& Devops 2. Containerization and deployment on Kubernetes 3. Kubernetes distributions 4. APPUiO Cloud 5. Break 6. Workshop Agenda The agenda for today is the following. First we’re going to learn how Cloud Native and DevOps are shaping the world of software development in the cloud. Then we will see how to containerize applications, what it means, and how to make them run in Kubernetes. About Kubernetes, we will see the different flavors of it, how to relate with one another, and then we’ll talk a bit about Red Hat OpenShift. This will lead us to learn more about APPUiO Cloud, and of course I will try to reply to any question you might have. Then we’re going to have a 15 minute break, and when we return we’ll put all of this knowledge in action, using APPUiO Cloud to deploy an application. Speaker notes 2
the first question that comes to your mind is, what is APPUiO Cloud? Well, it’s a "Platform as a Service" environment, or actually better yet, "OpenShift Project as a Service". Speaker notes 3
I started my career as a software developer in 1997, deploying an application on the "cloud" (we didn’t call it that way back then) involved literally uploading files to a server, and refreshing a browser. Of course all of this was very prone to errors and issues, like for example two developers uploading the same file and overwriting each other’s code. Did I mention we did not have source control? There was no Git, no GitHub, no GitLab, no Visual Studio Code, no nothing. PHP was in its infancy, so we used whatever Microsoft gave us to build stuff. Speaker notes 7
These days we call "Cloud Native Applications" those apps created specifically to run in the Cloud. That could be a public cloud provider, a hyperscaler, or some other platform. There are three basic words that appear all the time when you talk about Cloud Native Apps: IaaS, PaaS, SaaS. Let’s see each of them in detail to understand where we are, and where APPUiO Cloud fits. Speaker notes 8
Resources All IaaS providers always, always, always offer these four basic building blocks to create new systems: Storage resources, to be able to keep large datasets (petabytes!) online with convenience and securely. Compute resources, to run apps that perform calculations on those data. Networking resources, to make those apps available to the Internet and to other apps as APIs. And security mechanisms, so that we can specify when, how, and by whom data is accessed. Speaker notes 9
services on the Cloud in a nice pyramid. At the lowest level we have IaaS, or "Infrastructure as a Service". In this level we find all of the big hyperscalers like Amazon, Google, and Microsoft. Their primary customer is a system administrator, system architect, or app operator, who needs resources to build new services in the cloud. Speaker notes 10
original diagram. On top of IaaS we find PaaS, or "Platform as a Service". That’s where we find APPUiO Cloud, see the logo over there. PaaS are built for developers, who do not want to spend time launching compute resources and wiring networking policies, they just want to write and run code! PaaS offer command line tools, nice user interfaces, and even software development kits (SDKs) for developers to create applications with. Speaker notes 11
pyramid, we have SaaS, or "Software as a Service". These are geared to users; they do not want to build stuff, they want to get things done. They need to create and share documents, collaborate, exchange ideas, and for that they use more and more cloud services. This is the final pyramid of cloud services in the public Internet. Speaker notes 12
with cloud computing in mind Cloud-native ≠ cloud only ⇒ public cloud and on-premises Focus on interconnected (micro-)services Enabled by open source implementations and open standards What is a Cloud Native App? In this new Cloud paradigm, we need a new type of application, and hence was born the concept of Cloud Native applications. They are apps that are built with those cloud concepts in mind. However, that does not mean that they only run on AWS or Azure; they can run as well on "private clouds", that is, private datacenters that offer enterprise developers with similar tools and opportunities as a hyperscaler. Usually, Cloud Native apps are built around microservices. Each component in a microservice architecture is small, focused, doing one and only one task. Finally, Open Standards (and Open Source implementations thereof) are at the heart of this technology world. Speaker notes 13
it would be hard to find a better guide than the 12 Factor website at . The principles in this website provide a good practical starting point for developers looking forward to create modern Cloud Native applications. Speaker notes 12factor.net 14
