쉐도잉 연습: Kubernetes 101 - Episode 3 - Deploying Apps - 영상으로 영어 말하기 배우기

Thank you.

Hello, everybody.

Today we're going to focus on deploying, exposing, scaling, and updating an app in Kubernetes, along with a couple other things along the way.

So as we start all these live streams, I'd like to say welcome.

I'm here in St. Louis, Missouri.

It's getting pretty cold here, but please say hello on the live chat and where you're from.

And if you're watching this after the live stream is over, unfortunately you won't be able to participate in the live chat, but I still love hearing from you.

So feel free to leave a comment below.

So I'm going to switch over to my screen share here.

And one thing that I noticed was some people, especially in episode one, when we were starting to use Minikube, they were having some trouble installing it,

or some people had corporate policies that prevented them from installing it.

Or a couple people actually now have M1 Macs that you can't currently run Minikube on.

There are ways to work around those kind of things.

But in general some people have had trouble getting minikube working on their computers

So I actually recorded a separate little video clip for you to show you a way

that you can have a free cluster for testing in the cloud So

that you don't have to worry about anything running on your

local computer as long as you can get Qubicuddle running on your local computer You can work with

that cloud and all the examples today

And I'll try to make this the case for the rest

of this course too all the examples today will work just

as well in in this cloud cluster as they would in a local Minikube instance.

So here is that video.

Back in October, I had a crazy idea.

How fast can I get 50,000 Kubernetes jobs deployed in a cluster?

After this live stream is over, I'll link to the video I made exploring Kubernetes jobs in a card above me.

But if you're watching live, check the description for a link.

Linode generously sponsored that milestone video, and they also gave me a link that I could share with anyone to get a free $100 credit on Linode

to develop and test in one of their Kubernetes clusters.

And I'm happy to report, even though they aren't sponsoring this video, that's amazie.io, that link that they gave me still works,

and so if you're having trouble getting Minikube running on your computer, or if you just want a cluster with no fuss to test things out, please visit the link and sign up for a Linode account.

The link is in the description below, too.

Anyways, assuming you've created a Linode account, creating a cluster is really easy.

After you log in, go to the Kubernetes section of the Linode control panel.

Label your cluster.

In my case, I called it cube 101 after this series.

Choose a region, ideally one that's close to where you are so that things don't get too slow.

Then choose a Kubernetes version.

I chose the latest version of Kubernetes that Linode had available in December 2020.

That's today.

But if you're watching this a few years from now, it'll probably be really outdated, so just choose the latest version.

Then add some nodes to a node pool.

You could just add one server, but I like to use at least two servers in any cluster that I build, just to make sure I don't get lazy and build things that work great on one server clusters,

but break when you have two, three, or more.

Once you click create, it should only take about two minutes for the entire cluster to be up and running.

To be able to connect to the cluster, you have to download the kube.config file, and Linode gives you a handy download link.

So once all the nodes are running and the kubeconfig link is available, download it, and then switch over to the command line.

Minikube automatically configures our kubeconfig file for us, and we haven't really talked about it before,

but a kubeconfig file describes exactly how to connect to a Kubernetes cluster's control plane.

So we can use tools like kubicuddle to work with the cluster.

What I like to do is copy cluster kubeconfig files, try saying that five times fast, into my .kube folder in my home directory.

So I'm moving the kube101-kubeconfig.yaml file I downloaded earlier into the path home slash dot cube slash config dash kube101.

Now I need to tell kubicuddle what cluster kubeconfig to work with, and I'll do that by setting the kubeconfig environment variable.

And now I can start interacting with the Linode Kubernetes cluster

just as easily as any other cluster like a Minikube cluster running locally.

And with that, I'm going to hand it back to, I guess, myself.

Ah, that's a little awkward talking to myself, but that's kind of how you have to be when you're working remote alone all this time.

But anyways, so now we have this managed Kubernetes cluster, but I want to go one step further.

I'm going to talk about a lot of things in this series, in the Kubernetes 101 series, relates to setting up a cluster, getting the tooling around your cluster, all those kind of things.

And a lot of that is stuff that you might not even want to have to do.

You might have to spend a lot of time dealing with all the surrounding parts of your cluster.

And I mentioned in that video, Linode was not the sponsor, but I'm just extending that generous offer that they gave

because a lot of you might want to have a cluster to run.

But Amazee.io is actually sponsoring the series, and I thank them so much because they have made this possible.

And a company like Amazee.io basically takes all those parts

that require a lot of maintenance and building your Kubernetes cluster

and managing it and making sure all the right parts are put together inside of it

so that you can manage your applications on it.

They do all that stuff for you and they make sure it's monitored well, they make sure it's running well, and they'll do that in any cloud environment,

whether that's Amazon or Google or Azure.

And they They can work in all those different places

and help you to make it so that all you're managing is your applications.

So today we're going to focus on application deployment in Kubernetes.

And the end goal for any of us is to make it so that we're not managing Kubernetes, we're managing our applications.

And a company like amazie.io does a great job of helping you work on your applications

and they'll take care of all the complicated Kubernetes parts.

So anyway, we have this cluster running, and whether you run it in Linode or in Minikube or in some other environment,

these directions, one of the great things about Kubernetes in general is

that your definition of how to run an application will work anywhere, whether it's local or in the cloud or in some other environment.

