Cloud Native Computing Foundation KubeCon + CloudNativeCon Europe 2020 - Virtual, 4 Sep 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: An introduction to Helm - Bridget Kromhout, Microsoft & Marc Khouzam, City of Montreal

Description

Don’t miss out! Join us at our upcoming events: EnvoyCon Virtual on October 15 and KubeCon + CloudNativeCon North America 2020 Virtual from November 17-20. Learn more at https://kubecon.io. The conferences feature presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

An introduction to Helm - Bridget Kromhout, Microsoft & Marc Khouzam, City of Montreal

CNCF Helm is a package manager for Kubernetes manifests that helps you write, share, find, and manage applications on Kubernetes. In this session you will learn the basics of Helm. We’ll dive into the components that make up Helm, and what a Helm Package looks like. We’ll then look at ways you can share Helm Charts and we’ll show you how to deploy an existing Helm Chart as well as how to get started writing your own Helm Chart. We’ll finish off by talking about some of the other exciting tools in the Helm ecosystem.

https://sched.co/Zx1Z

A

Welcome to the kubecon eu 2020 intro to helm, I'm bridget krumhout. I live in minneapolis podcast with requested devops lead the devops days conferences and I'm a pm for open source kubernetes ecosystem software at microsoft.

B

And I'm mark kuzam, I'm a core maintainer for the helm project. I work for the it department at the city of montreal canada, where we've been using helm in production for three years now, I'm particularly happy to be part of kubecon eu because, as a montreal native, I speak french so see.

B

B

All right, so so here's what we're going to talk about today. First we're going to go through some helm fundamentals like easily installing software into kubernetes, then we'll show you how to get started with charts, which are the packages used by hum we'll. Also look at the migration from health version, 2 to helm version 3 and that's that's very important for anyone already using helm, v2 and then we'll wrap up with a look at the helm community and don't worry, we'll also have concrete demos throughout the presentation.

A

All right, so, if you're tuning in because you're hoping to find out what even is helm and why do people use it uh home? Is a package manager for kubernetes like homebrew app yum, it's the de facto standard for packaging and deploying kubernetes applications and helm's superhero origin story? Is it started as a hackathon project at a small startup called deus in october 2015., shortly after the kubernetes project itself launched and then google joined the helm project, microsoft, acquired, dais and helm became very important to many many people.

B

Many people like, for example, the city of montreal, as I mentioned, we've been using helm since may 2017, and so we offer online services to our citizens and many of those services are deployed in production using help and now, since april 2020 helm is a fully graduated cncf project.

B

It has tens of thousands of github stars over two million downloads of helm cli every month. Did you know that helm is relied on by over 70 percent of kubernetes users? That's a lot of users.

A

Probably more by now, I think that was a number from a cncf survey, maybe a year ago, so it's probably even more now.

A

Let's talk about what helm does for all these users, starting by managing complexity, how it makes it easier to define your applications as yaml charts, so you can abstract away business logic and make your apps easier to install it. Also lets you update your applications in a consistent repeatable way so that you can mirror behavior across all your environments.

A

You can share public or private charts in a decentralized way and then there's rollbacks, which are technically roll forwards to a past state. I I cannot stress this enough. um Helm does not provide time, travel.

B

So this is important to understand with a rollback any data. That's been changed in your database or on your file system. It won't be changed back. It's not like restoring a backup or anything. What a rollback really does is. It makes it very easy to redeploy an old version.

B

Maybe because I don't know the latest version you deployed had a big bug in it and you want to go back to a version that worked before. That's what the rollback does.

B

Okay, but enough talk, let's get our hands dirty. Now, let's look at how to get started with helm, for the demos, we're going to assume that you already have helm installed. But if you don't, you can look at the helm documentation for the particular installation, details that are right for you all right, so for our first demo, we're going to show you how incredibly easy it is to install software into kubernetes when using helm, we're going to install a web server, so an nginx web server into our brand new namespace today.

B

So the first thing I'm going to do is make sure that I have helm properly installed. So, as you can see, version 324, which is the latest as of today, the next thing I'm going to do is make sure that I can speak to my kubernetes cluster helm can communicate to my kubernetes cluster and it does that using the variable cubeconfig and a queue config file, just like cubectl does in this case I'm going to use a kind cluster which is kubernetes in docker, which is a test cluster. Just for today's demo.

