Kubernetes - how to orchestrate containers
4 likes1,479 views
The document provides an overview of Kubernetes, an open-source cluster manager for orchestrating containerized applications, detailing its architecture, including master and node components. It explains key concepts such as pods, replication controllers, services, and labels, emphasizing their roles in deployment and management of applications. Additionally, it discusses the requirements for running Kubernetes, including necessary tools and network setups.
![Kubernetes
{
"id": "myPod",
"kind": "Pod",
"apiVersion": "v1beta1",
"desiredState": {
"manifest": {
"version": "v1beta1",
"id": "myPod",
"containers": [{
"name": "myPod",
"image": "myPod-Container",
"ports": [{"containerPort": 80}]
}]
}
},
"labels": {
"name": "myPod",
"environment": "production"
}
}
How a pod definition lookslike](https://image.slidesharecdn.com/kubernetes-how-to-orchastrate-containers-johannes-scheuermann-docker-meetup-03-2015-150320103919-conversion-gate01/85/Kubernetes-how-to-orchestrate-containers-10-320.jpg)
Kubernetes - how to orchestrate containers
- 1. Kubernetes How to orchestrate Containers Johannes Maximilian Scheuermann Karlsruhe, 19.03.2015
- 2. Images: http://upload.wikimedia.org/wikipedia/en/thumb/9/9f/Twitter_bird_logo_2012.svg/1267px- Twitter_bird_logo_2012.svg.png , https://developers.google.com/+/branding-guidelines?hl=de About me ● Computer Science Student - Master @KIT ● Working at inovex GmbH ○ IT Engineering & Operations ○ Docker & Kubernetes ○ New Datacenter Technologies ● IRC @johscheuer A brief introduction @johscheuer +JohannesScheuermann
- 3. image: http://xkcd.com/908 Orchestration ● Who use Docker? ● Who use Docker productiv? ● Who use Docker (productiv) on more than one machine? ● How do you orchestrate/scale you Containers? Some Questions
- 4. Image: https://github.com/GoogleCloudPlatform/kubernetes/blob/master/logo.png Kubernetes ● Greek for pilot or helmsman of a ship ● Open Source cluster manager from Google ● Managing containerized applications across a cluster of nodes ● Kubernetes is: ○ lean ○ portable ○ extensible ○ self-healing ● Has Master and Node (Minion) components ● Easy Service deployments, updates and scalability ● Can run basically on every Linux platform Managing Containers
- 5. Kubernetes How it looks Bare Metal/VM Node Proxy Kubelet Bare Metal/VM Master API Server Controller manager Scheduler ● Kubernetes has 5 components ○ 3 master components ○ 2 node components ● Master and node can run on the same host
- 6. Kubernetes ● Can be physical or a VM ● Has the services which are necessary to run Pods and to be managed by the master ○ Includes Docker, kubelet and a network proxy ● Node status describes current status ○ HostIP, Node Phase, Node Condition ● Node is created by cloud providers or from physical or virtual machines ○ Kubernetes only creates a representation ○ After creation it will check whether the node is valid or not ● Node Controller manages Node objects ○ cluster-wide synchronization (create/delete representation) ○ single node life-cycle managment ○ unable to provision nodes Node (before known as Minion)
- 7. Kubernetes ● Clusters, the compute recources on top of which the containers are built. Kubernetes can run on every Linux-Distro. ● Pods, a colocated group of (Docker) containers with shared volumes. Is the smallest deployable unit which can be created, scheduled and managed. You should use a replication controller to create pods. ● Replication controllers, manages the lifecycle of pods. Ensuring that a specified number of pods are running at any given time by killing or creating pods as needed. ● Services, provides a single, stable name and address for a set of pods. They act as basic load balancers. ● Labels, used to organize and select groups of objects based on key:value pairs. Concept
- 8. Kubernetes ● Logical group/abstraction ● Ephemeral rather than durable ● Shared network and volumes ● Identified by ID or labels ● Should/Can be managed by replication controllers ● Why not just run multiple programs in a single Docker container? ○ Transparency ○ Decoupling Dependencies ○ Ease of use ○ Efficiency Pods (as in a pod of whales or pea pod)
- 9. Kubernetes Pods - how it looks like offers service(s) over same IP Shared pod storage (survives container crash) App 1 (Docker) Container App 2 (Docker) Container App 3 (Docker) Container Communication via localhost
- 10. Kubernetes { "id": "myPod", "kind": "Pod", "apiVersion": "v1beta1", "desiredState": { "manifest": { "version": "v1beta1", "id": "myPod", "containers": [{ "name": "myPod", "image": "myPod-Container", "ports": [{"containerPort": 80}] }] } }, "labels": { "name": "myPod", "environment": "production" } } How a pod definition lookslike
- 11. Kubernetes ● Should be used to control pods ● Creates pods from a template ● Ensures that the desired number of pods are operational ● The deletion of an replication controller does not affect the created pods ● Rolling updates ● Multiple release tracks possible with the labels Replication Controllers
- 12. Kubernetes { "id": "myPodController", "kind": "ReplicationController", "apiVersion": "v1beta1", "desiredState": { "replicas": 1, "replicaSelector": { "name": "myPod", "environment": "production", "track": "stable" }, "podTemplate": { … #Pod Template from before } }, "labels": { "name": "myPod", "environment": "production", "track": "stable" } } How a replication controller definition lookslike
- 13. Kubernetes ● Solves the problem that pods are ephemeral ● Service proxy runs on each node ● Offers an IP and port pair ● Adds a set of environment variables ● This imply an ordering requirement Services
- 14. image: http://www.centurylinklabs.com/what-is-kubernetes-and-how-to-use-it Kubernetes Services - How it works
- 15. Kubernetes { "id": "myPod", "kind": "Service", "apiVersion": "v1beta1", "port": 3000, "containerPort": 80, "selector": { "name": "myPod", "environment": "production" } } How a service definition lookslike
- 16. Kubernetes ● Labels are simple key/value pairs ● Can be attached to objects (like pods or nodes) ● Labels do not provide uniqueness ● Via a label selector a client can select a set of objects ● Let you categorize objects e.g. for complex service deployment ● Don’t underestimate them! Labels (in short)
- 17. Kubernetes ● Any Linux Distribution ● Docker ● etcd (distributed key value store) ● Overlay network (flannel) ● If you are running on your private Data Center you need the Kube-register ○ https://github.com/kelseyhightower/kube-register ● There are many getting started guides to start your Kubernetes cluster ○ https://github. com/GoogleCloudPlatform/kubernetes/tree/master/docs/getting-started- guides Requirements
- 18. Kubernetes ● Kubernetes supports DNS with skyDNS (with some custom logic) ● Monitoring with Heapster and cAdvisor ● fluentd + elasticsearch Cluster addons
- 19. Kubernetes ● You will need an HTTP-Proxy ○ Hipache/Vulcand ● You will need a Provisioning + Config Managment Tool ○ Foreman ○ Puppet/Salt/… ● You will need an overlay Network ○ Kubernetes assign new (public) IP address to each pod ● You should never use image:latest for a container image ● Kubernetes Master is SPOF ● No autoscaler at the moment ● Pod ordering ordering requirement ○ SkyDNS can help Some challenges
- 20. Maybe soon? https://github.com/GoogleCloudPlatform/kubernetes/blob/master/docs/design/persistent-storage.md Kubernetes Always the question about shared Storage
- 21. Kontakt Johannes Maximilian Scheuermann [email protected] inovex GmbH Office Karlsruhe Ludwig-Erhard-Allee 6 76131 Karlsruhe Thank you for listening!