As we go along further in the course, there are some things that are more specific to certain cloud providers and things like that, but I'll warn you about that when we get to them.

But we want to run that Hello Go app.

And let me bring up a browser window here.

It's on my other computer, so let me bring it up here.

Over on hub.docker.com, in the last episode, I pushed up this cube101-go app.

And it's a private repository on Docker Hub

that only I can pull from and only I can push to because it's in my private account.

And so we want to run that in Kubernetes.

So, and thank you to Andy.

And also, I can't see who else it was because it scrolled past already.

But thank you so much for those Super Chat donations helping make this series continue and be sustainable.

Although my node, I should say generously, is donating the cluster time for me that I'm still using for the series.

So thank you to them for that as well.

So I want to run this in Kubernetes.

and I have a Minikube cluster running locally here and it's running Kubernetes 1.19.

It's funny, Kubernetes 1.20 was just released in the past week and so now already the video that I just made,

it is the latest version you can get on Linode's version of Kubernetes

but I'm sure that they'll be coming out with 1.19 soon and probably 1.20.

So in the world of Kubernetes, anytime that you commit something to a video,

it's out of date within probably a few hours if not a day or week.

100

And thank you also to Roger.

101

Very generous.

102

Awesome.

103

So, I want to run this app inside the cluster, so I'm going to say kubectl create deployment,

104

and this is the same thing that we did in episode one.

105

Just creating a deployment.

106

I'll call it Hello Go, because that's our little app, the kub101 go app.

107

And then I'm going to give it an image, equals gearling guy slash, and this is just coming from here, gearling guy slash cube 101 go,

108

I'll even copy it out of there, save a little time.

109

And then I'm going to deploy version 1.0.0.

110

That was the version that we created last week, 1.0.0.

111

And I'm going to go ahead and deploy that.

112

And I'll watch this deployment.

113

watch kubectl get deployment and again i might i mentioned this before

114

but i'm saying kubectl just because that is an official way it's kind of like

115

if you're saying the word gif under protest i'm saying kubectl under protest

116

but when i say

117

that word you can substitute it for whatever you want to say a lot of people say kubectl

118

because it's like system control kubectl anyway off on a tangent there.

119

So I'm going to watch this deployment, get deployment, hello, go, and see what happens here.

120

And I notice that there are no pods available in the deployment, and that's not a great thing.

121

It says ready, zero out of one.

122

And this is a tiny container image, so it should download pretty quickly.

123

So I can say kubectl get pod, and I'm going to use a label of app equals hello

124

go now this is a special label kubernetes allows you to

125

label things to be able to identify them to be able to use them in other places

126

and by default

127

when you create a deployment like this it will create a

128

label the app is equal to whatever the deployment name is

129

so that's a handy way to get any pods

130

that are part of this deployment or get any other resources part of this deployment

131

so i'm just going to say get pod with the label app equals hello go

132

and it's saying that that there's an image pull back off status.

133

Typically if a pod is working, it should just say status running, or if it's a job pod, it should say status completed or finished or something.

134

So this is something bad.

135

And I want to dig into what is causing this.

136

So I'm going to actually use instead of kubectl get, I'm going to use kubectl describe.

137

Describe gives you a little bit more detail on the command line about what's going on.

138

So kubicuddle describe pod dash l app equals hello go, not ag.

139

And I can see here that it's saying fail to pull the image.

140

For most things in Kubernetes, the list of events will give you some good debugging information to get you started.

141

And here it's saying that it failed to pull, and the daemon that's running that is trying to pull the image says that pull access was denied.

142

It may require Docker login.

143

So it looks like we might need to actually authenticate to the cluster, or authenticate to Docker Hub, which makes sense.

144

Minikube on my computer right now doesn't know how to connect to Docker Hub as me, so I'm going to need to tell it about that.

145

And I made another little video, so I will hand it off again to myself

146

and let you guys see how I'm going to connect this Minikube cluster to Docker Hub.

147

The reason my Hello Go deployment is failing is

148

because Kubernetes is trying to pull a container image from a private image repository on Docker Hub,

149

but the Kubernetes cluster itself doesn't have any authentication credentials it can use to do that.

150

So we can create a special kind of Kubernetes secret to store our Docker Hub private registry credentials

151

and then tell Kubernetes to use that secret when it tries running the Hello Go application.

152

What's a Kubernetes secret, you ask?

153

Well, there's special objects you can store inside Kubernetes containing sensitive information like passwords, tokens, or keys.

154

There are some special types of secrets, like a Docker registry secret, which require a certain format when you're creating them.

155

And so, to connect our little cluster to the Docker Hub registry, I'll create a Docker registry secret and name it regcred.

156

The syntax for that is on the screen right now, and you can pass in a username, password, and email.

157

And if you're connecting to a registry besides Docker Hub, you would also need to pass in a server with the option docker-server.

158

And here's one important thing to remember.

159

The password is not the password that you use when you log into Docker Hub in your web browser,

160

but instead it's the value of an access token you configure in your Docker Hub account settings in the security section.

161

And oh my gosh, I hear you saying right now, did Jeff just expose his Docker Hub access token to the whole world?