B

The first thing I'm going to do is create a namespace for our demo since we're installing nginx. Let me call it nginx namespace, and then I'm going to show you that nothing's in it so nginx get let's say service, as you can see, there's nothing get pod. There's nothing and get config map there's clearly nothing.

B

I just created in the namespace now we're going to use helm to install nginx, so the command is very simple: it's helm install I'm going to tell it which namespace to use so nginx namespace, I'm going to give my release a name. So let's call it engineer's demo. This is what I'm going to name what I'm the software. I'm going to run and the recipe for it the chart in helm, terminology which we'll talk about in the second demo is bitnami ngenx, so I'm gonna.

B

This is the recipe I wanna deploy and I can just send that command. Helm then creates all the yaml necessary ships it off to kubernetes and kubernetes, then creates all the proper entities.

B

So if I go back and I check for a service in my namespace, you can now see that oh, I have a new service for nginx demo for pods. I have a running pod for my nginx and if I look for a config map, I have a config map for my nginx demo, so I'm all set up with my nginx all ready to go. I can also use helm to get information about.

B

What's running in my cluster, that's been deployed with helm. So if I ask for helm list into that namespace, it shows me that I have one thing deployed. I can also ask for the status also in the namespace of my nginx demo, and then it prints me again. Some information, the name the last time it was deployed, the namespace in which it is current status, revision, etc, etc, and some information provided by the software itself beyond installing extremely easily.

B

You can also uninstall very easily so you've seen that multiple things have been created. A service, pods config maps deleting each one manually. If you want to get rid of that, nginx is a little more complicated, so helm provides you a way to delete the whole package at once, so you'll guess it's called the uninstall command, I also specify which namespace so I want to uninstall from that namespace and I just simply give the name of my release, which is called nginx demo, so I send that command and then helm simply wipes out everything.

B

So if I go back- and I look for my config map- it's gone- the service excuse me it's gone and the pod it's gone as well.

A

Click intro mark. So once we have the helm package manager installed, we want to manage packages with it, which means we need to find some of these packages, which are known as charts uh helm, has taken a decentralized approach for its charts, unlike docker hub, for example, which is a centralized registry of docker images, though not the only one uh helm instead has catalogs of different repositories like artifact, hub.io or hub.home.sh, and using a catalog. You can find the registries that have the charts you want to install.

A

It may sound complex, but we'll take a look at how this works.

B

So for our second demo, we're going to dive into charts and we're going to show you how you can configure a helm chart to set up and install the software you care about the way you want it installed on your kubernetes cluster before you get started, let me just show you a little bit about helm completion. Shell completion, which allows us to work with the helm, command line tool. Much faster, so helm provides the completion command, which allows you to source a script.

B

That's going to tell the shell all the available commands flags arguments and you can use it to complete your commands much faster. So, for example, once I set this up, I can do helm, install press tab and it's going to complete it for me. So you're going to see, I use it extensively throughout the demo. So now, let's get started on charts.

B

If you want to install a chart, you first need to know which repository it belongs to. So if I use the helm repo list command you're going to see that helm, 3 does not ship with a default repository, it's up to you to define which repositories you want to use, as we mentioned before, there are catalogs available like helmhub or artifact hub, where you can go and look for charts once you've found which repository they belong to.

B

You can tell helm about it, that's by using the helm, repo ad command, and in this case I want to install nginx, and I know it's on the bitnami repository, so I'm going to add it to helm. In this case, when I list the repose now I have a repository stored. If I want to install a different software, for example nexus, I would then look on the catalog. I would find out that it's stored in the sona type regis repository, and I can add this one as well.

B

So if I now list my repositories, I have two so helm can now access both and I can decide which to use to install what software I care about. I can also use the helm search command by using this. I can by specifying that I want to search my repositories. I can say that I'm interested in the nexus software and it's going to search all the repositories I have configured and tell me what are the names of the chart. So then I can know which chart I want to install so now.

B

If we want to install nginx, I would do helm search repo nginx and it will show me that I can find it under bitnami nginx. Let's see if we have it installed already helm list everywhere, all namespaces. I have nothing installed. So, let's install nginx like we have done in the first demo, so that's helm, install tab. I specify the namespace. I can use completion again.

B

I want to name it nginx demo, as we had last time and I want bitnami ngx.

B

So, as you signed the previous demo, it's going to install nginx for us and helm.

B

List will show us that it is now available. So what did I just do when I did and install bitnami nginx? What is this bitnami nginx chart?

