Red Hat OpenShift OpenShift Coffee Break | Red Hat Livestreaming, 3 Nov 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: OpenShift Coffee Break: Single Node OpenShift

Description

Get your espresso ready for the EMEA OpenShift Coffee Break as we will walk you through Single Node OpenShift (SNO) 4.9, a single node cluster optimized for Edge computing use cases with Kubernetes. Join the discussion & live demos on SNO with Jaafar, Natale and Tero and our special guest Alessandro Rossi!

Twitch: https://red.ht/twitch

A

Oh welcome welcome back to openshift coffee break and today we have also uh tarot last minute, so I see the screen. You know in the preview four of us and then there are other three that that are asked. I don't know why, but it looks cool. I don't I don't know if we can fix this. Let me check uh better now, uh not sure, but it it looks, uh still cool. um I hope you can all see us and welcome back to openshift coffee break.

A

um Today, we're gonna talk with uh terro with jafar and alessandra our special guest about single node openshift. You see, terror is already having is a coffee shot? I hope you had yours.

B

C

Jafar has yours, alexander.

A

Where is your coffee?

A

Okay, that's the way man right, um yeah welcome back today, very hot topic, because we're talking about single node openshift and you.

B

Know what that'll say.

A

Because of the coffee, they'll say: no, I can't do it and then yeah appear yeah there is match. This is a live demo effect uh fantastic.

A

So um I'm very happy to have you all today, because uh we we're gonna talk about single, not open shift. This is something we were waiting long time in openshift right about.

D

Five years, five.

C

D

And then, then we did a lot of hacks that having a single node running uh virtually face and running multi-node, so yeah.

D

It is kind of it's a good thing to finally have it, but I can understand that it is pretty hard to get properly working since the possible chicken and egg problem that there is that you you need something to install like in openshift four you, you use the kubernetes to install googlenet and when you have a single node, you you it's basically pretty hard, because you have to use the single node to install that environment. But but one question, it's it's movember good cause. I started first of november. I I have this raise your mustache.

D

Where is movember.

A

November yeah, oh yeah yeah, you.

B

Know what I I just shaved mine this morning- uh yeah- maybe I I I I missed. I missed that so I have to yeah.

A

So for the next show we make sure we will have some uh some.

B

Next show we have.

A

Hipster one yeah: it's not! If you have always this uh look and feel.

D

Yeah yeah yeah, that's that it gives the senior kind of look yeah, but about the single note it is. It is good to have it since there are a lot of use cases that it's just insane to run the minimum three node uh three node environment and still but yeah I'm. I hope we have a demo.

B

A

Well, I also hope so, and actually we have a demo, uh it's just a spoiler, but just let me remind the uh hello everyone also in the chat. If you have any question, please write down in the chat in youtube and twitch. We will bring your question during the show um before we start, I would really would like to uh introduce alessandro so alexander, who are you why you are here? What do you want.

E

Thank you thank you for making me be here. It's a real pleasure. We had a big chats with natalie in the past. I'm I I'm alexander.

E

I generate that in august, actually as a solution architect, and I also worked on many site projects in in the past, uh buffalo to give ernetis and openshift like an installer for and and a provision that we worked also with with my friend and colleague, valentino bertie, in in in the past, when I worked as an instructor also for that when I was a working for a partner and yeah, my I'm completely in love both with overshift and kubernetes. So I'm very happy to see this single note happening.

E

Finally, so as aster was saying, so I'm really happy to be here- and I will enjoy this one.

A

Thank you for your introduction, alessandra, and you know what we already have uh terror and jafar and alison. We already have some questions in the chat, so lo narrow content, which is a nice name in italian, is asking is possible to downgrade a running cluster to a single node or we. You need to set up a new one and migrate the workloads. So the question is: can we downgrade an existing cluster in edge availability to a single node.

B

uh I don't think so, because um the um the requirements for for the for the vips are a bit different in terms of uh the ip that you reserve.

B

So I don't think it's it's something that is possible, but when we will speak about the architecture and the installation process, I will try to make that a bit clearer on how how it works.

A

Okay, cool thanks, yeah that this is an interesting thanks for the question which you know what it means in italy. It means the happy bagpiper so and we are going into christmas. So there's this! uh You know there are people playing during the street. So it's it's cool, hello, adrian, hello, everyone! So um jafar alessandro, we're talking about single node openshift. Can we recap a little bit? What is single, node openshift yeah.

