The core components and terminology k8s

What is Kubernets

  1. Distributed Architecture program container technology.
  2. Open development platform.  
  3. Complete support distributed platform.

Master

  Master refers to the control node cluster, each cluster k8s need to have a Master node responsible for the management and control of the entire cluster, basically k8s all control commands are sent to it, which is responsible for the specific implementation process. In order to achieve high availability, deploy multiple Master, run the following set of key processes on the Master node.

  Kubernets API Server: provides key service process HTTP Rest interface, the process is the only entrance entrance Kubernets in CRUD operations such as all the resources, but also the control of the cluster.

  Kubernets Controller Manager: Kubernets in automation control center for all resource objects, all resource management cluster.

  Kubernets Scheduler: responsible for resource scheduling process.

In addition, the Master node needs to start a etcd service, all Kubernets all the data in the resource object stored in ectd.

 

Node

  In addition to other machines Master, Kubernets cluster is referred to as Node node, Node node may be a virtual machine may be a physical machine. Node is the node in the cluster workload node, when a Node is down, the workload on it will be automatically transferred to other nodes.

  Every Node has the following key processes.

   kubelet: create complex Pod corresponding container, start, stop and other tasks, and collaborate with the Master, implement cluster management functions.

   kube-proxy: load balancing mechanism in communication Kubernets Service.

   Docker Engine: DOcker engine.

Under

  Pod k8s is also the most important basic concept, the root containers containing a special (Pause vessel) each Pod, Pause image corresponding to part of the container k8s platform, in addition to Pause containers, each comprising at least one further Pod a plurality of users or service container. Pod design, (if you want to include something that already exists, you have to come up with a concept has been bigger than something). Pod service container which share a plurality of containers the IP Pause, Pause shared container mounted Volume, Pod divided into two categories, static and Pod Pod general, not stored in the static Pod ectd k8s stored, but stored locally Node file, it can only exist in this Node.

 

Label

  Label k8s system is another concept, a Label is a key-val value pairs, Label can be attached to any resource object. Given a resource object to a Label is equivalent to give it a label, followed by Label Selector queries and filters have some Label resource object.

  1.kube-controller process to filter the number of copies to be monitored by Pod Label Selector defined on the resource object RC.

  2.kube-proxy process to select the corresponding Label Selector Service by the Pod, to automatically establish each Service request forwarding table corresponding to the Pod, thereby achieving load balancing Service.

  3. By the definition of some specific Node Label, and the use of such labels NodeSelector Pod scheduling policy definition file, kube-scheduler Pod oriented scheduling characteristic can be realized.

 

Replication Controller

  RC defines the number of copies Pod at any time in line with a predetermined value, so the RC reads as follows.

  1.Pod the number of copies expected.

  2. Users of screening target Pod Label Selector

  3. When the number of copies Pod is less than the expected number, the template for creating new Pod.

 

Deployment

  Deployment is introduced in order to solve the scheduling problem Pod. Deployment with respect to a maximum upgrade RC is that we can always know the progress of the deployment.

 

Horizontal Pod AutoScaler

  HPA enhanced custom metrics for the application of automatic volume reduction and expansion functions. It is a resource object within k8s, by tracking changes in load analysis of all target Pod RC control, to determine whether the number of copies of the targeted adjustment target Pod.

  1.CPU% utilization

  2. The application custom indicators, such as TPS, QPS.

StatefulSet

  In k8s in, Pod of managed objects RC, Deployment, DaemonSet and Job are for stateless services, but in reality there are many services the state, especially some middleware clusters, such as MySQL, ZK. They each node has its own identity ID, cluster size is fixed.

Service

  k8s inside each Service is actually a micro-service architecture in a micro-service that defines the entry address to access a service, front-end applications to access their group behind a cluster instances of Pod composed by a copy of this entry address, Service and its back-end Pod it is achieved through seamless Label Selector copies between clusters, and the role of RC actually ensure service capacity and quality of service service is always in the expected standard.

IP

  k8s cluster memory of the three IP.

  1.Node IP, IP physical machine.

  2.Pod IP, IP address of the Pod. It is allocated Docker Engine according Docker0 bridge.

  3.Cluster IP: Service IP address. Cluster IP is a virtual IP, but more like a fake IP,

    3.1.Cluster IP merely acting on the object Kubernets Service by Kubernetes manage and assign IP addresses.

    Can not be 3.2.Cluster IP ping, the network entity in response to the absence (Ping protocol is a network layer protocol)

    3.3.Cluster IP Service Port can only be combined to form a specific communications port.

    3.4. Kubernets within the cluster, Node IP network, communications between the Pod IP in Cluster IP uses a special routing rules K8s own design.

Volume

  Volume Chinese name storage volume, which is the Pod in a plurality of containers can be shared directory. Its life cycle and related Pod, k8s support for multiple storage volumes, such as GlusterFS, Ceph and other advanced distributed file system.

Namespace

  Namespace resources used to implement multi-tenant isolation. Namespace by allocating resources inside the object to a different Namespace cluster formed of different items or groups of users logically grouped, while facilitating different groups are respectively managed in the resource sharing using the entire cluster.

Annotation

  Annotation Label and similar, using the key-val form, except that the Label strict naming convention, which defines the metadata K8s object, and additional information is Annotation user arbitrarily defined, in order to find an external tool.

 

ON. Posted 2019-06-18 23:33 Well, Cai Cai is read ( ... ) Comments ( ... ) edit collections

Reproduced in: https: //www.cnblogs.com/shuiyonglewodezzzzz/p/11048622.html

Guess you like

Origin blog.csdn.net/weixin_34408624/article/details/93245655