B

What we're going to do is use the pull command to actually download the charts locally so that we can look at what it it contains, so helm pull and I'm going to tell it to immediately antar the download- and I use the same name nginx from bitnami.

B

Do this and now it has downloaded for me a repository called nginx and if I look inside that's the actual content of a chart, so a chart is made up: let's not go over everything, but the chart is made up of a chart file which describe the chart and a template directory which contains all the different yamls that helm will tell kubernetes to instantiate.

B

So in this case, there'll be a deployment there might be a horizontal pod, autoscaler, an ingress, a service, etc. There's also a values.yaml file which gives all the different configurable values that I can change from my own release of nginx, my own deployment of engineering. So if I look at that file, I'm sorry I have to go inside the nginx repository or directory.

B

If I look inside, I have all kinds of different tags that I can change. For example, by default, it specified that the nginx version- I'm installing- is this one 119 debian 10r0 and by default it has all kinds of different things like there's, no part annotation specified, but you could specify some the liveness probe, the readiness probe and many many more things.

B

So now that I have this, I also want to show you that you can look at this values file without having to actually download the the chart itself. You can use the helm, show command, show and helm show allows you to show all the details about a chart.

B

Only this charts file its readme file or its values file. So in this case, let's show the values file and again I specify bitnami nginx type it to less, and this same file that I just showed you helm shows it to me by downloading it off the internet. Just for my browsing pleasure.

B

All right, so we saw in the file here that by default the tag that I've installed for nginx is uh this version. So let's confirm that that's actually installed on my kubernetes cluster.

B

If I run this command- and I describe my deployment and I look at the image version- it actually tells me that yeah, that is correct. I've installed the nginx version, 119.1 debian, 10 r0.

B

Let's say that I want to upgrade to a newer version. I would use the helm upgrade command, so that would be helm upgrade. I can do the help and I get all the different flags that I can use to configure it to configure this command properly. So, let's go back and say helm, helm upgrade. I want to upgrade in the namespace that I've been using nginx namespace. I want to upgrade the release that I've been using nginxdemo.

B

I want to use that bitnami chart and linux, and if I leave it like this, it's going to reinstall the same version of nginx. So what I want to do is tell it no I'd like to use a different value for the tag, so we're going to use the flag set to tell the chart that I want to tweak it so that I set the image tag, yaml value to a different one. So here I found a more recent version, which is the r14.

B

Version so I tell nginx to I tell sorry helm to upgrade and it installs nginx again with the new values. So in this case it replaces the old one and if I do a helm list minus all all namespaces. I see that I'm on revision2, so I've upgraded my nginx version and if I look at cubectl and I again check the version, I should now see the new r14 version. Let's see if that worked, and yes, kubernetes are now aware that it should be downloading the r14 version.

B

The helm upgrade allows you to override certain values, like I've showed you directly on the command line with the set flag. But when you have many commands that you want to override it's better to use a file, so helm upgrade help again has here the minus, f or minus values flag, which allows me to specify a file where all my different overrides are set up. So, if I show you, I have here a demo values file.

B

Which contains a new version, so let's say I wanted to the r14 version. Had a problem I wanted to downgrade to r13. I can specify a particular pull policy.

B

I can set resources which are potentially not set already so, for example, let's look at the resources if I want to set this up, let's look at what's currently set up. So if I look at my deployment and I grep for limits, because I want to set limits, if I graph for limits currently they're not set, so in my upgrade and my override, I want to set these extra limits to to to control the nginx deployment.

B

The way I want so I would use the helm upgrade command, but in this case I would specify the minus f flag and, I would say, use the demo values.yaml file that I just showed you so it's kind of like using the minus set flag multiple times, so it will overwrite the image tag, the pull policy and set the resource limits. So let's do this and then.

B

I should point to the right file, which is the directory above and now again helm upgrades. The nginx version follows my instructions and now, if I do helm list, I can see I'm at revision three and at vision. Three. It should now have limits. Let's check do the same.

B

If I do a describe on my deployment, we actually look at the details. You see that now I do have limits and they have been set to what I had originally requested in my overwrite file.

B

If I want to look at the content of what's been installed by helm, I can use the helm, get command.

B

Helm get allows me to access my kubernetes cluster and download the values file that I've used or the actual yaml files that helm has sent to kubernetes. So in this case I would do helm get from the namespace. I want the actual command is. I want to get the values file.

B

I want to do it for my nginx demo and I'll pipe it to less, and then you can see that helm tells me when I installed last. I used these user supplied values.