162

Is Redshirt Jeff trying to create a little mayhem?

163

Relax, by the time that you're watching this pre-recorded segment, I already deleted that access token.

164

Or at least, I hope I remembered.

165

But back to the task at hand.

166

The regcred secret can be used wherever you need pods to pull container images from the private registry.

167

To use it, you can edit a deployment template, like the hello go one we just created.

168

For example, run kubicuddle edit deployment hello-go, then go to the spec section under template and then spec,

169

and add image pull secrets, which is a list, and add one list item with the name regcred.

170

At this point, if you describe the pod that's part of the Hello Go deployment with the same kubectl-describe pod command from earlier, you should see it's pulling the image.

171

There are a few things you need to keep in mind with pull secrets.

172

Every container registry provider uses slightly different terminology and may use different types of access control,

173

so refer to the official documentation for how to connect to your registry if you're not using Docker Hub.

174

And some container platforms even do all the authentication for you, assuming you use a container registry that's connected to the Kubernetes platform like with Amazon EKS.

175

And one last note, we aren't going to dive into the concept of namespaces yet

176

because everything we've done so far is in the default namespace, but pull secrets are specific to a Kubernetes namespace.

177

So when I start talking about namespaces in the next episode, you'll need to remember that any namespace

178

that pulls an image from a private registry will need a pull secret defined in that same namespace.

179

Now, for the second time today, I'll hand things back over to me.

180

I don't think that'll ever be not awkward, but hello again.

181

It is me, and I was just talking to you as myself, but that was actually yesterday, so very strange, but it at least lets me take a little drink

182

and take a break without you hearing me drinking my water and being weird on the stream.

183

So anyways, it looks like things are working now, so that's awesome.

184

The pod is actually deploying because we added the image poll secret,

185

and we can talk a little bit more about some peculiarities with images

186

and polling and things like that in different situations in the future.

187

One thing to keep in mind too is some people have like air-gapped Kubernetes clusters

188

and there's some special things that you need to do if you're going to set that up But this is Kubernetes 101, not Kubernetes 404, so we're not going to talk about

189

that in this series So we have the deployment running

190

and that's awesome But we need to actually use it right now if I go in my web browser, there's no way for me to actually get to this

191

So we need to expose the application to the world so that we can get to it from the outside So,

192

somebody says yesterday same shirt.

193

You notice it said 9am so technically I could be lying.

194

In my terminal it was saying that the times were around 9am.

195

So, but yes it was yesterday.

196

Using the magic of video editing I actually changed shirts.

197

I did not wear the shirt three days in a row.

198

So anyway, we're going to expose this application using Qubicuddle expose and I'll get into how this works in just a second.

199

So I'm going to expose this deployment that we created, HelloGo, and we're going to expose it on port 80,

200

although we won't access it on port 80 on the cluster itself because we're going to use a node port, which I'll talk about in a second.

201

And the target port, this is what it's going to look at in the containers running inside this deployment.

202

The target port is 8180, and you'll remember that if I go to cube 101

203

and go to episode 3 and go to the examples and go to the docker file.

204

That is the port that we're exposing and in the little go app

205

that we ran that is the port that it's listening on right here.

206

So 8180 is the port that we're going to use.

207

We're going to target request to that port and we're going to say type equals node port.

208

And one reason I'm using node port here in case you do know something about Kubernetes is

209

that this will work on the Kubernetes cluster running in Minikube

210

and it will run on a cluster running in Linode if you're doing it that way very easily.

211

And what this does is a node port service and

212

when you run kubectl expose it's going to create a service in Kubernetes.

213

A node port service will run in Kubernetes in its networking layer

214

and it will it will take requests from a port on every node in the cluster.

215

So it'll randomly assign a port on every node, and it'll take all that traffic into this little service, and it'll route those requests to one or more,

216

however many pods are running behind that deployment.

217

And you can expose, you can have services that expose different things.

218

It's not just for deployments, but for the most part in what we do, especially deploying web applications and things like that,

219

be exposing deployments and having services for deployments.

220

So we're going to create this,

221

and I can get the service information with kubectl get service and hello go.

222

And again, when you do things with the defaults, it's just going to take the name of the deployment and make a service for it,

223

just like it took the name of the deployment and added an app label to all the pods that run for it.

224

So we did that, and you can see that the port that it's running on is 32630.

225

So we can get to that.

226

Now this is the cluster IP.

227

This is an internal IP address using Kubernetes' own internal software-defined network that it runs on its own.

228

So we can't access that.

229

If I try going to that in my browser, I'm not going to get anything because it's a private IP address inside the cluster.

230

So this node port is actually on the IP address of any node running in the cluster, and for For Minikube, I can get that IP address with minikube.ip,

231

and I mentioned this in the first episode but might as well cover it again here.

232

This is the IP address of the actual Minikube instance running on my computer.

233

So I'm going to go here, I'll open up a browser window and go to the port that it's running on,

234

the node port, and I should get hello you requested slash.

235

So it's working and if I give it a different URL, it'll give me that.

236

So with Linode, it's a little bit different.

237

There's not an, I mean, there might be a command somewhere using a utility, Linode IP.

238

But for Linode, you just get the IP address of one of the nodes in your cluster, and that's the IP address you could use.