automation Suitable for modern cloud environments Minimal difference "dev" vs "prod" ⇒ continuous deployment Twelve-Factor Patterns For example, among the various principles, we find that Cloud Native apps should be described with textual formats, which leads to automation. Cloud Native apps should be portable across environments; that is, ideally we should be able to avoid platform lock-in and be able to migrate to other cloud providers as required. These applications should be built with those cloud environments in mind, in terms of security, connectivity, latency, and other factors. There should be almost no differences between development and production environments, and ideally, there should be a continuous flow of applications from one environment to the other. Finally, they should be scalable. Scalability should be a primary concern, so that apps can respond and adapt to traffic spikes. Speaker notes 15
Maximum automation through "infrastructure as code" Cost efficient ⇒ Lean Agile ⇒ react to changing requirements faster Continuous improvement built-in DevOps But what kind of teams are able to create Cloud Native apps? The "DevOps" movement, itself an evolution from the ideas of the Agile Manifesto in 2000, brings a social framework for such applications to exist. First of all, by ensuring communication between development and operations. There shouldn’t be barriers among those different groups. The use of automation and "infrastructure as code" ensures that every element in the infrastructure is written down in text files, versioned in Git repositories, and handled with automated tools. Being lean means being cost efficient, taking advantage of automation to lower costs. Being agile means being more reactive to the demands of customers and stakeholders. And finally, improving continuously with new features, bug fixes, in a continuous flow of work from development to production. Speaker notes 16
like to learn more about DevOps, the most important book I can recommend is "The Phoenix Project", a novel about DevOps that has become a classic in the genre. It is very well written! Speaker notes 18
Automatic releases Faster time-to-market & more security Infrastructure as Code Repeatability & testability More Observability Resilience Best Practices for Cloud-Native In summary, when you want to create a Cloud Native application you should follow these best practices. But regarding the first point in the list, how can you componentize your apps in microservices? Turns out, containers are the solution for that. Speaker notes 19
reminds us of these things, seen at harbours and train exchanges all over the world. Why do we use the word "containers" for software nowadays? Speaker notes 21
"intermodal" freight transport Help reduce cost and times of transport Universal standard open to anyone Ecosystem of vehicles, training, accessories, tools… Shipping Containers It turns out that the invention of standard shipping containers have been one of the major inventions in trade of the 20th century. They are "intermodal", which means that you can put them on a train, on a boat, on a truck, and they will work; it does not matter what they contain, they can be moved from support to support. This of course reduces the cost of transport, making it faster and more reliable. And it’s a universal standard, open to anyone, which means that companies that support it can immediately join a large… … ecosystem of vehicles, trainings, accessories, tools, etc. Speaker notes 22
app + dependencies + libraries + runtimes + configuration + … Runtime isolation through OS-level virtualization Not a new idea: Version 7 Unix "chroot" (1979-1982), FreeBSD "jails" (2000), LXC (2008), Docker (2013) Application Containers Software containers aim to reach the same kind of productivity for software as shipping containers did for international trade. They are standardized by independent bodies, and their specs are open to anyone. An application container is completely independent, and carries everything that it requires: libraries, runtimes, configurations, etc. They provide some isolation, which means that their behaviour is guaranteed, and can be controlled from the outside. And no, they aren’t really a good idea; Docker provided the final element required for them to become popular, but the chroot mechanism already existed in Unix Version 7, back in 1979! Speaker notes 23
tools used today by developers to create standard application containers are Docker and Podman, both free to use and very similar to one another. Speaker notes 25