B

Sure so um traditionally I don't think we use slides in this uh openshift coffee break, but since we have a nice one, maybe I could uh I could use that. But for for for people who have been uh with openshift for some time, you probably have uh heard what we call the openshift all-in-one insta installation, which is basically uh it has. It was when we were in openshift3.x.

B

All the services running into one single uh machine uh into one linux, so it was very convenient for many things. For so for us as solution architects, it was convenient for our demo environments. It was very easy to set up and and very easy to run on, our local machines and uh at that time containers were just starting. So there was not really that need for for edge and and such things, but since the the the last years, um a lot of things happened in the kubernetes ecosystem.

B

We saw massive adoption uh across all industries and one of the industries that is now really uh into containerizing. The services is the telco industry, for instance, where we have the uh containerized networking functionalities, what we call nav and such things, and basically when you are trying to deploy those services on an openshift cluster, uh you had, I would say the the first pattern was you need a full flavored uh openshift cluster, with three control planes and at least two workers.

B

So that was like the the thing before I believe in four seven, we released a new uh pattern.

B

That was the compact cluster, which was the ability to have three nodes that play both the roles of masters and nodes, so it was already a way to, for example, address the regional data center, or maybe some offices where you want to have less infrastructure to to manage your openshift and and kubernetes workloads, and uh one of the the the big um asks we had from uh customers was to to be able to run all of that into a one uh note uh pattern so, for example, for edge where you want to have your workloads really close to your customers or to your local branches, etc.

B

So uh engineering has worked on that. It was not easy because, as tiro mentioned, we have automated a lot of things in openshift installation, and one of them is what we call the bootstrapping process. The boots trapping process is basically, you use an ephemeral external machine that starts all the uh containerized services that configure your openshift cluster. So basically you use containers to deploy containers and when you are doing all of that inside the same vm, there are some weird things that can happen. You have the load balancers.

B

You have some vips that are going to be hosted on the bootstrap when you are bootstrapping the cluster and then they will be held on the uh on the the masters uh etc or the workers. So so it's a complex process, and uh hopefully uh or luckily, engineering- have have worked very hard on it and they have been able to to make that work on one instance where everything happens inside the same node.

B

So you you do what we call in place bootstrapping and then the uh the work, the node becomes, a a master, slash uh worker, it's not really a worker, it's a master that is uh schedulable. Basically, so um that's, I would say the single node uh cluster. In a nutshell, the the so some people have already heard about code, ready containers so code ready containers is, I would say, the developer friendly flavor of openshift, but it's not a full fledged openshift instance, whereas the single node cluster is is really one.

B

You get all the features of a full open shift clusters uh with things like metering, etc uh that you don't get with a with an openshift code, ready containers. So it's upgradable. That's great news compared to to the other, I would say, developer.

B

Developer friendly, I would say uh platform uh or or pattern so um yeah. So that's the that's the the I would say one of the biggest differences.

B

So, of course, because everything is running on one node, you don't have the aha so, for example the hcd when you are running a traditional cluster, you have three control planes with three lcd instances to be able to have the column and all of those lcd things.

B

But now you have one single lcd running in the node, so you don't have a. But of course you can't have h a within one machine anyway. So so that's I would say one of the differences with a traditional openshift cluster.

A

B

A

For this explanation, I think you already answered one of the questions we had. um What that was. Are the single node cluster upgradable? I guess yes, you say the during the talk. There are other two questions. Maybe we can help with terra and alessandro. Also to answer so one is uh um one is: is there a list of supported host distribution for single node, open shift, uh and for this I guess it's uh only one. No, it's valid for the same concept for openshift there's only one os uh correct me.

A

If I wrong, which is rel core os and and probably the question were more around hey. Where can I start on on-prem or virtualization with other us? I think it goes in the case of uh open shift in general right.

B

So um so uh um I I believe openshift sno is uh supported on bare metal in installations.

B

uh That's my my understanding, uh so that's I would say the primary target is, is to have that on on on on bare metal in cells.

B

Of course it works also if you are deploying it on on virtualized environments. So in my case, that's what I'm running, but I I I think in terms of support, it's it's aimed toward a bare mineral deployment.

D

What about aram or power, so is it only x86 or any.

B

Other architectures, uh I think, for the moment it's x86, but I have to check actually for the uh for for the the supported platforms. Maybe we, if we look at the assisted installer we can, we can have a hint on on what's supported.

D

Yep, because the ocp you can run it on our arm also. So probably it's coming yeah.

