title
stringlengths 3
46
| content
stringlengths 0
1.6k
|
|---|---|
20:701 | The endpoint check is quite similar: |
20:702 | ... |
20:703 | livenessProbe: |
20:704 | exec: |
20:705 | httpGet: |
20:706 | path: /healthz |
20:707 | port: 8080 |
20:708 | httpHeaders: |
20:709 | - name: Custom-Health-Header |
20:710 | value: container-is-ok |
20:711 | initialDelaySeconds: 10 |
20:712 | periodSeconds: 5 |
20:713 | ... |
20:714 | |
20:715 | The test is successful if the HTTP response contains the declared header with the declared value. You may also use a pure TCP check, as shown here: |
20:716 | ... |
20:717 | livenessProbe: |
20:718 | exec: |
20:719 | tcpSocket: |
20:720 | port: 8080 |
20:721 | initialDelaySeconds: 10 |
20:722 | periodSeconds: 5 |
20:723 | ... |
20:724 | |
20:725 | In this case, the check succeeds if Kubernetes is able to open a TCP socket to the container on the declared port. |
20:726 | Similarly, the readiness of containers once they are installed is monitored with a readiness check. The readiness check is defined in a similar way as the liveness check, the only difference being that livenessProbe is replaced with readinessProbe. |
20:727 | The following subsection explains how to autoscale Deployments. |
20:728 | Autoscaling |
20:729 | Instead of manually modifying the number of replicas in a Deployment to adapt it to a decrease or increase in load, we can let Kubernetes decide for itself the number of replicas needed to keep a declared resource’s consumption constant. Thus, for instance, if we declare a target of 10% CPU consumption, then when the average resource consumption of each replica exceeds this limit, a new replica will be created. If the average CPU consumption falls below this limit, a replica is destroyed. The typical resource used to monitor replicas is CPU consumption, but we can also use memory consumption. |
20:730 | In actual high-traffic production systems, autoscaling is a must, because it is the only way to adapt quickly the system to changes in the load. |
20:731 | Autoscaling is achieved by defining a HorizontalPodAutoscaler object. Here is an example of the HorizontalPodAutoscaler definition: |
20:732 | apiVersion: autoscaling/v2beta1 |
20:733 | kind: HorizontalPodAutoscaler |
20:734 | metadata: |
20:735 | name: my-autoscaler |
20:736 | spec: |
20:737 | scaleTargetRef: |
20:738 | apiVersion: extensions/v1beta1 |
20:739 | kind: Deployment |
20:740 | name: my-deployment-name |
20:741 | minReplicas: 1 |
20:742 | maxReplicas: 10 |
20:743 | metrics: |
20:744 | - type: Resource |
20:745 | resource: |
20:746 | name: cpu |
20:747 | targetAverageUtilization: 25 |
20:748 | |
20:749 | spec-> scaleTargetRef->name specifies the name of the Deployment to autoscale, while targetAverageUtilization specifies the target resource (in our case, cpu) percentage usage (in our case, 25%). |
20:750 | The following subsection gives a short introduction to the Helm package manager and Helm charts and explains how to install Helm charts on a Kubernetes cluster. An example of how to install an Ingress Controller is given as well. |
20:751 | Helm – installing an Ingress Controller |
20:752 | Helm charts are a way to organize the installation of complex Kubernetes applications that contain several .yaml files. A Helm chart is a set of .yaml files organized into folders and subfolders. Here is a typical folder structure of a Helm chart taken from the official documentation: |
20:753 | |
20:754 | Figure 20.9: Folder structure of a Helm chart |
20:755 | The .yaml files specific to the application are placed in the top templates directory, while the charts directory may contain other Helm charts used as helper libraries. The top-level Chart.yaml file contains general information on the package (name and description), together with both the application version and the Helm chart version. The following is a typical example: |
20:756 | apiVersion: v2 |
20:757 | name: myhelmdemo |
20:758 | description: My Helm chart |
20:759 | type: application |
20:760 | version: 1.3.0 |
20:761 | appVersion: 1.2.0 |
20:762 | |
20:763 | Here, type can be either application or library. Only application charts can be deployed, while library charts are utilities for developing other charts. library charts are placed in the charts folder of other Helm charts. |
20:764 | In order to configure each specific application installation, Helm chart .yaml files contain variables that are specified when Helm charts are installed. Moreover, Helm charts also provide a simple templating language that allows some declarations to be included only if some conditions depending on the input variables are satisfied. The top-level values.yaml file declares default values for the input variables, meaning that the developer needs to specify just the few variables for which they require values different from the defaults. We will not describe the Helm chart template language because it would be too extensive, but you can find it in the official Helm documentation referred to in the Further reading section. |
20:765 | Helm charts are usually organized in public or private repositories in a way that is similar to Docker images. There is a Helm client that you can use to download packages from a remote repository and to install charts in Kubernetes clusters. The Helm client is immediately available in Azure Cloud Shell, so you can start using Helm for your Azure Kubernetes cluster without needing to install it. |
20:766 | A remote repository must be added before using its packages, as shown in the following example: |
20:767 | helm repo add <my-repo-local-name> https://charts.helm.sh/stable |
20:768 | |
20:769 | The preceding command makes the packages of a remote repository available and gives a local name to that remote repository. After that, any package from the remote repository can be installed with a command such as the following: |
20:770 | helm install <instance name><my-repo-local-name>/<package name> -n <namespace> |
20:771 | |
20:772 | Here, <namespace> is the namespace in which to install the application. As usual, if it’s not provided, the default namespace is assumed. <instance name> is the name that you give to the installed application. You need this name to get information about the installed application with the following command: |
20:773 | helm status <instance name> |
20:774 | |
20:775 | You can get also information about all applications installed with Helm with the help of the following command: |
20:776 | helm ls |
20:777 | |
20:778 | The application name is also needed to delete the application from the cluster by means of the following command: |
20:779 | helm delete <instance name> |
20:780 | |
20:781 | When we install an application, we may also provide a .yaml file with all the variable values we want to override. We can also specify a specific version of the Helm chart, otherwise the most recent version is used. Here is an example with both the version and values overridden: |
20:782 | helm install <instance name><my-repo-local-name>/<package name> -f values.yaml --version <version> |
20:783 | |
20:784 | Finally, value overrides can also be provided in-line with the --set option, as shown here: |
20:785 | ...--set <variable1>=<value1>,<variable2>=<value2>... |
20:786 | |
20:787 | We can also upgrade an existing installation with the upgrade command, as shown here: |
20:788 | helm upgrade <instance name><my-repo-local-name>/<package name>... |
20:789 | |
20:790 | The upgrade command may be used to specify new value overrides with the –f option or with the --set option, and a new version with --version. |
20:791 | Let’s use Helm to provide an Ingress for the Guestbook demo application. More specifically, we will use Helm to install an Ingress Controller based on Nginx. The detailed procedure to be observed is as follows: |
20:792 | |
20:793 | Add the remote repository: |
20:794 | helm repo add gcharts https://charts.helm.sh/stable |
20:795 | |
20:796 | |
20:797 | Install the Ingress Controller: |
20:798 | helm install ingress gcharts/nginx-ingress |
20:799 | |
20:800 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.