of application hosting. On the left we have the situation in 1997, where we would have a hardware server in a datacenter, and we used FTP to deploy code into it. In the 2000s virtualization became very popular, with solutions like VMWare and Vagrant, which provide isolation and much more flexibility. But how are containers different from virtual machines? It turns out that containers share the kernel with the host, which makes them faster to start and stop, and of course, less hungry in resources. Every VM must have its own copy of the operating system kernel, of course. As a downside, of course, containers must run the same operating system as the host; that means that, for example, Windows containers cannot run in Linux hosts, and vice-versa, while Windows virtual machines can run on Linux hosts. Speaker notes 26
communication Continuous Integration ⇒ Continuous Deployment Speed: fast to start, fast to stop Portable: across operating systems, hardware, cloud platforms, environments… Isolation: controlled resource utilization Security: each container runs isolated from others Benefits of Containers Containers have so many benefits that the industry has adopted completely. Here’s a list of the major advantages. Speaker notes 27
containers; one image can be used to create many containers. Docker Containers Running instance. Each container is created from one image. "Containers" and "Images" A common confusion among engineers new to containers is the distinction between "Docker Images" and "Docker Containers"; here’s the difference. Speaker notes 28
r := gin.Default() gin.DisableConsoleColor() r.LoadHTMLFiles("templates/fortune.html") hostname := hostname() version := "1.4-go" r.GET("/", func(c *gin.Context) { contentType, response := buildResponse(c, hostname, version, false) respond(c, contentType, response) }) r.GET("/metrics", gin.WrapH(promhttp.Handler())) return r } gitlab.com/vshn/applications/fortune-go Here we have a very simple web API written in Golang, using the Gin web framework. A perfect candidate to be containerized! Speaker notes 29
github.com/go-playground/validator/v10 v10.9.0 github.com/golang/protobuf v1.5.2 github.com/json-iterator/go v1.1.12 github.com/mattn/go-isatty v0.0.14 github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd github.com/prometheus/client_golang v1.14.0 github.com/ugorji/go v1.2.6 golang.org/x/crypto v0.0.0-20211108221036-ceb1ce70b4fa golang.org/x/text v0.3.7 golang.org/x/tools/gopls v0.7.3 gopkg.in/go-playground/assert.v1 v1.2.1 gopkg.in/go-playground/validator.v9 v9.29.1 gopkg.in/yaml.v2 v2.4.0 ) As most software projects, this Golang application has some dependencies, which are handled in a standard "go.mod" file. Speaker notes 30
binary FROM alpine:3.14 AS build RUN apk update && apk upgrade && apk add --no-cache go gcc g++ WORKDIR /app COPY go.mod ./ RUN go mod download COPY *.go ./ RUN CGO_ENABLED=1 GOOS=linux go build # Step 2: deployment image FROM alpine:3.14 RUN apk update && apk upgrade && apk add --no-cache fortune WORKDIR /app COPY --from=build /app/fortune-go /app/fortune-go EXPOSE 8080 USER 1001 ENTRYPOINT ["/app/fortune-go"] To create a new container image for this Golang application, we need a Dockerfile like the one shown here. Speaker notes 31
1143a36ec66b ---> 049aba3fb8c2 Step 7/10 : COPY app.py . ---> 22f4e095f281 Step 8/10 : USER 1001 ---> Running in 50359efd384f Removing intermediate container 50359efd384f ---> d42d57937270 Step 9/10 : CMD [ "python", "app.py" ] ---> Running in 42d120bc89f1 Removing intermediate container 42d120bc89f1 ---> d1746acd7dff Step 10/10 : EXPOSE 9090 ---> Running in 3a32121e27e5 Removing intermediate container 3a32121e27e5 ---> 72b4eca87a0b Successfully built 72b4eca87a0b Successfully tagged vshn/fortune-cookie-service:1.0 asciinema $ dock 00:00 This demo movie available online at shows how to use the Dockerfile to create a new container image, and then how to run a new instance of a container with the new image. We see how it can be also pushed to Docker Hub, so that other developers can run it and use it if needed. Speaker notes asciinema.org/a/333322 32
build .base: stage: build image: docker:stable before_script: - docker login -u "$CI_REGISTRY_USER" -p "$CI_REGISTRY_PASSWORD" $CI_REGISTRY docker-build-master: extends: .base script: - docker build --pull -t "$CI_REGISTRY_IMAGE" . - docker push "$CI_REGISTRY_IMAGE" only: - master You can also use your Continuous Integration and Continuous Deployment (CI/CD) system, in this case GitLab, to build your application automatically into a container image, and then push this image to an image repository. GitLab includes its own container image repository. Speaker notes 33
