add slides

.gitignore

@@ -0,0 +1 @@
+/_static

slides.md

@@ -0,0 +1,337 @@
+---
+title: From 0 to Continuous Deployment with Kubernetes, Helm and GitLab
+theme: moon
+highlight-theme: github
+separator: ------
+verticalSeparator: ---
+---
+
+## From 0 to
+## Continuous Deployment
+
+**with Kubernetes, Helm and GitLab**
+
+<small>by Stefano Torresi</small>
+
+------
+
+### Continuous Integration, Delivery and Deployment
+
+- Merge developer changes into a shared mainline
+- Automate integration checks
+- Keep mainline in a production-ready state
+- Deploy mainline automatically
+
+---
+
+### But why?
+
+- Enforce best practices
+- Grant confidence and safety
+- Foster collaboration
+- Collect code metrics
+
+_"Release early, release often"_
+
+Notes:
+See "The Cathedral and the Bazaar" by Eric S. Raymond
+
+---
+
+## The pipeline
+
+![](assets/pipeline.png)
+
+---
+
+## Let's build something
+
+_Demo_
+
+Notes:
+```shell
+mkdir k8s-cd-demo
+git init
+git remote add origin git@gitlab.com:stefanotorresi/k8s-cd-demo.git
+```
+
+------
+
+## Kubernetes
+
+- Container orchestration platform
+- Infrastructure abstraction
+- Automates container coordination and management
+- API driven, declarative, stateful
+
+Notes:
+- Born from Google's Borg, which was written in C++ instead of Go.
+- Donated to CNCF in 2014
+- The API is fully transparent, there are no hidden internal details. This allows a very high degree of flexibility.
+
+---
+
+### Kubernetes Architecture
+
+![](assets/k8s-arch.png)
+
+Notes:
+- Losely-coupled, master-worker
+- Cluster centric, as opposed to how Docker Swarm is designed, which scales out from a single Docker host and then incrementally joining others.
+- Master is a.k.a. the cluster control plane.
+- etcd is used as the central cluster data storage.
+- The `cloud-controller-manager` interacts with cloud specific details via dedicated controllers, when hosting Kubernetes on a cloud provider.
+- Minions are handled by the `Kubelet` agent
+- `Kube-Proxy` handles the networking abstractions.
+- Master components are usually all run on the same node, which is separated from worker nodes, but this is ultimately an implementation detail. Minikube, for example, runs everything on one single node.
+- HA deployments of the Master involve clustering etcd and load balancing the API server.
+
+---
+
+![](assets/i-can-has-clasterz.jpg)
+
+---
+
+## Caveat Emptor
+
+- Very high cognitive overhead
+- Operating a control plane is hard
+- Easy to learn, hard to master
+- Multi-tenancy is a WIP
+
+Notes:
+Hard multi-tenancy can be achieved by having one cluster per tenant
+
+---
+
+## Managed master to the rescue
+
+AKS, GKE, EKS, DOKS, ICKS... all the Ks!!
+
+![](assets/minions-rejoice.gif)
+
+---
+
+## Managed master to the rescue
+
+```shell
+gcloud container clusters create foobar
+```
+
+```shell
+aws eks create-cluster --name foobar
+```
+
+```shell
+doctl kubernetes cluster create foobar
+```
+
+Notes:
+Skipping CLI options for brevity, but AWS is actually much more cumbersome than that.
+
+---
+
+_Demo time_
+
+Notes:
+```shell
+doctl kubernetes cluster create demo \
+    --region fra1 --count 2 --wait
+
+kubectx do-fra1-demo
+
+watch kubectl get nodes
+
+kubectl create deployment hello-world \
+    --image=
+
+kubectl expose deployment hello-world \
+    --port=80  \
+    --target-port=8080 \
+    --type=LoadBalancer
+
+kubectl get deployment hello-world -o yaml
+kubectl get service hello-world -o yaml
+```
+
+---
+
+#### Whole lotta APIs
+
+Container, Pod, Deployments, ReplicaSet, ReplicationController, StatefulSet, DaemonSet, Job, CronJob, Endpoints, Ingress, Service, ConfigMap, Secret, StorageClass, Volume, PersistentVolumeClaim, VolumeAttachment
+
+and many others.
+
+Notes:
+
+[API reference](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.13)
+[OpenAPI spec](https://github.com/kubernetes/kubernetes/tree/master/api/openapi-spec)