B

If you want to look at the entire manifest the entire yamo that helm has sent, you can say, helm, get manifest again for the nginx demo and type it to less and you'll, see all the different things that helm requested kubernetes to create. So there was a config map. There was a service with different values. In them there was a deployment that was created.

B

A

B

For our nginx to work properly so now, you've learned how to configure charts using either a values, file or the set flag. But what, if the chart that you want to use, doesn't allow you to configure those particular things? Let's say you want to change something that is not configurable now you're in a little bit of a pickle.

B

What you want to do in that case is probably improve the chart itself. You probably want to make it more configurable than what it is right now, so you would download it using the pull command.

B

I would get the bitnami nginx chart. I would go into that directory and then I could actually edit the templates and modify them the way I want so that I can do more things than what the configuration of the chart currently allows, and if you really want to be nice, you might want to consider open sourcing, contributing that change to the chart itself, so that future deployments other people can do the same thing and benefit from your improvements once you've done that, let's say I've modified my nginx chart.

B

I can tell him to upgrade using instead of the repository instead of using here, bitnami nginx, I would say no, I want to use my local version, which would be dot so because I'm in the nginx repository I can then upgrade to the recipe that's stored locally on my disk helm makes an upgrade helm list minus a, and you can see now. I have yet another revision that has followed the content of the chart that I had locally on my disk.

B

The last thing I want to show you is what, if you have a product that you want to deploy into kubernetes? How do you have? How do you create a chart that might not exist already, so a helm helps you with this with the helm, create command, it's very simple, helm, create and simply give it a name. So I would do my product and helm creates a directory here called my product.

B

If I do tree it creates, as I had shown you earlier in the demo- a charts file, if it you want to use dependencies uh in other charts templates, is all the by these are example, templates. You would put the templates that you want to use the yamls that you want to use and a values file, and then you can set that up. And then you have your own configurable chart for your own product and then you're ready to go and install your product with helm into kubernetes.

A

Awesome demo mark, uh of course we can customize and even make our own charts and there's lots of details in the docs about that. So I have a link here. You can check that out.

A

So maybe you yourself are just getting started with home, but your organization already uses home too, and it works great. Why change.

B

Well, the helm project is nearly as old as kubernetes itself, so over the last few years, kubernetes has evolved a lot and, as a result, helms original design. Well, it predates many advancements in kubernetes like custom resource definitions.

B

Kubernetes are back so helm. 3 addresses all this.

A

Seriously, like your helm, 3 time is now we care about stability and reliability and version. 2 is already finished, with bug, fixes and it'll finish receiving security patches in november 2020..

A

So I understand everyone competes for your attention and it's hard to know which projects you have to undertake and if you do have helm 2 in your organization that you inherited now is the time to convince everyone that home three is the future or the present, as a case may be.

B

Absolutely so, what's really different about helm3. The most important change is that helm3 directly uses kubernetes api server and that's instead of helm's old server side component, which was called tiller with that helm, is significantly more tightly integrated with kubernetes, but really the real reason behind the getting rid of tiller is all about security.

B

Now the entire kubernetes are back system applies directly to helm. That was something very important for our users.

A

And hey, if you're new to helm, you may see names like galloping hippo in the docks and think what all those fun animal names are: no longer a default. But if you miss them, you can still opt in with generate name.

A

Otherwise, instead of names being auto-generated, you'll get an error unless you request auto-generation, so you have to produce or provide the name when uh doing a release.

B

All right, another improvement in helm3 is that we now have library charts a library chart, as the name suggests, is a chart that allows sharing functions across multiple charts. So, for example, you could set security policies in the library chart and then they would automatically apply to all your other charts uh library chart does not deploy anything itself, so you have to understand that what it does is define elements to be used by other charts.

B

So really it's about define, you'll see the actual code word define and- and these charts are really all about code, reuse,.

A

Yeah, and also, if you're new to home, when you're reading, docs and trying out tutorials, keep an eye out for that mention of tiller, which we mentioned before. um That means that those materials need to be updated for home three, because there's no tiller in home three. uh The short version as kubernetes has improved to the point where tiller isn't needed anymore.

B

So a side effect of removing tiller is that, although the releases are simpler, now they're actually not backwards compatible, so helm, v3 stores its metadata in the same name space as the resources it creates. So there's no more tiller name space for those that have seen it. None of that.