Versioned in repository Sharing Dockerfiles How do you share your Dockerfile with your colleagues? Very easy; you add them to your projects, as one more file for your team to include in your Git repository. Speaker notes 34
available to team members, outside contributors, or other deployment environments Public & Private Ready to run Sharing Container Images Of course you might want to only share the container, and not the Dockerfile with your colleagues or customers; in those cases, you can use what is called a "Container Image Repository", where you can make images available to others, either in public or private fashion, and where they are ready to run. Speaker notes 35
Red Hat Quay Amazon ECR GitHub On-Premises GitLab OpenShift Harbor Container Repositories There are many container repositories available; some are public and free, available online; like Docker Hub, GitHub, or Red Hat Quay. Others are installable on-premises, like for example Harbor or the ones included in GitLab and OpenShift. These ones are private, so that developers of a same organization can share containers with one another. Speaker notes 36
components: Databases Web servers Message queues Monitoring systems … all deployed in various environments: "dev", "test"… … all connected to each other… … and with redundancy! Running Many Containers Complex Cloud Native applications are made of multiple containers: a database, web servers, application servers, message queues, monitoring systems, and all of these apps must be installed in various environments, they must all be connected to each other in similar ways, and they must all support redundancy and scalability! Clearly, it becames very complex to run all of this by hand. Speaker notes 39
a container… … each with its own IP or port… … each loosely connected to one another… … we need a "Container Orchestrator" How to Coordinate those Components? To run many containers together at once, in a repeatable fashion, and with automation, we need something that acts like a "container orchestrator." Speaker notes 40
Container orchestrator platform originally created by Google Open Source and cross-platform Part of the CNCF ecosystem Latest version: 1.26.2 (March 1st, 2023) What is Kubernetes? And precisely, Kubernetes is a container orchestrator. It is not the only one (there are others like Docker Swarm and Apache Mesos) but it has become a standard in the industry. Pretty much every major application you use in the Internet these days is running in a Kubernetes cluster. Speaker notes 41
Meaning: governor, commander, or captain Root of "government" and "cybernetics" How do you pronounce it? Here’s the funny part; you’re going to hear lots of various pronounciations of the word, but this is the true one in Greek. Speaker notes 42
together. Node A machine (virtual or not) in a cluster. Pod Minimum unit of code execution, with one or many containers inside, and which can be scaled up and down. Ephemeral! Kubernetes Terminology 1/3 Kubernetes manages a cluster of compute resources, usually virtual machines provided by some IaaS. Each of those resources is called a Node. One (or two) of the nodes in a Kubernetes clusters are called "Master Nodes", and their job is to work as the director of the orchestra (after all, it’s a container orchestration system, get it?). Master Nodes also expose the Kubernetes API, used by DevOps engineers to "talk" to the cluster and to manage it. The other nodes are "Worker Nodes", and their job is to run containers following the instructions of the Master Node(s), and that’s it. Inside of those nodes, Kubernetes runs containers inside of small units called Pods. A pod can contain one or many containers, running together. Pods can be killed, scheduled, shuffled around from node to node, as required and as possible, depending on the conditions of the cluster. Kubernetes can for example restart pods if they crash, automatically. Containers inside of a pod can talk to each other using localhost and specifying the port number of each application exposed inside of a container. Speaker notes 43
"Services" and "Persistent Volumes" running in the nodes of a cluster. Service Used to expose a deployment to the internal network of the cluster. Ingress Exposes services to the wider Internet. Kubernetes Terminology 2/3 Pods are grouped together into Deployments. A deployment specifies the identity and number of containers inside of each pod, and provides information to Kubernetes about replication, scheduling, and various types of metadata. By default, deployments are not exposed to other deployments in the cluster, or to the outer world of the Internet; for that, you must create a Kubernetes Service. Services expose a network port and address for deployments to talk to one another. They can also provide load balancing, for applications to have a higher level of availability. Services only work at the level of a cluster; to expose things to the outer world of the Internet, one must use an Ingress. Ingresses are Kubernetes objects that expose and create a route for applications to be reachable from the outer world. Speaker notes 44