239

And you could use any of the nodes.

240

You can confirm that this is working by using all the different nodes in your cluster.

241

They'll all be exposing the service on this port.

242

So that's the simplest way.

243

If you have an accessible cluster on the internet or on your machine, a node port is an easy way to expose services

244

and test that they're running in a web browser or something like that.

245

So the next thing that we want to do, though, is a lot of applications like HelloGo.

246

This application can probably serve a lot of traffic very quickly, just running one instance.

247

But if we do have multiple nodes in our cluster, or if we want some redundancy

248

when we're doing upgrades of the application where new ones come in

249

and it's not killing off the old one before the new one comes, something like that, we can have multiple instances of it.

250

And that's very easy to manage with Kubernetes.

251

You don't have to create extra servers and things.

252

All you have to do is in a deployment, say, give me multiple replicas of this.

253

And by default, when you create a deployment, it will just give you one replica, which is one instance, but that's not very highly available.

254

And if you have multiple nodes in your cluster for redundancy, if one of the nodes dies and it's the one that was running that one instance,

255

that one replica, then your application will be down until Kubernetes identifies that and brings up another one in its place.

256

So that's never a good thing.

257

So we want to have multiple replicas

258

so Kubernetes can spread them across our cluster and allow them to use more resources and serve more requests.

259

The cool thing about Kubernetes services, thank you very much, Jay.

260

The cool thing about Kubernetes services is

261

that they will continue routing those requests to all the backend pods that you have running, all these different replicas.

262

So we can change that by editing the deployment, just like we did in the video a few minutes ago.

263

Qubicuddle, edit, deployment, hello, go.

264

And if you go down here under spec for the deployment, it's replicas.

265

Now, I didn't mention this too.

266

Kubernetes has a lot of, you're going to deal with a lot of YAML.

267

You can actually do things in different, different syntaxes.

268

You could do it all in JSON if you wanted and do, use different tools and things, but you're going to want to know the basics of YAML.

269

And I'm not going to cover that too much in depth other than what we need to know for this series.

270

But YAML is fairly simple.

271

The one thing to keep in mind is you need to mind your white space.

272

So if I go in Vim here and I'm adding in tab characters,

273

Vim is going to automatically detect that tabs are two spaces.

274

And YAML is in favor of spaces, as am I, but I know that I probably just lost two or three subscribers for saying that.

275

But anyway, Vim is going to auto detect that.

276

But whatever editor you're using, you need to make sure that you're not going to use mixed tabs and spaces because that can break the YAML formatting.

277

And it's important to have the right level of indentation in YAML, otherwise things can mess up.

278

Anyway, for this case, we're going to change replicas from 1 to 3.

279

And when I save this file, so I just changed it, and then I'm going to say WQ to write and quit.

280

And Kubernetes is going to check the syntax.

281

If there's a problem when you edit one of these files, Kubernetes will output the problem at the top of the file and leave you in the file.

282

As long as there's no problem though, it'll dump me back to the command line and start applying the changes

283

that I made to the cluster right away So I'm going to do that and then I'll say kubectl get deployment

284

hello go and We can see that already.

285

It's it seems like it already did the change

286

So I'm also going to say kubectl get pods dash L app equals hello go and make sure

287

So it looks like two of them looks like two of them.

288

These two new ones are already up and running So it's very quick, especially if you have applications that are lightweight and small and don't have to do a lot of setup work.

289

Sometimes applications might need to do extra setup work when you start running them in the cluster.

290

And we'll talk about that further down in the course.

291

But if they're lightweight applications and the container images are pretty small, then scaling up is extremely quick.

292

And later on in the series, we're going to talk about how to make this happen automatically in response to demand.

293

Like if you start receiving a lot of web traffic or if the CPU starts going high, you can scale up very quickly in Kubernetes.

294

And that's one of the best things that I like about Kubernetes is I don't have to worry about things like, you know, if I were on Amazon,

295

I'd have to set up an automatic load balancer and set up some rules for it and do the scaling.

296

And the scaling usually took a couple minutes per instance when they came up.

297

So this is super fast, super efficient and very nice.

298

Now, the other thing that you might be wondering is how did Kubernetes, like how is Kubernetes going to route requests to these things?

299

So if I go back over here and I'm refreshing this page, it's still working, but how do I know that it's actually going to spread the requests over all these different instances?

300

So one quick way to monitor in Kubernetes, the logs for all these pods, because in the, I think it was last episode,

301

I showed you that our application is actually logging all the requests, which is good because then we can debug the thing if we ever need to.

302

Kubernetes has a nice little handy kubi cuddle logs functionality built in

303

and I'm going to follow the logs with dash f just like I would if I were doing a tail

304

And then I'm going to use the same label app equals hello go

305

And the advantage of using labels is I can use this label

306

and it's going to get me all these different pods

307

because all of them have that label attached to them because they're all part of this deployment

308

And I'm also going to use another option called prefix equals true.

309

And what that does is it lets Kubernetes show me which pod and which container in the pod is running.

310

So if I do this, you can see that there's a bunch of requests from n92kc.

311

Those were the initial ones.

312

And those were this initial pod.

313

And then there's some from mrzxc.

314

Long term, as you go down this path, you'll notice that there's a lot of these kind of things.