B

So so that's one of the incompatibilities between version two of helm and version three. So it implies that you can't use helm three with a helm, 2 release, you can't upgrade a helm, 2 release directly using helm 3..

B

What you must do instead is first convert your helm, 2 release to the helm, 3 format, using the helm, to three plug-in: that's provided by the helm community. Then you can start upgrading using helm directly.

A

Yeah and that's a good point, so let's take a look at what we can do with that helm, 2-3 plug-in all right, let's jump in. Let's look at helm version now with helm3. The first difference you note is that there's no more server version only the client, because there is no more tiller in helm3 and there was much rejoicing.

A

So our first step is going to be installing the migration plug-in into helm3.

A

And we can see it downloads it installs. We can see that the plugin is installed and another difference is that we can see that uh helm3 doesn't come with a stable repository by default. This is what we call foreshadowing we'll see what to do with that in a minute. So let's take a look at the migration plugin.

A

What it does is, it can convert releases and it can migrate configuration. You can also do some cleanup.

A

Let's try a dry run. This is dry run mode. It'll just show us what it's going to do. It won't actually do it, so you can see that it will create a bunch of folders, and you know, data directories will copy over anything that was in our helm to install.

A

So, let's take a look at what's in our home to install right now we have several repositories that are not the default repository we've added a few other ones again. Helm3 has no repositories by default. Let's look at our plugins in our home 2 cluster. We have a few plugins just to have some plugins.

A

What plugin do you think we have in home 3? Well, we have the one we just installed. We have the migration plugin, but that's it. No other plug-ins, okay cool. So what we can do here is we can do the migration. It warns us that we might lose v3 configuration. So you.

B

A

Obviously take precautions, but this has moved the configuration again from our home 2 to our home 3.. Let's take a look: hey look at that home 3. We now have all those repositories as an option and we have the new plugins from helm2, but it didn't overwrite the two to three plugin that we already had in our home three.

A

So that was plugins, but you may be saying: well, that's fine, but I have a bunch of releases that already exist in my home two cluster and I would also like to preserve them I'd like to move them over. Okay. So again, we've moved configuration, but let's take a look at what's in the helm2 cluster, we have some stuff that was released yesterday and just to make sure you see, we have nothing up our sleeve our helm. 3 has nothing installed in it. Right now so say we want to pick one of these.

A

um Let's take a look at what the convert tool does and, of course it has a dry run option as well. It has options for if tiller is in a different name space as options if tiller is not even running in your cluster. So, depending on what your helm 2 looks like, you may need some of those options, but we're going with a really simple path here: we're just going to convert that postgres release, we're doing it in drive red mode first, and we can see all right.

A

It's just going to create a copy of that postgres release, which we only had one version again. This is the simplest possible case, but you can extrapolate from this. So, let's take a look.

A

We converted the postgres release and note that it says that the v2 still remains so again. This doesn't do anything to our v2 cluster that we already had running. That was hope. You know presumably production and important, and we definitely didn't want to break in any way. It's all still there, but- and this is what where it gets interesting check it out- the helm3 stuff that we just brought over.

A

It was running in the old cluster a day ago and we've brought over that exact release that was updated at that exact date. We didn't do a new deployment of postgres. We brought over the one from before.

B

Nicely explained bridget. That was very good. So so, let's talk about what we just saw. V2 and v3 have different resource artifacts. They have different ways of working with kubernetes. So what does it really mean? It means that you must take the time to migrate. Your helm, v2 releases to helm v3. You cannot simply do a normal in place upgrade using hill.

B

That's the whole point of the 2-3 plug-in bridges has just shown you.

A

So if that seems like a lot the tldr is, you want to make sure you're using home three today, but we're here to help the community has written a detailed faq, that's on the doc site, and we even did some cncf webinars all about helm3 and the conversion process.

B

Absolutely- and I think people should know that helm is a very active and welcoming community. If you want to get involved, there are many ways to do so. You can improve documentation by doing documentation, pull requests. You can also take a look at existing issues and see if any interests you you could work on them.

B

Some of them are even marked as good first issue or help wanted. The bottom line is that we look forward to meeting you and don't be shy.

A

Yeah and the project twitter handle will tweet out releases and other news and kubernetes slack. You can also connect with the project on channel home, dash users and channel helm, dash, dev and channel charts.

B

And we also have mailing lists and a community call every thursday.

A

You could hang out with us every week and thank you for spending your time with us today. See you in the community.

B

And happy homing everybody bye-bye.