B

Yeah yeah so uh again uh keep in mind that this is fairly new. uh The single node open shift has been available since openshift openshift49, so I I think it was released two weeks ago or something like that. I don't know exactly when was the ga date for 449, but it was somewhere around yeah mid uh mid october. I think, or something like that.

A

It was announced also at kubecon, so it's a pretty fresh kind of uh available availability is uh together with fort.

C

B

A

It's very fresh on october-ish, let's say.

B

Exactly yeah, so we will have more hands-on experience with that as we move forward, but yeah today, it's it's something a bit fresh, even for us, I would say so um anyway. I've been waiting for this for a long time, because, uh again, uh it's easier to have. You know your own uh open shift uh environment running in an all-in-one uh pattern, so uh this is aimed toward production.

B

So uh this is something that customers can use to have their. You know real workloads running in production on one single open shift, uh contrary to openshift containers, which is you know just for development purposes,.

C

B

Alright, so any more questions.

A

um Actually, yes, alessandro.

E

Can you yeah yeah? There are another interesting question that is regarding the subscription model for uh for single node openshift. uh There is no content. This is asking if whether the the subscription model will be the same as the classic or open shift uh subscription. So how does it work.

B

Yeah, it basically uses the same entitlement as a traditional openshift. It's just that it uses less subscriptions.

D

But it's free because masternodes are free, so it's free, well yeah.

C

B

But but only if you are hosting non-application workloads, okay,.

D

C

D

For the classification.

B

And since you are running application workloads, then it's not free.

B

D

um One comment about this slide: that when would you run a remote worker and when you have a single node, it's cluster, because when I remember when remote workers came it, those raised more questions than uh like answered questions because of the kubernetes network is.

B

D

Thing so is there really really use cases uh anymore for remote workers now that there is single node at servers.

B

So um so one big difference I would say, is that if you are in a disconnected environment, you can run a single node uh openshift, whereas the remote worker will not uh work, because the remote worker of course needs the the the control, the the communication back to the control plane and to to to the registry, uh etc. So so the remo remote worker can be useful in some situations where you want to. For example, you have already a central.

B

Cluster and you want to extend capabilities to remote locations, so you can pop up nodes and remove them, but you are still managing your cluster centrally. You have one cluster that you manage and you are adding and removing nodes. So in terms of management say you have a hundred facilities if you deploy remote workers and they are compliant with like networking requirements and stuff like that, maybe it's easier to manage that than having a hundred.

B

Single node clusters, so that's in theory, but single node clusters can also be managed centrally through acm, which is our advanced cluster management solution that runs on openshift. So you can also have like one single configuration for your snos that is replicated through your 100 sno clusters. So basically I would say that the biggest difference is this.

B

One is compatible with disconnected sites, whereas the other one is not, and also the network bandwidth.

B

The remote workers will need uh some proper networking uh to to be reliable uh and to have a you know, reliable connection with with the control planes. Even if, of course, uh we we have done a lot of improvement uh compared to a traditional um deployment, but yeah, that's, I would say the biggest differences. Maybe you guys see other ones, but that's the the main ones. I see.

D

Yeah, I don't think that that was quite explains that it explained the differences between the approaches.

B

Okay, okay, cool, so um so yeah. So, as I mentioned, um the the sno can be installed with acm. It can be installed with the openshift, install command line and there's a new feature that I like pretty much, which is the assisted installer, which is basically you go to console.redhat.com.

B

So it's a sas based, uh I would say, portal uh that red hat has created for the customers to be able to manage their subscriptions to manage their openshift clusters and they added support for single node cluster as a tech preview, and that's basically what I used actually to to deploy my my own instance.

B

So if you, if you want uh guys, I can if you, if there are no more questions, I can walk to walk you through the the process of doing that. I'm not gonna do a full installation, but I can show you some. You know some of the main steps and also explain uh some requirements in terms of prereqs like what ip and dns host names are required and such things.

A

Yeah, that's definitely definitely very interesting. Go go for it. We have a couple question in the chat we'll try to answer. Also uh during the during uh the chat, I think one interesting while you switch to the demo one.

C

A

uh That we we had a question like is this available on kd and yes, it's also available on ok. Did I link it in the chat?

A

Another cool, a good question is, uh is arm architecture support or for a single node openshift, because I think uh at the end the edge use case is gonna be multi-arc, uh but I I don't remember if arm is still, I think it's in depth preview somewhere, but I'm not sure it's uh supportable for a bare metal. I don't know if you have a more info about it.

