's some optimization. We can make there to make this happen faster.

C

So I'll um so I guess a couple of things. So one I know we're coming up on an hour and 15 minutes and I know for you all. It's late in the day um I see the sun setting over there. um So we don't want to keep you on too much longer. So for the audience.

C

If you have any questions comments, thoughts, uh please don't hesitate to post those into chat or, as always, uh if you're not watching us live reach out to Johnny and I uh andrew.sullivan, redhat.com or Johnny redhat.com, um and then I want to get back to uh Johnny. You started asking a question there, which was you know what's what's next? What's the future of hosted control planes, look like yeah.

A

Yeah, exactly without without going into too much roadmap stuff right like because I don't want to I, don't want to put you in a predicament but like yeah, like I.

C

Mean clearly GA right, but.

A

Yeah yeah, but like what's what's some of the next, uh you know like hotness, with the host of control planes, and especially like the this is installer, because this is I mean. This is awesome right, like obviously I think you said it really well right. You have the single pane of glass for sres and that really Loops into like the governance aspect of it right. So it's like when you start talking about like application, governance and cluster governance and stuff like that and really understanding your environment. This is It's hard to beat this.

A

This is really awesome.

D

Yeah mix managing much easier. You could even we didn't mention this, but you can even upgrade the worker nodes separately from the control plane. So if you have like a maintenance window, but you need to patch a critical CV on the control plane, so you could do that in terms of what's next, um a lot of effort went in to Rosa in the lately, and the focus now, as you said, is GA um in terms of more roadmap around. Do you know, I, don't want to speak for product management.

B

I think mainly, the focus is going to yeah.

A

Nice well, this is awesome. I I, think that, like there's, I can see a lot of use. Cases for this I can see my I had. Somebody hit me up last night, literally about the same topic about bare metal and assistant installer So, like um you know, it's definitely it's kind of funny how we've gone from like this big cloud conversation now we're back to on-prem with bare metal again, so I think it's. It's really good.

B

Okay, so it took a.

A

B

Minutes and now we have the second worker here and hopefully all pods are running Let's see we got five out of five, so that's nice and if we scale it down and wait it long enough, it's I think it's about 10 minutes.

B

It will come back, but since I don't really want to wait that long uh we could just edit the notebook, and so we can send it back to use instead of motor scaling.

B

We wanna do the replicas again and.

D

I'm gonna say one another: scaler takes some time to kick in. It doesn't want to take away the nodes too quickly right.

B

Yep so, as you can see, you know, it's just went away uh if we go to compute and it's going away.

B

D

To Hope cluster.

B

D

Is doing the cleanup and properly removing that node.

B

I feel that Observatory, you should say that okay, it's removing from the cluster and soon enough, this should become, should reboot back to the python that we were booting it with and become available again to be used by other clusters.

C

I love it I, love that it's just this continuous pool of hosts that you can use and reuse over and over and over again and I. I've talked with several Partners lately, mostly storage Partners, who like they go in and they, when they're doing a CI run for their CSI plug-in.

C

You know they'll need to deploy a bunch of different clusters with a bunch of different versions and just constantly cycle through those I feel like this is going to be a huge benefit to doing something like that: hey I need to test with openshift, 412 and 413, and even different Z streams.

C

um I, don't know, that's a question kind of unrelated to or slightly related to this can I have or how do I create different versions of openshift to deploy into each one of those because it is set like I can only deploy a 4.13.1 or 4.12.17, or can I go in and create each one of those versions. Yeah.

A

That's what I was trying to ask earlier I just did a really terrible job.

B

So I think what you're referring is. You want to add more note pools and you want to have different workers on different versions.

C

No, like a whole separate cluster like if I wanted to stand up a whole new hosted cluster. That's using whatever version of of openshift I would assume that there's probably a limit, or only a certain window of versions of openshift that I can deploy as a hosted control, plane. Cluster.

B

So I think you can probably like with the self-manage you can probably use whatever assist it has been taught to install so I mean.

D

I think Coastal control planes is only supports clusters from a specific version through it, but okay but yeah you can use any Z stream or whatever you want.

A

Okay and Andrew like if you wanted to do like an arm with like a hyper or a hosted, control plane, you change the release image so like when you're doing your hypershift uh you're going through the hypershift CLI, you just say, use this release image. You know, for like say the the multi-arc and then it'll go out and it'll it'll deploy an arm cluster as long as it the cluster can or the the management cluster can support it. Okay,.

C

Yeah I think what you're, showing there is, what I was referring to, that openshift version. Is that something that you know me as an administrator, you know, can I control what versions show up there or is that something that comes from.

B

C

Mce, okay, yep.

B

So so this this comes from the agent cluster config, the agent service config. It's something I, think I can show it so yeah, okay, I'm not going to create a new one and I. Think if we go to here, I can show you exactly where this is coming from.

B

So when we configure the assisted.

A

B

Tell it okay: this is what you can install. These are the release images you are aware of um so I think they're here, okay, so for 411, we have this series image for 412. We have this one at 415. We have this one, and this is what you can see over there.