disk storage available for pods to save data. Persistent Volume Claim A declaration from a deployment about storage requirements, that Kubernetes tries to fulfill as best as possible. Namespace Mechanism for isolating groups of resources within a single cluster. Kubernetes Terminology 3/3 In order to store data, deployments require Persistant Volume Claims or PVCs. They are requests for Persistant Volumes or PVs, that the cluster tries to fulfill if possible. DevOps engineers setting up a Kubernetes cluster can configure storage providers, so that PVCs are fulfilled appropriately. This level of abstraction clearly separates applications from the storage they require to save data. Last but not least, namespaces provide a way to group objects within the same cluster. Object names must be unique in a namespace, but not across namespaces; we’re going to use this capability in the workshop today. Speaker notes 45
Pods. Storage with PVs and PVCs. Network with Services and Ingress. Security with… well, nothing really. Kubernetes does not provide any user management features! Kubernetes as an IaaS Remember what I said in the beginning; IaaS providers always provide basic building blocks, but Kubernetes only provides part of the solution. User management is one of the features that the Kubernetes project creators decided to leave to implementers to add to their own Kubernetes distributions. This makes the Kubernetes project very similar to the Linux Kernel project. The Linux kernel provides useful features like process scheduling and support for file systems, but if you want to edit documents, you’re better off installing a distribution that includes LibreOffice; and that’s what Red Hat Enterprise Linux or Ubuntu provide on top of the kernel. Speaker notes 46
24s fortune-deployment-7b6b9f6596-hqkqm 1/1 Terminating 0 2m39s asciinema $ curl $(minikube service fortune-service --url) Fortune cookie of the day #441: Saw a sign on a restaurant that said Breakfast, any time -- so I ordered French Toast in the Renaissance. -- Steven Wright asciinema $ kubectl delete -f fortune-service.yaml service "fortune-service" deleted asciinema $ curl $(minikube service fortune-service --url) 💣 Service 'fortune-service' was not found in 'default' namespace. You may select another namespace by using 'minikube service fortune-service -n <namespace>' . Or list out all the services using 'minikube service list' curl: try 'curl --help' or 'curl --manual' for more information asciinema $ minikube delete 🔥 Deleting "minikube" in docker ... 🔥 Deleting container "minikube" ... 🔥 Removing /home/akosma/.minikube/machines/minikube ... 💀 Removed all traces of the "minikube" cluster. asciinema $ 00:00 This video, available at , shows the deployment of an application to a Kubernetes cluster. The application shown is called , it’s a TUI (Text User Interface) tool that is very popular to manage Kubernetes clusters. Speaker notes asciinema.org/a/333331 K9s 49
that manages clusters Clusters consist of Nodes Kubernetes runs Deployments in Nodes Deployments usually consist of Pods, Services and Storage Services expose network ports to the outside world Pods consist of Containers Containers are built from Images Images are built with a Dockerfile So, a short summary about Kubernetes here. Speaker notes 50
APPUiO Cloud ("OpenShift Namespace as a Service") Amazon Web Services: Elastic Kubernetes Service (EKS) Microsoft Azure: Azure Kubernetes Service (AKS) Google Cloud: Google Kubernetes Engine (GKE) Kubernetes Distributions There are many flavors of Kubernetes, just like there are various flavors of Linux. Amazon, Azure, and Google Cloud all offer a "managed Kubernetes" solution, which is a great way to have a Kubernetes cluster in a short amount of time. Speaker notes 52
kind SUSE K3s Canonical Microk8s Red Hat OpenShift CodeReady Containers (CRC) Kubernetes in your Laptop If you would like to try Kubernetes, no need to pay for it at any of the big hyperscalers; you can use any of these services to run a small Kubernetes cluster within your laptop. Speaker notes 53
of the Kubernetes distributions above, one of the most popular among big enterprises is Red Hat OpenShift. It offers lots of very interesting features, and this is the reason why we have chosen it to be the basis of APPUiO Cloud. Speaker notes 54