B

um Yeah, no, not really. uh uh We will need to check back uh on on that topic, but yeah that that's you know um tarot already asked about it. uh um There's.

C

B

Support for uh you know, arm on openshift, uh like it's, it's a clear um target architecture uh so for sno we need to. We need to check where.

A

I I have seen their their support on aws for arm.

D

A

We need to check uh you know a single node of machine bare metal. uh I think there's no support yet, but uh techno technology, technologically speaking, all the container and all the binary are are available. uh I think it's not just it's just not supported yet. Yes,.

D

Another cool thing would be running a single open system, mainframe, ibm power.

B

Yeah but uh the the the mainframe can host, maybe thousands of snows, maybe.

B

So you will have a whole data center in there, okay. So uh so now we are looking at the console.red console.redhats.com.

B

Which is a as I mentioned, the portal that our customers can create can use to manage their subscriptions to to manage their openshift clusters. So uh what I did here is, I just went to console.com.

B

I selected like the clusters uh part and what I do now is I go to create cluster so from here you have different options and each one will walk you through the steps that you need to do to install openshift. So you can see there are managed services uh where you can run uh the um the installation.

B

uh If you want to install them on the cloud, you can run them on a data center and that's the option that we are going to use. So uh I'm going to go to create cluster okay and and then it asks you for like what are the uh the details that you want for for that installation.

B

So we will say coffee break just and I would say, coffee, break dot, open shift, dot affair, so I happen to to have to have that that domain. So I can fool around a little bit with it, and the interesting thing here is that you can choose to have a single node installation and what this installer does is.

B

It will basically generate an iso image that you can then load on your uh you can bootstrap on your machine or in your virtual light environment, and it will uh talk back to openshift to the management service and it will uh automatically configure uh the the the instance for you. So it's very, uh very nice uh too, to have that so, okay, we will say: okay, we see that we have openshift for nine. I go next and, as you can see here, you can generate a discovery uh iso.

B

So basically what it does it's. It generates all the bootstrapping information and, as you load this iso image, it talks back and and gets configured.

B

So there are three um important requirements which are: if you are familiar with openshift4, there are um two endpoints that need to be configured in terms of uh dns, which are the first one is the api. So it's basically api dot, something so it's the api dot, cluster name, dot, uh domain etc, and the second one is uh so.

B

The api is, is the end point that you're going to use to talk to uh to the kubernetes and openshift api, and the second endpoint is the um white card extension that is added to all the routes that we create. So it's a star that, for example, apps.openshift dot cluster name, so those two um dns endpoints needs to have fixed ips that you reserve that will be managed by the um the the ip failover services on on the installation itself.

B

So you will have those floating ips that are managed by the uh the openshift single node, but you also need to redirect the the traffic to um to you. You need to reserve a fixed ip address for the host, uh so this is the part that is a little bit different from the from the the traditional installation of openshift.

B

So here you have the api and the the the white card that are two floating ips that you will get and then you will have a fixed ip address that you reserve for for the snow. Basically, so once you generate the discovery, iso, it tells you if you want to have a full image, meaning that it's going to be about one gig or if you want a minimal image that will have like.

B

I think it's 100 megs or something like that, and once you start it it's going to pull uh the the information that it needs. So if you want to be able to troubleshoot and connect to your snow so same way as we do with openshift, you can upload your ssh key in here and then it's going to be added to the ssh key store uh in your uh open shift. Sno.

B

So that's it! You generate the iso you boot it you wait for for me took about. I think uh 45 minutes.

C

B

No a bit more, maybe one hour to have uh the the the process completed, and the good thing is that you have a nice visual ui on the progress. You see all the steps you can access the logs. uh So it's it's very nice experience to do that this way. So basically, what happens afterwards? Is you see that the the node registers itself here and then you see the progress uh in this ui so fairly simple?

B

uh I really liked the experience and I have one here that was installed, so you can see the link in here. You can download your cube config, you can check the basically the the settings etc. So uh very nice experience.

B

uh I really liked it and you can try also to see if it works or not.

B

So any questions about this.

A

Well, this is very cool, I don't know uh terror. Also, if you have a question about the this assisted installer procedure from console red.com, I don't know if you already have, if you, if you ever use that or if you plan to use that with single node openshift for your workloads.

D

Nah, no, no actually don't know yet, but I have a few questions. So basically, the assistant installer cannot be installed in an aircraft environment because it needs the connection to the console.