315

you don't really think of these containers as pets anymore because you really can't, because you can't even identify them by a human-readable name.

316

So Kubernetes helps break you free of that and think of all of your applications running in cattle mode.

317

And really, if I delete this pod, in fact, I'll go ahead and do that really quick.

318

Qubicuddle, delete pod, that.

319

And let's go back once it finishes deleting.

320

Let's see.

321

So I'll go back to here, and you can see that it just brought up a new pod in its place.

322

Kubernetes doesn't really care what happens to your pods and knows that pods come and go, containers die.

323

That's not a big deal, and now a new one's running in its place.

324

Anyway, I'm going to go back and go into the logging and go back over to Safari,

325

and I'll refresh this you can see it seems like there's some sort of pinning based on the client

326

or the session or something because every request here

327

that I'm doing is going to Z7 but

328

if I open up another browser so I'll go to Chrome

329

and I'm gonna go to hello I'll go in Chrome

330

and go here and it's still doing that same one but yeah Let's try curl Go here.

331

Okay, so curl hit a different one. So

332

Kubernetes kind of divides things up You can actually get some more control over

333

that and we're going to talk later about something called ingress in Kubernetes

334

Which gives you even more control over how traffic comes into your cluster and is routed to backends

335

But Kubernetes services by default if I say kubi cuddle get service.

336

And another note that I should say here, because you'll see me do this just because my brain works in different modes,

337

you can also use shorthand for some of these things.

338

So SVC equals service, you know.

339

So instead of writing out service all the time, you can write SVC.

340

A lot of people also alias cubicuddle to K.

341

I don't do that just because when I train people on things, I don't want them to think that K is the way of things.

342

And, you know, you can set up that shortcut if you want, but I'm not going to do it here.

343

kubicuddle get service.

344

Hello, go.

345

And actually, I'm going to use dash o yaml.

346

Dash o changes the output of it, so this output is just a quick at-a-glance output for the screen, but for any object that's in Kubernetes,

347

you can get the yaml for it directly by just saying kubicuddle get service, or you could edit it too, but if you edit it and accidentally make a change, that'd be bad.

348

So I'm just going to get the entire output for how the service is defined in Kubernetes and how it's running.

349

And down here, you can see that it has this session affinity setting, and it's set to none.

350

And that means Kubernetes is doing whatever its default is, which seems like it's kind of all over the board.

351

You can see it, you know, a lot of requests went to this container, then it started going to this container.

352

I don't know exactly how Kubernetes routing engine works in the back end.

353

And as somebody who's worked a lot with like Varnish and HAProxy and Nginx and things like that.

354

There's a lot more control for whether you're using round robin or session affinity or IP affinity or whatever kind of thing.

355

So I don't know how this exactly works when it's set to none, but it seems like it's kind of doing something session affinity-like.

356

But anyway, you can get into that, and I'm sure somebody probably knows a lot more about the way

357

that service routing works in Kubernetes and could school me on that, but we aren't going to go deep into that today.

358

All right, so we have this application running, and we have the pods up.

359

There are three pods right now, and if I kubectl describe the deployment hello go,

360

I can see that right now it's running that version 1.0.0.

361

Now, the marketing team came to me and said, let me go over here, they said that this string of, let me go back to,

362

let me go back to an earlier version of it.

363

Oh, this is the initial version.

364

This is the new version of the app.

365

So marketing came back though, and they said that the version that I had for version 1.0.0, which is here in episode two from from last time,

366

it used the word hello.

367

Hello, you requested.

368

And they said, you know what?

369

In today's day and age, the word hello is way too formal.

370

We want to change things up.

371

We want to be more hip and cool.

372

And so you guys need to drop all of your work on everything else you were doing

373

because obviously maintenance and technical debt are not important.

374

It's more important to be hip and trendy.

375

So, um, and I, there are some good marketers out there, but I'm just having a little fun.

376

So anyway, they want us to change it to high because that's a lot more trendy.

377

That scores well on our A-B testing and it creates more engagement.

378

People stick on the site longer.

379

So yeah, we want to deploy that.

380

And so I created this 1.1.0 version that has that in it.

381

And we want to deploy that.

382

We want to roll that out to the cluster.

383

And so there's a few different ways that you can do that.

384

And long-term, we're going to get into ways

385

that you can automate this process with git or with a system like Jenkins or something like that.

386

But I'm going to show you two ways to do it.

387

One way is to say Qubicuddle edit deployment.

388

Hello go.

389

And if you go in here, you can go down to where the template is and then change the container image here.

390

I could just say 1.0.

391

There's a shorthand way to do it though.

392

Oh goodness, I'm screwing up my thing here.

393

Let me get out of here without.. no, quit.

394

All right.

395

Sometimes when I do Vim while I'm talking, I just can't.. my brain stops working.

396

Anyways, that's one way to do it.

397

The other way is to use some shorthand that kubicuddle has, and I can say, let's see,

398

kubicuddle set image, and that allows me to change the image for a given deployment or a pod or something like that.

399

So I'm going to say set image deployment slash hello go uh qb101 go equals

400

and where i'm getting this from is this is the definition of

401

that the container that's being uh used here

402

so you can see by default it takes the the name of the container not the namespace

403

and container but just the container itself and it labels the container name like

404

that you can specify the name of this container i could have called it my special app or something, but I didn't do that so this is the default that it uses.