One-click app deployment CI/CD Observability Integrated container registry Enhanced security settings kubectl ⇒ oc Openshift Features Let’s take a look at some of the things that OpenShift adds to standard Kubernetes off-the-box. It even provides its own tool, called oc instead of kubectl. We will use this tool during the workshop this afternoon. Speaker notes 55
gist of this presentation. Now we’re going to talk about APPUiO Cloud, and the various challenges we had to face to put the system into production. Speaker notes 56
brand and a series of products offered by VSHN together with our partner Puzzle ITC, an IT service provider from Bern. This collaboration started in 2016. And what does the word "APPUiO" mean? It’s a word in Esperanto meaning "Support". Speaker notes 57
Managed Kubernetes Services EKS, AKS, GKE Self-Managed APPUiO "Flavors" There are several types of APPUiO. APPUiO Cloud is billed per usage; that is, you only pay for what you use. APPUiO Managed consists of offering OpenShift clusters to their customers, entirely dedicated, so that nobody else can use them. (Of course this is more expensive!) VSHN and Puzzle ITC take care of the management of the cluster (updates, backups, security, etc.) APPUiO Self-Managed is like APPUiO Managed, but we train the customer to be able to manage the clusters by themselves, which can be interesting for bigger customers with a dedicated IT team. Speaker notes 58
Backup Pre-Installed and Configured Operators Community Support Support packages available at extra cost APPUiO Cloud Features APPUiO Cloud has various interesting features that set it apart in the market. Speaker notes 59
DevOps & CI/CD Pipelines Machine Learning Production App Hosting Mobile App Backends Education! Target Audience Who have we built APPUiO Cloud for? These are typical customers who need quick access to a managed OpenShift 4 cluster, without the hassle of installation and management. Speaker notes 60
guarantees of resource availability SLA: Best-effort. Fair-Use Policy Privileged containers can’t run on APPUiO Cloud Log retention: 72 hours max, then deleted Shared Platform Restrictions status.appuio.cloud APPUiO Cloud is a shared platform; the same way you would buy "shared PHP + MySQL hosting" for your hobby website in Hostpoint or Infomaniak, you can get a "shared OpenShift cluster" for your enterprise in APPUiO Cloud. This means that there are a few limitations and restrictions in place. Speaker notes 61
the APPUiO Cloud zones are managed through a single Identity Provider (IdP); we use Keycloak for that. Keycloak is an open source software product under the stewardship of Red Hat, written in Java, to allow single sign-on with Identity and Access Management aimed at modern applications and services. Whenever a new user requests an APPUiO Cloud account, they get an account at our Keycloak, and that allows them to access any of the APPUiO Cloud zones without distinction. The other logo belongs to Kyverno , a policy engine for Kubernetes which we use to enforce policies and quotas in APPUiO Cloud. Speaker notes www.keycloak.org kyverno.io 64
documentation at Documentation for end users at Documentation products.docs.vshn.ch kb.vshn.ch docs.appuio.cloud VSHN is a document-driven company. We document everything we do, at every step, and this attitude has helped us a lot during the pandemic; new hires could find all the information they need to start working, and whenever somebody goes on holidays, we don’t need to call them to ask how to do stuff on their behalf. So as expected, we have created lots of documentation for APPUiO Cloud; in fact, we have documented in written every single step of the creation of the platform. And even better, all of this documentation is freely available online: These three websites are completely open, free, and ready to read; actually, this very presentation is in a way a summary of the information contained in those three websites! Speaker notes 67
quick overview of APPUiO Cloud before we get to use it in a real-world scenario. There’s a lot of information out there for you to learn more about it! Speaker notes 68
Description Status Portal kb.vshn.ch/appuio-cloud/ docs.appuio.cloud/user/ products.docs.vshn.ch/products/appuio/cloud/ status.appuio.cloud portal.appuio.cloud Here are some links where you can find all the information you need about APPUiO Cloud. Speaker notes 69
akosma
ysknsid25
asei
fujiwara3
bgpat
nullpo_head
tarotaro0129
fnifni
shujisado
muziyoshiz
duelist2020jp
mochizuki875
fujisawaryohei
addyosmani
tanoku
quramy
tammielis
speakerdeck
bkeepers
tammyeverts
kneath
trishagee
swwweet
sergeychernyshev
colly