B

D

But then can you, uh but if you do the full image the one gig then, is that possible to run in a aircraft or do you always need the uh connectivity to upstream console.

B

So uh if you are using uh the uh the service, like the assisted installer, uh I think even when you are running the full image, because that's the installation that I used, uh you still need to have um the network connectivity to the to to this endpoint.

B

uh But if you are using the uh openshift command line, uh then you can do a disconnected install.

D

Yeah, okay, but there was a support for proxy. So you can. You can set up the connectivity with the proxy yeah.

B

D

E

All right, another good thing that is coming from the installer is that sometimes one of the obstacles that were found by the customers and also people that just wanted to try it. It was that it was a bit complex. Let me say to just have everything up and running to just try it and try it and then destroy the cluster and then recreate the cluster so having something that is you pick it and you start moving on from scratch.

E

It's a very great thing, so yeah. This was one of the things that I think that are more appealing about the the this kind of installer and also download the config and have everything in place. Starting from the beginning, it's very cool yeah.

B

So honestly, it was a a great experience, so I didn't plan to walk you through the installation, but maybe we can do that in another session or something, but because you know it takes some time, but basically it was never that easy, at least for me to install an openshift cluster. uh Of course we have a lot of ways to do that. You have the yaml files, you have the openshift install you have the acm, so all of that works perfectly fine and we've done all of that before.

B

But now all I did really was go to the assisted installer. So again, this is in tech preview, but it works for uh it's going to be supported soon. uh So you generate the iso, the the the image the the the os boots up talks back to the cloud service, and then it gives you a pop-up asking for your for information like what's the api uh vp. What's the uh wildcard vip, what's the fixed ip etc, and once you enter that information like it asks you do, you want to set up additional disks, etc.

B

So once you you enter that information in the portal, it sends back the configuration bootstrapping to the sno and the sno reboots with that correct information and and starts the install process, so very smooth, very nice and easy to to do all of that.

B

Okay, so uh enough talking about the installer, so here we are looking at the sno. How can we make sure? So? Basically you go here. You see that it's one single node that has both roles, it's master and worker.

B

But if you look at the cluster settings you see that there are no dedicated workers, so it's one master that is schedulable, so it plays the role of a worker and the nice thing here is: you can see that it can be upgraded. Actually, so there's an upgrade that is available. I'm not gonna. Do it now, because we don't know what can happen.

B

A

Live demo effects, yeah yeah.

B

We want to to show you the thing uh working and not wait for the installation to happen in the background. So so, basically you have all the features that you have with the traditional uh openshift cluster.

B

You can see uh here in the overview that you have the metering that is working. You can use uh storage, classes, etcetera, etcetera. Of course. This there's an indication here that there's no high availability, because we are in an sno pattern, but other than that everything looks exactly the same as openshift, and since we are in openshift 4, you can see that some changes happened in the ui.

B

So, let's go to you know a fairly simple project from the developer perspective and see if everything is working, so this is not the one I had prepared. I think that was this one.

B

So here I see I have two workloads that are deployed, one node.js app and one quarkx app and basically all the services are working in there. Okay, so the application is started.

B

The other service, the quarkx 1, is started also, so let's try and create maybe a new project and see if everything works, as we do with a traditional openshift project. So let's call it coffee break.

D

One one question: meanwhile: what are the storage options that you can use? Is it then local, storage or nfs, or what are the choices.

B

So uh very good questions, so you can so what I did in this uh is just you know something for my home lab uh and it's not something uh you know to use uh in your production environment. So I I did set up an nfs storage class using the nfs provisioner, and you can see that I have a storage class in here, but you can create other storage classes and use.

B

So they are available. I don't know to what extent they are going to be adapted for an sno, especially if you're running ombre metal, but probably the most suitable thing is. If you go to the operator operator hub, there's the openshift.

B

Local storage operator, okay- and this is probably the one that you could use if you wanted to have everything running in one single node. So basically you need to have a free disk that is unpartitioned on your machine and once you configure the local storage operator, you can use that unmounted disk as the local volume provisioner for your openshift cluster.

A

That's interesting in general for openshift right uh when you.

B

Have some disk you.

A

Want to touch and it takes care. Moreover, this is not the case, because it's single node, but if you have multiple nodes, these take care of the complexity of.

C

A

The intelligence no to attach a volume only on certain node if there is red right only so it's a cool feature.