405

So this is the container name and then I'm going to set it to gearlingguy cube101-go 1.1.0. Do that.

406

Then I'm going to watch the pods.

407

So watch kubicuddle get pods-l app equals hello go.

408

And I'm going to see that it already already brought up the three new pods

409

so since this is such a lightweight app it's very quick to bring up the new version even pulling down

410

that container it's it's only like six megabytes

411

but it was very quick pulling that down

412

and it terminated the old pods what kubernetes does by default

413

is it will bring up the new pods make sure they're running

414

and then after they're running

415

and working it switches the service to point to those new pods

416

and then it terminates the old ones

417

and it might take a few seconds to clean them up

418

but it does that and it's a rolling deployment

419

and you can actually control a lot about how that happens

420

and this episode i'm not going to deep dive into controlling rolling deployments

421

and things but if you look at

422

if you look at the deployment options here it does have

423

a few things here like strategy rolling update these are the defaults

424

that it sets up and it will it will make sure

425

that there are no more than 25 percent in this case there's three three replicas

426

so that would be one of the three i guess uh it makes sure

427

that there's no more than that percent unavailable at any given time

428

so while it's doing the update it's going to bring up a few it'll tear down a couple make sure

429

that all the new ones are running then it'll bring up a few

430

if you had 10 or 20 replicas i'm talking about

431

and it'll do that so that your application is never actually down which is awesome.

432

And I'm going to talk a little bit about some caveats there, especially because some applications that are out there might not take it

433

that well if you have multiple versions of it running at the same time.

434

So we have to work with our applications

435

and with our application developers to make sure that when they build things and when they update them, they aren't done in a way that can break when you're updating things.

436

So let me get back out of here.

437

and let's see I need to get back to where I am in the notes

438

because I've talked way ahead of where I have my marker all right

439

so we can also make sure that you know we saw

440

that the pods were rolled out I'm also going to say kubi cuddle describe deployment

441

hello go and make sure that the pod version is now 1.1.0 and I can also confirm in my browser browser, if I go here, it has hi.

442

Now, we deployed this, and this is not from experience in any way, of course not.

443

That's sarcasm if you can't tell.

444

But marketing just called up my manager and said, you know what, we just found out that the word hi actually means a very,

445

very bad word in some random foreign language that nobody actually uses and nobody accesses our site.

446

But you know what?

447

This is an emergency.

448

And I know that you guys are going to get back to fixing

449

that other huge bug that's causing you to have to do things every night.

450

But this is the biggest emergency ever.

451

And we're not going to admit that it's our fault that we caused all this work for you.

452

But you guys need to drop everything you're doing and fix that and roll it back to the previous version.

453

So luckily, Kubernetes has an easy way for us to do this.

454

And it's called kubectl rollout.

455

And when you do a deployment, when you change a deployment, Kubernetes rolls that deployment out.

456

And you can also view the history of deployments and also undo things or switch back to an older revision.

457

So we can get the history, just to show you, of the deployment.

458

Hello, go.

459

And it's going to show me that there are three revisions.

460

Now, there's this change cause field.

461

This might actually go away at some point.

462

There's a way to put a message in here.

463

It's a little complicated.

464

If you're building automation for it, you might want to set this up, but doing it by hand is a little annoying.

465

But there's a way that you can add a little message in here that shows like updated to version 1.1.0 or something.

466

But all we need to know is that the first revision was when we deployed this deployment initially, and it wasn't working.

467

The second revision was when we added the image pull secret and then it started working.

468

This third one was updating it from 1.0.0 to 1.1.0.

469

So I want to undo this previous,

470

this last revision, and the quickest way to do that is to do kubicuddle rollout undo deployment.

471

Hello, go.

472

And that's going to just roll back the deployment that I just did, and if I go up to the pods and watch them, you'll see that it already has done it.

473

It has the three new pods running, and we'll confirm with kubectl get described deployment,

474

we'll confirm that it has switched back at this point to 1.0.0.

475

And if I go in the browser, marketing is again happy because it says hello, even though apparently that's not very trendy and hip.

476

It's a safer word.

477

Now, obviously, the word hi, as far as I can tell, the word hi does not have any negative connotations.

478

But I am speaking from experience that these kind of things happen all the time.

479

And you have to deal with them.

480

So Kubernetes helps us to do these things and it's pretty nice

481

Now I mentioned earlier

482

and I'll mention it again Your applications have to be designed in a way

483

that lets you do these things To be able to deploy with a rolling update

484

and to be able to roll back if you have an application And I have some experience with this

485

Applications like Drupal and Magento.

486

I'm I'm traditionally a PHP developer

487

applications like

488

that what they would do a lot of times is you would apply an update to your system

489

and the update would all of a sudden change database schemas

490

and would make a breaking change like it would change the the column name of something

491

or it would change an entire table to something else and

492

so the new version would work with that

493

but then you're like oh my gosh I forgot there's a bug in this the schema translation

494

and we need to roll back so

495

when you roll back it's not just rolling back your application

496

you have to roll back a database to a previous state

497

and so things get really complicated if you applications that aren't built for this kind of rolling update.