+
+- Pods are ephemeral.
+- Containers inside a Pod are co-located and run in a shared context, e.g. they share IP address and IPC.
+- Deployment, ReplicaSet, DaemonSet, StatefulSet, Job are all controllers.
+- Controllers provide a declarative approach to their workflow: the user declares a desired state, the controller ensures that state is present.
+- Services provide access to Pods, in spite these latter being ephemeral.
+- Ingresses provide external access to services, featuring load balancing, SSL termination and name-based virtual hosting.
+- The spec is big: 100.000 lines long!
+
+---
+
+## You can do it
+
+- Documentation is excellent
+- Plenty of resources available
+- Community is huge already and vibrant
+
+---
+
+## The YAMLPOCALYPSE problem
+
+**Deployment example:**
+
+```yaml
+apiVersion: extensions/v1beta1
+kind: Deployment
+metadata:
+  annotations:
+    deployment.kubernetes.io/revision: "1"
+  creationTimestamp: null
+  generation: 1
+  labels:
+    app: hello-world
+  name: hello-world
+  selfLink: /apis/extensions/v1beta1/namespaces/default/deployments/hello-world
+spec:
+  progressDeadlineSeconds: 600
+  replicas: 2
+  revisionHistoryLimit: 10
+  selector:
+    matchLabels:
+      app: hello-world
+  strategy:
+    rollingUpdate:
+      maxSurge: 25%
+      maxUnavailable: 25%
+    type: RollingUpdate
+  template:
+    metadata:
+      creationTimestamp: null
+      labels:
+        app: hello-world
+    spec:
+      containers:
+      - image: registry.gitlab.com/stefanotorresi/k8s-cd-demo:56487b78
+        imagePullPolicy: IfNotPresent
+        name: k8s-cd-demo
+        resources: {}
+        terminationMessagePath: /dev/termination-log
+        terminationMessagePolicy: File
+      dnsPolicy: ClusterFirst
+      restartPolicy: Always
+      schedulerName: default-scheduler
+      securityContext: {}
+      terminationGracePeriodSeconds: 30
+status: {}
+```
+
+---
+
+## The YAMLPOCALYPSE problem
+
+**Service example:**
+
+```yaml
+apiVersion: v1
+kind: Service
+metadata:
+  creationTimestamp: null
+  labels:
+    app: hello-world
+  name: hello-world
+  selfLink: /api/v1/namespaces/default/services/hello-world
+spec:
+  externalTrafficPolicy: Cluster
+  ports:
+  - nodePort: 32276
+    port: 80
+    protocol: TCP
+    targetPort: 80
+  selector:
+    app: hello-world
+  sessionAffinity: None
+  type: LoadBalancer
+status:
+  loadBalancer: {}
+```
+</div>
+
+------
+
+## Helm
+
+Kubernetes on steroids:
+
+- Package and dependency manager
+- Template engine
+- Release tool
+
+---
+
+### Helm Architecture
+
+- _Chart_: bundled application manifest
+- _Release_: running instance of a _Chart_
+- ```helm``` CLI: gRPC client, manages _Charts_
+- _Tiller_: gRPC server, manages _Releases_
+
+Notes:
+[interesting blog post to point people at](https://medium.com/@gajus/the-missing-ci-cd-kubernetes-component-helm-package-manager-1fe002aac680)
+
+---
+
+### Helm Architecture
+
+![](assets/helm-arch.png)
+
+---
+
+_Demo time_
+
+Notes:
+```shell
+kubectl apply -f tiller-rbac.yaml
+
+helm init --service-account tiller
+
+helm version
+
+helm inspect stable/external-dns | less
+
+helm install stable/external-dns \
+  --name external-dns \
+  --namespace kube-system \
+  --set provider=digitalocean \
+  --set extraEnv[0].name=DO_TOKEN \
+  --set extraEnv[0].value=$DO_TOKEN \
+  --set rbac.create=true \
+  --set rbac.serviceAccountName=external-dns \
+  --set policy=sync
+
+kubectl annotate service hello-world external-dns.alpha.kubernetes.io/hostname=k8s-cd-demo.torresi.io
+
+helm install stable/nginx-ingress \
+  --name nginx-ingress  \
+  --namespace kube-system \
+  --set-string controller.config.disable-ipv6=true
+```
+
+------
+
+## That's all folks
+
+Questions and feedback are welcome!
+
+- Twitter: [@storresi](https://twitter.com/storresi)
+- Email: [stefano@torresi.io](mailto:stefano@torresi.io)
+
+**Thanks**