B

Yeah, so um yes, uh of course, as we said, it works for a traditional open shift cluster and in case of an s o, so basically the question of the storage.

B

It's always you know, there's no one single answer for that. So, basically you are already in a non-ha environment right, so.

B

If you think that your data can be safely hosted on the sno, for some reason, you can do that if you believe the data should be better stored elsewhere and have your sno set up to access the data, there then go for that.

D

B

D

Of course, you have to consider the type of workload that you you.

B

D

B

And the criticality of your data, so if it, if it's data that you can afford to lose and re-load, then okay, if it's data like that, is mission critical, then you don't fool around with that and- and you know, risk losing your your data if, if something goes wrong with your uh snow uh for, for instance,.

D

Yeah, but if we take a use case, remote side, maybe using low car storage for message buffering or.

B

Yeah logging, yeah yeah exactly okay, so basically uh that's that's it. So you can see you have a full uh openshift uh experience. You have uh the dashboards. uh Everything works exactly as with the.

B

The the traditional openshift cluster, uh let's switch back to the developer perspective and maybe try to add quickly uh a sample, so the the four line perspective has changed a little bit, so I needed to get used to it.

B

But now I can find my way around it, but yeah it's a nice improvement to the to the ui. So I'm just deploying a simple node.js app. uh You can see that the build started. It's cloning, uh the code from github and once it's done, it's gonna be be running in here. So very nice very easy to set up very happy with it.

A

Pretty cool and jafar I wanted also to share my single node openshift, which is in this intel nook, very cool. You know what, because I I'm doing the I read that act, fest yeah. Let me share in the chat the link there is in this week that read that hackfest. If you are interested to join it's happening now so, but if you are interested to join the community about the the red attack faces, uh is um it's using the single node openshift and then uh quercus aqua software stack.

A

So a java software stack to implement server-side and client-side pollution data to be collected and analyze it. So I think this is a cool use case, so you can put as jafar that did your single node openshift in your uh bare metal art, where you can put on an intel nook uh to come back to your question about the architecture at the moment it's intel based, but the support of other architecture is also coming.

A

um I really recommend you to go into that links because there's the procedure also on how to install now to install additional software on single node open shift, but I really enjoyed the this demo jfr because it was visible.

A

How was easy to install it and at the end, that is an openshift cluster, which is a master worker together, I don't know if we, if we can go into the compute zone and we can see that that is a master worker uh and and so um it's a single node, where all your app will will run. um This is the um I think this is the cool thing. I have one question to be honest: what are what are the hardware requirements? What is the minimal minimum ram that you have to have.

B

So so the the documentation says uh eight cores to 32 gigs of ram. So that's the the minimum because of course there are services that will be running. You know all together on the instance, the lcd, the api servers, the monitoring. So you need to make sure that you have enough room for those workloads and your applications.

B

So, depending again on on the the type of workloads that you are going to run, you will probably need to beef up that platform to have enough room if you are going to run.

B

Php micro services that take 50 megs of memory- you are probably fine with 32 gigs of ram. If you are going to use more specialized workloads that need gigs of ram for each pod, you probably need a little bit more than that. So for my instance, I had uh provisioned, I think around uh 90 gigs of ram for the incense, so that leaves uh enough.

D

How much the how much the cluster is currently consuming memory and cpu. Let me check.

B

uh Let me check because.

D

That's pretty vanilla, it doesn't have everything anything.

B

Yes, yeah exactly so um so the cluster, but again keep in mind that some services will stretch depending on how much memory you have available.

B

So the more you have, the more they can take right, based on the the the limits and the uh like the default uh requests and limits that you that are set up so currently my cluster takes, I would say, about 10 gigs of ram and about three cores about out of the. I have 14 cores available.

B

But if you deploy it on a lesser configuration, it will probably take less memory, because the monitoring and the logging will take less than that.

D

And one other thing would be nice to learn that: how long is the service break when you do an update on a single note,.

B

Oh so yeah I can. I can run the update and let you know how much it takes. uh I didn't want to run the update before uh the uh or I mean on the session, uh but yeah. Now I can. I can hit the update button and and see how long it's gonna take.

B

So um it's it's. It's totally different from. I think I have one cluster that I upgraded since openshift41 that I still have so when you, when you look at the upgrade history, it's like pages of updates.

B

This one is gonna, be fresh, so uh the other one. I know how long it takes this one. It's gonna be a new uh thing for me. I guess somewhere around maybe 30 minutes. I don't know.

