Geekly Articles each Day

Kubernetes 1.16: Highlights Overview



Today, on Wednesday, the next release of Kubernetes will be held - 1.16. According to the tradition that has developed for our blog, for the tenth anniversary time, we are talking about the most significant changes in the new version.

The information used to prepare this material is taken from the Kubernetes enhancements tracking table , CHANGELOG-1.16 and related issues, pull requests, as well as Kubernetes Enhancement Proposals (KEP). So let's go! ..

Knots


A truly large number of notable innovations (in the alpha version status) are presented on the side of the nodes of K8s-clusters (Kubelet).

Firstly, the so-called " ephemeral containers " (Ephemeral Containers) , designed to simplify the process of debugging in pod'ah . The new mechanism allows you to run special containers that start in the namespace of existing pods and live for a short time. Their purpose is to interact with other pods and containers in order to solve any problems and debugging. For this feature, a new kubectl debug command is kubectl debug , similar in essence to kubectl exec : only instead of starting the process in the container (as in the case of exec ) it starts the container in pod. For example, such a command will connect a new container to the pod:

 kubectl debug -c debug-shell --image=debian target-pod -- bash

Details on ephemeral containers (and examples of their use) can be found in the corresponding KEP . The current implementation (in K8s 1.16) is the alpha version, and among the criteria for its transfer to the beta version is “testing the Ephemeral Containers API for at least 2 releases [Kubernetes]”.

NB : In essence and even the name of the feature resembles the already existing kubectl-debug plugin, which we already wrote about . It is assumed that with the advent of ephemeral containers, the development of a separate external plug-in will stop.

Another innovation, PodOverhead is designed to provide a mechanism for calculating overhead costs for pods , which can vary greatly depending on the runtime used. As an example, the authors of this KEP cite Kata Containers, which require the launch of the guest kernel, kata agent, init system, etc. When the overhead becomes so large, it cannot be ignored, which means that a way is needed to take it into account for further quotas, planning, etc. To implement PodSpec , the Overhead *ResourceList field has been added to PodSpec (mapped to data in the RuntimeClass , if one is used).

Another notable innovation is the Node Topology Manager , designed to unify the approach to fine-tuning the distribution of hardware resources for various components in Kubernetes. This initiative is caused by the growing demand of various modern systems (from the field of telecommunications, machine learning, financial services, etc.) for high-performance parallel computing and minimizing delays in the execution of operations, for which they use the advanced capabilities of CPU and hardware acceleration. Such optimizations in Kubernetes have so far been achieved thanks to disparate components (CPU manager, Device manager, CNI), and now they will add a single internal interface that unifies the approach and simplifies the connection of new similar - the so-called topology-aware - components on the Kubelet side. Details are in the corresponding KEP .


Topology Manager Component Diagram

The next feature is checking containers during startup ( startup probe ) . As you know, for containers that run for a long time, it is difficult to get the current status: they are either "killed" before the actual start of operation, or they end up in a deadlock for a long time. A new check (enabled through the feature gate called StartupProbeEnabled ) cancels - or rather, postpones - the action of any other checks until the moment the pod has finished its launch. For this reason, the feature was originally called pod-startup liveness-probe holdoff . For pods that start for a long time, it is possible to conduct a state survey in relatively short time intervals.

In addition, immediately in beta status an improvement for RuntimeClass is added, adding support for “heterogeneous clusters”. With RuntimeClass Scheduling, now it’s not necessary for every node to have support for each RuntimeClass: for pods, you can choose RuntimeClass without thinking about the cluster topology. Previously, to achieve this — in order for pods to appear on nodes with support for everything they needed — they had to assign appropriate rules to NodeSelector and tolerations. KEP talks about usage examples and, of course, implementation details.

Network


Two significant network features that first appeared (in the alpha version) in Kubernetes 1.16 are:


The finalizer presented in the last release called service.kubernetes.io/load-balancer-cleanup and attached to each service with the type LoadBalancer advanced to the beta version. At the time of removal of such a service, it prevents the actual deletion of the resource until the "cleansing" of all the corresponding resources of the balancer is completed.

API Machinery


The real "stabilization milestone" is fixed in the area of ​​the Kubernetes API server and interaction with it. In many respects, this happened due to the transfer to the stable status of CustomResourceDefinitions (CRD) that did not need a special presentation , which had beta status since the distant Kubernetes 1.7 (and this is June 2017!). The same stabilization came to the features related to them:


Another mechanism that has long been familiar to Kubernetes administrators: admission webhook , has also been in beta status for a long time (since K8s 1.9) and has now been declared stable.

Two other features reached beta: server-side apply and watch bookmarks .

And the only significant innovation in the alpha version was the rejection of SelfLink - a special URI that represents the specified object and is part of ObjectMeta and ListMeta (i.e., part of any object in Kubernetes). Why refuse it? The “simple” motivation sounds like the absence of real (insurmountable) reasons for this field to continue to exist. More formal reasons are to optimize performance (removing an unnecessary field) and simplify the work of generic-apiserver, which is forced to process such a field in a special way (this is the only field that is set right before the object is serialized). The real "obsolescence" (in the beta version) of SelfLink will happen to Kubernetes version 1.20, and the final one - 1.21.

Data storage


The main work in the field of storage, as in previous releases, is observed in the field of support for CSI . The main changes here are:


The function for cloning volumes that appeared in the previous version of Kubernetes (using existing PVCs as a DataSource to create new PVCs) has also now received beta status.

Scheduler


Two notable changes in planning (both in the alpha version):


In addition, the opportunity is presented to create your own plug-ins for the scheduler outside the main Kubernetes (out-of-tree) development tree.

Other changes


Also in Kubernetes 1.16 release, one can note an initiative to bring existing metrics in full order , or more precisely, in accordance with official K8s instrumentation requirements. They basically rely on the relevant Prometheus documentation . The inconsistencies were formed for various reasons (for example, some metrics were simply created before the current instructions appeared), and the developers decided that it was time to bring everything to a single standard, "in line with the rest of the Prometheus ecosystem." The current implementation of this initiative has the status of the alpha version, which will gradually increase in subsequent versions of Kubernetes to beta (1.17) and stable (1.18).

In addition, the following changes can be noted:


PS


Read also in our blog:

Source: https://habr.com/ru/post/467477/

More articles:


All Articles