C

Okay, that's really cool I I didn't know that you could control that. So.

B

uh I think this I'm pretty sure this is what behind.

C

B

Maybe I'm Wrong.

C

uh Allen asks if this works with microshift.

B

So, what's what's this like.

D

As a management cluster, no.

C

Yeah I I think so I think the answer to that in all respects is going to be no um I, don't know, does ACM work with microshift I think it might I, don't think so not today, because microshift isn't really openshift. It's the kubernetes API or just enough kubernetes on the host to be able to deploy pods. None of the other I think things.

D

Are really for far roads use cases, not uh you know, hosting multiple other clusters.

C

B

Okay, I, don't know if you noticed, but the host is back in the pool and you can definitely reuse it for whatever other hosted cluster. You want to.

D

A

If you want to create.

B

A regular cluster works as well.

C

And also this is one way: there's no such thing as single node openshift with this right, because your single node openshift is a single node that is both control, plane and worker. Node, so you'll be deploying the control plane, which is a highly available control plane, and then you can have you know one or more worker nodes. So it would be a single node in that respect, but it's not a combined cluster.

B

I want to say maybe to make it more accurate, so you can definitely create a hosted control plane with one worker node. That's one thing you can do that. We.

A

B

A

B

Can also use this the same UI go to clock to the central infrastructure management use.

A

B

Hosts that we just use as worker- and you can create your SNL, your single node openshift from here and just get seeing a little bit shift. You can reuse those this the pool to do whatever you you want and you can install openshift in whatever way.

C

I also wanted to point out, because I've been having a number of conversations about this as we're getting ready to update the uh uh uh subscription guide if the cluster that's hosting the control plane workloads, if it's only hosting those workloads they're in for workloads. So you don't need to entitle that cluster. You only need to entitle the compute nodes that are being added in to host whatever applications are being deployed. I thought that was particularly interesting with openshift virtualization as well.

C

You know, I have a physical cluster that I'm, using with hosted control, planes and I'm, deploying all of those virtual machines that are the worker nodes, the compute nodes for the hosted clusters. I only need to entitle those virtual compute nodes for those those hosted clusters.

C

uh Can this be done with AWS govcloud.

C

A

C

Yeah I don't know the answer to that. I I would assume so but I don't know, I, don't know, that's a Johnny.

D

Our favorite host the control thing PM, is answering questions so yeah.

D

C

You added Adele and what his role is and why he's answering questions.

C

Thank you for joining Adele and helping to answer answer. Questions.

C

um So again, kind of going back to being sensitive of time, especially for you all over there in Israel. If anybody has any questions, if there's anything that comes to mind, please don't hesitate to post that now, while we're kind of closing out everything or uh yeah blew your cover uh or send it into email, all right, send it in email or slack or any of the other ways that you can reach out to us.

C

um So I will take a moment to thank everyone who has joined us today, really appreciate all the chat, all the questions that you all have had. um This has been really interesting to me. This is the again why we've been watching this evolve over time and seeing it all come together here and seeing what that final experience or close to the final experience we're not at GA yet, but the close to final experience looks like it's really really cool to me.

C

um So thank you both for joining us. We really appreciate it. We love having we love having you engineer, folks, like engineering folks, come and join us, because I always learn something, and especially with assistant installer in single node, openshift and again, the CV that you all have like it's. It's been I I've loved. Having you all correct me on things that I say.

D

Really fun yeah.

B

C

And then uh yeah, so thanks to our audience, thank you to YouTube for joining us, um really appreciate it um Johnny any anything from your side.

A

uh Just looking forward to the next couple weeks right, uh you know thanks thanks guys for coming on. It's really awesome. We appreciate your time. uh We know you're busy and uh it's um thank you for coming here and enlightening us with all this, because this is really it's really incredible. What you guys have done- and your story is awesome too- the fact that you guys have known each other for a long time and then you come over here together. I think that's awesome.

A

So uh thanks for coming on and uh yeah Andrew just looking forward to the next couple weeks, man, we gotta got a full schedule. It's gonna.

C

Be awesome, I I realized a moment ago that I might miss one or two of those due to traveling and stuff, like that I gotta go to Denver in early August and all kinds of other stuff, so we'll see and getting fancy I. Don't know about that. It is remember summit's in Denver next year, so oh yeah, but that's not why I'm going to meet with a customer. Oh okay, so uh Iran uh anything from you all before we uh close out.

D

Nope, it was a pleasure yeah, very.

C

Much so invite.

B

C

Yeah, you all are welcome anytime, anytime.

A

You know yeah Johnny's.

C

Johnny's kind of persistent when it comes to those things he he includes me on all the emails and all the messages that he sends out to guests all right, Joe thanks a bunch really appreciate everybody's time today. um Again, if you think of anything, if you're not watching us, live you're, watching the recording, don't hesitate to reach out andrew.sullivan redhat.com, johnnyjo, NY redhot.com or, of course the kubernetes slack workspace uh look for the openshift users channel in there, and with that we will see everyone next week.

C

Still struggling with this new UI Johnny.

C

All right, bye, y'all.