D

Yeah because it has to update the api, lcd and the dusty and and then it has to download the new. The.

C

Metadata messages, yeah.

D

Yeah and coreos and boot new, so it has to reboot. Yes,.

E

Correct, I I actually tried it uh in the last days and it takes yeah about 45 one hour. It depends basically on mostly on the networking so for downloading. All the new images for the core os and also the cpu requirements are.

B

Yeah, quite a few.

E

Yeah the the cpu requirements are quite, let me say, strict mostly for the bootstrapping phase, because one of the issues that you can run also when installing production- and you are just too restricted on the cpus- is that the booster process is not going to do so. Well,.

B

So actually, I ran into that issue to be fair. On my first install my first install, I did not provision enough cpu.

B

I really like tried, even maybe below that minimum requirement to see what would happen and the bootstrap did not complete, because you know there was not enough power horse horsepower to to you know, download the images bootstrap, the services and everything that needs to to happen in the background, because uh if you look at the container image, you have cans of container services that run in there and yeah just just to to show you that it's basically a full-fledged openshift cluster. Let me try to connect to my openshift sno.

B

So do you see my screen.

A

Yep, can you can you increase a little bit the font yeah.

B

Yeah thanks better so.

D

I have glasses, I don't need, I don't need it to be increased.

B

So so you see the the the tens of uh processes that are, you know, containers that are uh already running in the cluster, so.

D

But now you have a situation that you need a second snow cluster uh with the load balancer that you, during the updates. You have something providing platform for the workload so.

B

D

You have three of those and then you can actually.

B

Yeah, so so yeah, so of course that's why they say: there's no h a of course. You need to bear in mind that you don't have h a if h a is uh important and you need continuity. Then probably you need to go with the three node compact cluster pattern. If you still want to know to minimize the infrastructure, that's probably the safer choice for that type of use. Cases.

A

uh This answer the question we had in the chat jaffar because uh los angeles uh on edge scenario. What is the h a for a single node operation, because you say I prefer to have two nodes rather than five, but if, if somebody can use the compact cluster, so three, I think it's. uh I think it's the right choice right instead of having two single node openshift, one compact cluster of three uh master worker.

C

D

Yeah yeah, but there is also also this that in some its locations you might have two sections in your data center, so you might actually have better to choose two single node clusters. One in each room then have uh all in one three node setup where you have two nodes in one room, which is then a single point of failure, so you don't basically get all.

D

Of course you will have two clusters to maintain, but you will have the same fail tolerance than you would have if you could do two uh data centers uh with three node clusters. Yes,.

A

D

Constant constant debate between you need a cool room. What, if you have two data? Centers.

B

Yeah exactly and if you can manage externally the load balancing stuff uh I mean if the load balancer is not installed on one of the machines like if you don't require any additional machine, and if you have hardware that can dynamically switch the routing and stuff like that, then yeah. You can probably go with that type of scenario where, during the upgrade you switch you you, you move the containers to to the other cluster or you have them as an active, active or active passive.

B

Depending on the architecture of the of the workloads you redirect all traffic to the one. That is not upgrading. You wait till it's finished and then you you, you know the the other way around.

A

I thought of to your point: probably you have two different clusters. Then you need some kind of a lot but load balancing operator or when you you can route the traffic transparently or submariner. I don't know this is a multi-cloud multi, let's say multi-cluster use case.

D

Yeah, it depends. I would like java said that it depends so much on workload. How is it just that you need to have it like? Sla is 20 uh three hours per day. You just need it for to be there on office hours.

D

It doesn't matter, then what you run, but if you need to have it's in it, it's taking a lot of traffic, doing something and then sending it to forward to my data center it it depends on how critical can it go down for one hour or should it be hca, and how often you need to update it, because the the kind of mentality which opens it for is you always upgrade?

D

But if you take lts versions, do you actually have to upgrade that often, can you run like years without upgrading? So it also it's a it's compared to the how critical the workload? What is the security requirements slo sla? So there is no.

D

uh Let's say there is no silver bullet, but I would say that this kit gives a good uh option to have an multi-cluster environment that you can have and with two data centers, you can have an hsa application workload so that you can meet external, it's load balancer.

D

You can actually use two clusters that are and the environment is supported, because that that's important thing also, but you have to think about. You- have two clusters and you will still have that single cluster. It's not a fail to learn, so it compared what data you store and the clusters don't have like. Maybe share storage, so they are isolated.