498

So you need to make sure that you're talking with application developers

499

and using software that has the ability to update and roll back cleanly.

500

And one way to do that, and one way that I've built all of my applications that use things like databases, is instead of having an update,

501

modify a schema, what you do is you like create a duplicate table

502

or you add a new column and start using the new column in the new version, but also translating it and keeping the old column up to date.

503

And then at a future point, maybe three or four point releases down the line, then you drop that old column once you know all your new code is working.

504

So it requires a little bit more discipline, but it allows you to do this update

505

and rollback process a lot easier because you're not breaking your entire application doing a rollback.

506

So that's something to keep in mind.

507

It's one of the things that you start realizing, even if you're just using, even if you're deploying a big application in production in general,

508

everybody has had to roll something back at least once in their lifetime, if you've been in this industry long enough.

509

And I'm pretty sure that most of us have a painful story about a rollback that's gone horrendously wrong.

510

And anything that we can do to help prevent that is a good thing.

511

So let's see.

512

So that is all I have for today's episode.

513

And in the next episode, this was kind of a convoluted example.

514

It's really simple.

515

It's just a little thing that prints out a line.

516

It has no state that it has to track.

517

It has no persistence.

518

There's no database it's interacting with.

519

But it does show all the different things that you can do with deployments, and it's important to get a basic understanding.

520

And a lot of people have also asked me on Twitter especially, oh, are you going to talk about this magical tool

521

that you can use with Kubernetes or this system that you use with Kubernetes?

522

And I'm not going to get into some of those until very, very late in this course, mostly because just like with Git, I don't like the idea of using a tool

523

that provides magic layers on top of something until you understand the magic layers underneath.

524

So we're going to stay in the command line most of the time

525

and do things manually always first so

526

that we can understand how those manual parts work before we get into these fancy tools

527

that automate more for you and do more for you behind the scenes.

528

So next episode, we're going to take a look at an actual real world application running inside Kubernetes, in this case, Drupal.

529

And the reason I chose Drupal is there's actually two reasons.

530

One is the company that is sponsoring this Amazie.io.

531

I knew them a lot from back in the day from a lot of the Drupal things they do.

532

They do tons of different things too, but they have been highly involved in the Drupal open source community, as have I.

533

So I know Drupal really well, but it also is an application that is legacy and modern.

534

So it has some features that are hard to deal with in Kubernetes

535

and some features that are easy to deal with in Kubernetes.

536

And I want to show some of those things

537

because I know a lot of us are probably going to be running clusters

538

that run some apps that might be stateless and modern and fancy and amazing, but a lot of us are going to be running applications

539

that are more practical and old and have a history to them

540

that are not as easy to get running cloud native like Drupal.

541

And we're also going to find some of the new challenges that brings and how to deal with them.

542

And we're going to get a lot deeper into YAML.

543

So today was a very light introduction to YAML using Vim on the command line, but next episode we're going to get a little bit deeper dive

544

and I'll talk about a little bit how I make sure that my YAML is correct and things like that.

545

And thank you, Ricky Holland 87 as well.

546

It's great to see where everybody's from.

547

I am so happy with this course.

548

I've been planning it for so long.

549

Please consider subscribing if you want to keep watching these videos.

550

And again, everything is online at cube101.jeffgirley.com.

551

All the apps as I upload them will be available here.

552

And this will be a reference in case you want to go back and learn more.

553

So consider subscribing.

554

My book, Ansible for Kubernetes, is still available for $5.

555

and until next time, I'm Jeff Geerling.

556

Thank you.

557

Thank you.

📺 같은 채널

✨ 추천 영상

28 문장

Kubernetes Explained in 100 Seconds

Business EnglishUniversity StudentsWorking Adults

177 문장

Mini Project - Implementing a Simple Web Identity Federation Application

Business EnglishUniversity StudentsWorking Adults

545 문장

WHY YOU NEED TO GET OUTSIDE YOUR COMFORT ZONE | getting over your fears & anxieties + just doing it!

Business EnglishDaily ConversationUniversity Students

423 문장

Free CCNA | VLANs (Part 2) | Day 17 | CCNA 200-301 Complete Course

Business EnglishUniversity StudentsWorking Adults

78 문장

How I deploy serverless containers for free

Business EnglishUniversity StudentsWorking Adults

231 문장

Redis Crash Course - the What, Why and How to use Redis as your primary database

Business EnglishUniversity StudentsWorking Adults

214 문장

New Skills! v1.3 brings /pr, /implement-spec, and /retro

Business EnglishUniversity StudentsWorking Adults

106 문장

A Career in IT

Business EnglishUniversity StudentsWorking Adults

333 문장

Master n8n Triggers in 20 Minutes – Automate Anything

Business EnglishUniversity StudentsWorking Adults

747 문장

MIT Introduction to Deep Learning | 6.S191

Business EnglishUniversity StudentsWorking Adults

208 문장

My HomeLab Monitoring Rules // feat. Checkmk

Business EnglishUniversity StudentsWorking Adults

189 문장

"SHIP GOES HERE" for flight 14 - Starbase Summary

Business EnglishUniversity StudentsWorking Adults

실제 상황: 클라우드 클러스터 설정 가이드

이 비디오는 쿠버네티스 클러스터를 로컬이 아닌 클라우드에서 설정하는 방법을 설명하는 대화입니다. 발화자는 Minikube 설치에 어려움을 겪는 시청자들에게 Linode를 이용한 무료 클러스터 생성 방법을 알려주고, 단계별로 안내합니다. 일상적인 설명과 실용적인 팁이 섞여 있어, 기술적인 내용을 설명하는 대화의 흐름을 익히기에 좋습니다.

