Shulou(Shulou.com)05/31 Report--

Today, I will talk to you about how to run the database service in Kubernetes. Many people may not know much about it. In order to make you understand better, the editor has summarized the following for you. I hope you can get something according to this article.

1. Why is it difficult to deploy stateful applications on Kubernetes?

Can we attach a volume to the Pod template? Isn't that enough? In theory, your application can be written to disk now, but if the container is restarted or moved to another node, the volume will be reattached to the new location of the container.

For simple cases, this is true. But for services such as Elasticsearch, etcd, Consul, etc., the situation is much more complicated.

These services have some requirements that cannot be met by a conventional Kubernetes deployment controller.

For example, you might need to provide a predictable DNS name for each Pod to make it easier to form the initial cluster. Or, the system you deploy may need to make sure that Pod will start the container in a predefined order.

In addition, you may want to create and attach separate volumes for each Pod that will be bound to it throughout the life cycle of the Pod. For regular Pod, you can only attach one volume, which will be shared among all Pod created by the same deployment.

We also didn't mention how to manipulate the database. You also need to make sure that you have a plan to determine when and how to perform backups, or how to perform recovery / failover if an error occurs.

two。 Available options for running stateful applications

Here are some options on how to deploy the database on Kubernetes.

(1) StatefulSet

StatefulSet, a built-in controller, is essentially similar to the deployment of Kubernetes, which first appeared in Kubernetes 1.4 and later renamed StatefulSet in 1.5.

Eventually, it will create and manage a set of Pod based on the Pod template you will specify.

The main difference is that it provides the following guarantees for underlying applications:

Each Pod has a stable, unique network identifier.

Each Pod may have a stable, persistent storage volume.

Deployment, expansion, or termination will all be performed in an orderly and elegant manner.

Here are some examples of deployment implementations using StatefulSet's open source database:

Kubernetes Elasticsearch Cluster, author: Paulo Pires

Https://github.com/pires/kubernetes-elasticsearch-cluster/tree/master/stateful

Consul on Kubernetes, author: Kelsey Hightower

Https://github.com/kelseyhightower/consul-on-kubernetes

StatefulSet is generic, so you can use them to model the constrained clustering construction (unique cluster formation) or master / slave architecture of the database.

However, the end result will be lacking in operation. You will need to add additional resources or automation to ensure that you can perform regular backups or add scripts that deal with edge situations such as failover.

In the end, using StatefulSet to model more complex stateful services may feel a bit clumsy, not native to Kubernetes, and, as mentioned above, it will lack management automation. This is where Operator works:

StatefulSet is a load management controller API provided by Kubernetes to manage stateful applications. On the basis of Pod management, ensure the order and consistency of Pod. Unlike Deployment, where StatefulSet uses the container's Spec to create a Pod, Pod created by StatefulSet maintains persistent tags (such as Pod Name) throughout the life cycle. In a nutshell, StatefulSet is a controller that provides a unique flag to Pod, which ensures the order of deployment and expansion.

(2) Operator

If one of the reasons you decide to run data on Kubernetes is to unify the management of all application components, then Operator may provide the experience you want to have!

Instead of putting the application into the StatefulSet model, write (or use someone else's) custom controller.

As a user, this allows you to use Kubectl CLI to control stateful applications as local Kubernetes resources. For example, if you deploy etcd Operator, you can check the backup status of the cluster using the following kubectl command:

Kubectl get EtcdBackup example-etcd-cluster

The main advantage of Operator over StatefulSet is that they add an automation layer that is unique to stateful applications of their operations. You don't have to worry about how to add backup cron to the Elasticsearch cluster, which is implemented using StatefulSet. With Operator, you only need to specify the bucket to store this backup.

Unfortunately, because writing a new Operator requires understanding not only the details of stateful applications, but also Kubernetes and its API, there are not many Operator currently available, and the existing Operator is still relatively new.

Operator is a framework developed by CoreOS to simplify the management of complex stateful applications. It is a state-aware controller that automatically creates, manages and configures application instances by extending Kubernetes API. Operator extends new application resources based on Third Party Resources (CRD) and ensures that the application is in the expected state through the controller. For example, etcd operator simulates the behavior of managing an etcd cluster through the following three steps: observing the current state of the cluster through Kubernetes API; analyzing the difference between the current state and the expected state; and calling etcd cluster management API or Kubernetes API to eliminate these differences.

(3) other

There are few definitions mentioned in this section, mainly to illustrate that for specific databases, such as the PostgreSQL examples we will see later, there are other options for deploying and managing them on Kubernetes as Docker containers.

Sometimes, there are other options available in addition to StatefulSet or a dedicated Operator implementation.

For example, Stolon is a "cloud native PostgreSQL manager for PostgreSQL high availability", and although I haven't had a chance to use it personally, I've seen some posts about Stolon.

To deploy Stolon on Kubernetes, you can use the provided StatefulSet definition. However, due to the capabilities of Stolon, you do not need to add your own cluster management automation to control PostgreSQL clusters. Stolon provides its own CLI for this purpose.

After reading the above, do you have any further understanding of how to run the database service in Kubernetes? If you want to know more knowledge or related content, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.