D

So, if request goes to, as you know, cluster one store something. The data is not the messenger as an old cluster two. So these are the considerations, but it's a it's. A good extra option. uh How you set up your environment on the edge.

B

Yeah and so so just to keep in mind something that um the goal of all of this is, I would say, to ease not only the proximity of your workloads to your other business, related applications or data, but also to simplify multi-cluster management.

B

So the the fact that our uh openshift sno is like a plain, full-fledged, openshift cluster. uh It means that it works seamlessly with the acm and you can use acm to do this kind of things like if you want to move workloads from one sno to another one.

B

You can use the same things as you do with acm, like by tagging, the target clusters and removing flags- and you know, labels and stuff, like that. You can centrally manage the workloads from your acm to your hundreds of sno clusters.

B

For instance, uh say you have like a 100 branches where you have an snl running everywhere and you have connectivity to a central acm or you have two snows per per branch or whatever you can manage all of that uh centrally from acm, as you would do with a multi uh worker uh open shift cluster.

D

This is kind of the the use case that, because director didn't have the proper edge use case earlier, the one solution would be just run rail and containers, but this now the rail is now open. It opens if this is the new rail.

D

I said way back so this is that you don't actually install really you install opencv, but you run uh you get all the good things from kubernetes and you run containers and it's still single node.

B

And- and actually you remind me of uh of a conversation that we had with one of uh my previous customers, but it was years ago, I think it was almost three years ago or something like that where they wanted to have that use case. So they were in a manufacturing.

B

uh You know plants and they wanted to have that minimal footprint where their services are uh containerized.

B

So basically, that's the the choice that they went with. So uh uh they had a rail with. uh You know, containers running on it and that's it. But of course you didn't have the the management capabilities of openshift, the the ui, the the metrics, and all of that you know the developer.

B

Experience that you can also because remember that also, if you are using an sno one of their use, cases was that they wanted to from their local plant if they had an update that they needed to push to their applications that they could have a local development environment where they can make changes on the fly, run their ci cd on the plant and update the application without being connected to the central facility.

B

So openshift was a great use case for that, of course, but it was not a pattern that was uh possible with openshift4, so now they can do that because they can have a fully fledged open shift with maybe call ready workspaces with the pipelines with whatever operators they want to run on that single node.

B

But it's going to be one machine that hosts all those services.

A

Yeah, oh and you know what we we are.

B

A

Have uh I think, alessandro's fan, there's lots of people say grand alessandro, hello,.

C

A

You know I have to explain that alexander is very active in our openshift community in telegram, is uh we invited him because he did he do he does lots of things with uh on github we've open lots of experiments and he he did also on uh single node openshift. Then alexander, any final thoughts about it.

E

Yeah and yeah, I also have to uh to thank rodrigo for the precious gifts that he produced for his following, obviously from the bash, and I just adapted this one to to an ansible playbook that is public, available and yeah. My final talks are that all the single lot of issues is something that it will probably change, how we're going to approach edge customers and the different scenarios that can require disconnected workloads or something that should be also local. So uh I'm really happy to see how this will evolve and yeah. It's very. It's been.

C

E

Great to discuss this to this together today, so thank you again for this opportunity.

A

We we, we are happy to have you today and thanks everyone for joining. Today we have. uh We are at the end of this session, this coffee break, unfortunately, because it was super interesting with demos and talk, but we keep talking on openshift tv uh jafar. That is true, because today we have additional uh show- and I put the link in the chat about our streaming calendar uh java. Are you this afternoon again on on the stream.

B

uh I don't, I don't know.

A

B

A

Be if we look at the at the calendar, that should be the level up and then, if you go to the calendar, there are other sessions this afternoon. If you like to stay connected with us on openshifttv and uh for our next appointment uh is actually not on the november 17, because uh there's an event in between. uh Let me share also the chat there. We we will be in kdc italy, which is a kubernetes event, and um I will share in the chat we will come back uh on of the openshift.

A

Coffee break will come back on december, the uh with another show on our series um and uh in the while um um I put in the chat uh uh the next session on openshift tv. uh Please keep in touch uh with us and uh subscribe to our openshift channel. uh Let me share also in the in the chat the link to uh to to be connected with us to like the show to stay in touch with us and and yes, I will uh really thank you, uh jafar tero alessandra, for joining us.

A

It was a pleasure and um see you see. You soon see you soon on openshift tv, coffee, break.

B

Thank you very much. Ciao bye.

C