Downloaded 17 times
More Related Content
Similar to Kubernetes Introduction
Kubernetes Introduction
- 1. Kubernetes Introduction Milos Zubal WaterfordTech. Meetup 31.01.2018
- 2. Who’s this?! @MilosZubal Software Engineerat (we’re hiring! ;) ) 10+ years of backend Java happiness https://www.linkedin.com/in/miloszubal/ https://github.com/mzubal https://twitter.com/MilosZubal
- 3. What are we goingto cover today? ● Big Picture of Kubernetes (k8s) ● History and main features of k8s ● Small recap of containers ● Architecture of k8s ● Main k8s primitives + examples ● Other features ● Running Kubernetes ● Demo ● Bonus: Czech accent all along the way! :) (all of that in 30-35 minutes - managers would call this “ambitious plan”)
- 4. Why Microservices? Totame the kludgy Monolith? (the whole story at: https://cloud.google.com/kubernetes-engine/kubernetes-comic/)
- 5. Let’s go Microservices,but what about... ● Service Discovery ● Load Balancing ● Networking ● Storage ● Monitoring ● Clustering ● Logs ● Many other things...
- 6. Kubernetes - BigPicture
- 7. Is this likeServerless? ● Better term would be “Server-agnostic” ● K8s API tries to be as server-agnostic as possible ● K8s API is declarative ○ You defined the desired state rather than actions
- 8. Kubernetes - fewfacts ● Open-source container orchestration platform ● Kubernetes == greek for “Helmsman” or “Pilot” ● Often abbreviated as k8s ● Based on Google’s internal system (Borg), which reflects their experience of running containers for more than 15 years ● First released in 2015 and later donated to the Cloud Native Computing Foundation ● Big and vibrant community ● Biggest contributors - Google and RedHat
- 9.
- 10. Containers recap ● Containers== lightweight “VMs” ● They are basically cleverly isolated processes running on the same OS/kernel (using cgroups/namespaces) ● You can adjust networking/volumes between host and containers ● Current container techs. - docker, rkt, lmctfy
- 11.
- 12. Kubernetes - Architecture ●Master - a node holding the configuration and managing the cluster ● Node - a worker machine (bare metal, virtual, cloud) hosting Pods ● Kubectl - console for the API, allows modifying/uploading configuration ● API - REST api to modify the configuration (which is stored in etcd) ● Scheduler - gathers info about Nodes (utilisation) and decides where to spawn new Pods ● Controller Manager - Checks and enforces the current config in the cluster (e.g. ReplicaSet) ● Kubelet - agent running on each Node, providing info to master, creating Pods ● Kube-proxy - serves for inter-Pod and inter-Node communication ● Pod - container(s) running the actual application
- 13. Pod - containsthe app ● Basic unit in k8s, can expose ports ● One or more containers sharing the same namespace, resources, volumes, network (localhost), ports, etc. ● Each container is defined by its image ● Limits for CPU / RAM ● Set of labels (we will learn more about them later) ● Can have health / readiness checks ● Containers in pod get automatically restarted in case they crash or fail the health check ● Pod might get rescheduled to other Node ● You can also mount Volumes to Pods or inject some configuration (via ConfigMaps and Secrets)
- 14. ReplicaSet - scaling/ self-healing ● A set of Pods based on a Number of Pods, Pod Template and a Selector ● Allows defining clusters of the same Pod (application) ● Number of Pods - the desired number of replicas to be running in cluster ● Pod Template - basically the same like Pod definition (image, labels, resources) ● Selector - a set of matchers to match the Pods in cluster by labels ○ E.g. “app: elastic AND version: 6.1.2 AND env: prod” ○ Usually matches the labels defined in the Pod Template ● Kubernetes makes sure there is the exact number of defined Pods in the cluster
- 15.
- 16.
- 17.
- 18.
- 19.
- 20. Service - servicediscover / load-balancing ● Addresses/ports of Pods are dynamically assigned every time a Pod is created ○ You cannot rely on them when trying to call an app running in other Pod ● Service solves the problem by providing a stable clusterIp/port for Pod(s) ● The Service is defined by a selector (the same as in ReplicaSet) and a port ● E.g. “app: elastic AND version: 6.1.2 AND env: prod” + port: 9200 ● 3 modes of operation to choose from: userspace, iptables (default) and ipvs ○ K8s will alter the iptables accordingly on all the Nodes ● The Service Discovery is implemented either by injecting variables or using DNS ○ E.g. ELASTIC_SERVICE_HOST / ELASTIC_SERVICE_PORT get injected into every Pod ○ Or there is ‘elastic’ DNS record, so all the other Pods can rely on that
- 21.
- 22. Volumes / PersistentVolumes - Storage ● Not exactly the same thing as Docker volumes ● Volume - tight to a Pod lifecycle (survives container restarts) ● Persistent Volumes - can be reused by another Pod (of the same type) ● Many adapters supported - local volumes, AWS EBS, GCEPersistentDisk, Ceph, Gluster, ScaleIO and many others ● Persistent Volume Claim - a way to decouple the definition of storage and apps claiming that storage (claim can be reused) ● Storage Class - a way to define properties of the storage (IOPS, zone, etc.) - the claim can be based on these (e.g. demanding a fast volume in us-east)
- 23. Deployments ● Deployments providea way to define a deployment strategy for a Pod ● Its definition is very similar to ReplicaSet, but it adds the strategy ● 2 strategies - Recreate and RollingUpdate ● RollingUpdate ○ maxUnavailable - max. number or percentage of Pods unavailable during update ○ maxSurge - max. number or percentage to create over the replica count ○ You can also define a time period for which the Pod must be healthy before moving to another
- 24.
- 25.
- 26.
- 27. Some of theother Primitives ● Network Policies - defining access in the cluster and to/from outside ● Ingress - inbound http rules, TLS termination, external loadbalancing, etc. ● Jobs - one-time jobs ● StatefulSet - solution to master/slave stores (e.g. MySQL, Redis, etc.) ● Namespaces - means to separate different “parts” of cluster (e.g. by environment)
- 28. Deployment Options ● Local- minikube (used in the demo later) ● Hosted - Google (GKE), RedHat (OpenShift), Microsoft (Azure CS), AWS (EKS) and many others ● VMs - Vagrant, vSphere (Vmware), oVirt ● Bare Metal - CentOS, Fedora, Ubuntu, CoreOS
- 29. Resources ● https://kubernetes.io/ ● https://cloud.google.com/kubernetes-engine/kubernetes-comic/ ●https://speakerdeck.com/luxas/intro-to-the-cloud-native-world-of-kubernetes-helsinki- october-meetup ● http://lmgtfy.com/?q=kubernetes ● http://shop.oreilly.com/product/0636920043874.do
- 30. (well, in casewe have the time...)
- 31. Thanks! Any Questions? Head isgoing to explode? (Don’t worry, mine is too!) See you later! Later Gator!