유용한 구문과 어휘 조합

have trouble + 동사-ing: "some people were having trouble installing it" (일부 사람들은 설치하는 데 어려움이 있었습니다). 문제 상황을 설명할 때 자주 사용되는 구문입니다.
work around: "There are ways to work around those kind of things" (그런 문제를 해결할 방법이 있습니다). 어려움을 극복하는 방법을 말할 때 유용합니다.
sign up for: "sign up for a Linode account" (Linode 계정을 등록하세요). 서비스 가입을 설명할 때 자주 사용됩니다.
switch over to: "switch over to the command line" (명령줄로 전환하세요). 다른 도구나 화면으로 전환할 때 사용하는 표현입니다.
configure for: "Minikube automatically configures our kubeconfig file for us" (Minikube는 자동으로 kubeconfig 파일을 설정해 줍니다). 설정 과정을 설명할 때 유용합니다.

당신의 쉐도잉 챌린지

영어 쉐도잉(shadow speech) 연습을 위해, 비디오에서 "creating a cluster is really easy"부터 "download the kube.config file"까지의 부분을 선택하세요. 다음 단계를 따라 해보세요:

원본을 2-3번 듣고, 발음과 억양, 말하는 속도를 파악합니다.
원본을 멈추지 않고, 1-2초 지연시켜 따라말합니다 (shadow speak). 발음 뿐만 아니라 강세와 쉼표의 위치도 주의깊게 따라하세요.
녹음기를 사용해 자신의 목소리를 녹음한 후, 원본과 비교합니다. 어디가 다른지 찾고, 다시 연습합니다.

이 연습은 IELTS 스피킹과 같은 실제 의사소통에서 자연스러운 흐름을 만들어 주며, 기술적인 용어도 함께 익힐 수 있습니다. 매일 5분씩 연습하면 빠르게 실력이 향상될 것입니다!

이 영상의 문법

화자가 가장 많이 쓰는 문형을 영상 속 실제 표현과 함께 정리했습니다.

문형	영상 속 표현
현재완료 have/has + 과거분사 — 과거의 일이 지금도 관련이 있을 때	you've created · haven't really talked · have made
수동태 be + 과거분사 — 누가 하는지보다 무슨 일이 일어나는지에 초점	be really outdated · are put · was just released
관계절 who / which + 절 — 사람이나 사물에 대한 추가 정보	Hub, which makes · secrets, which is · replica, which is

쉐도잉이란? 영어 실력을 빠르게 키우는 과학적 방법

쉐도잉(Shadowing)은 원래 전문 통역사 훈련을 위해 개발된 언어 학습 기법으로, 다언어 학자인 Dr. Alexander Arguelles에 의해 대중화된 방법입니다. 핵심 원리는 간단하지만 매우 강력합니다: 원어민의 영어를 들으면서 1~2초의 짧은 지연으로 즉시 소리 내어 따라 말하는 것——마치 '그림자(shadow)'처럼 화자를 따라가는 것입니다. 문법 공부나 수동적인 청취와 달리, 쉐도잉은 뇌와 입 근육이 동시에 실시간으로 영어를 처리하고 재현하도록 훈련합니다. 연구에 따르면 이 방법은 발음 정확도, 억양, 리듬, 연음, 청취력, 말하기 유창성을 크게 향상시킵니다. IELTS 스피킹 준비와 자연스러운 영어 소통을 원하는 분들에게 특히 효과적입니다.

섀도잉 방법: 단계별 전체 가이드 읽기 →

쉐도잉 연습: Kubernetes 101 - Episode 3 - Deploying Apps - 영상으로 영어 말하기 배우기

I'm making my Ansible and Kubernetes DevOps books free this month

Provision LEMP with Ansible on a Raspberry Pi Cluster (Pi Dramble #4)

I tried gaming on Linux...

Kubernetes Explained in 100 Seconds

Mini Project - Implementing a Simple Web Identity Federation Application

WHY YOU NEED TO GET OUTSIDE YOUR COMFORT ZONE | getting over your fears & anxieties + just doing it!

Free CCNA | VLANs (Part 2) | Day 17 | CCNA 200-301 Complete Course

How I deploy serverless containers for free

Redis Crash Course - the What, Why and How to use Redis as your primary database

New Skills! v1.3 brings /pr, /implement-spec, and /retro

A Career in IT

Master n8n Triggers in 20 Minutes – Automate Anything

MIT Introduction to Deep Learning | 6.S191

My HomeLab Monitoring Rules // feat. Checkmk

"SHIP GOES HERE" for flight 14 - Starbase Summary

실제 상황: 클라우드 클러스터 설정 가이드

유용한 구문과 어휘 조합

당신의 쉐도잉 챌린지

이 영상의 문법

쉐도잉이란? 영어 실력을 빠르게 키우는 과학적 방법