Table of contents
Kubernetes is designed for automation. Out of the box, you get lots of built-in automation from the core of Kubernetes. You can use Kubernetes to automate deploying and running workloads, and you can automate how Kubernetes does that.
Kubernetes' operator lets you extend the cluster's behavior without modifying the code of Kubernetes itself.
What is Kubernetes Operator?
Let's say you have three copies of an application.
If one of the copies failed then Kubernetes will automatically create a new one. This is the functionality that mostly works in stateless workloads because Kubernetes works best in stateless workloads.
Stateless workload
If you don't know about stateless workloads then the definition is that it does not store the data of the past transaction.
Examples of stateless workloads are web app requests and search engine requests. Every request will be considered as a new one because it won't store the information of the previous request.
Stateful workload
Stateful applications, store the data of the previous transaction. An example is a database. It requires additional backup and recovery knowledge that Kubernetes doesn’t have.
in order to provide this knowledge to scale and upgrade the applications, there are some extensions that will be attached to Kubernetes and they will execute the stateful workload in Kubernetes.
These extensions are called Operators.
Operators allow you to extend the functionality of Kubernetes with custom resources.
Usage
You can run the custom resources with the same command kubectl apply -f filename.yaml .
How do Kubernetes work with stateful applications?
Following are the steps that will be followed in order to make the kubernetes work with stateful applications.
Control Loop Mechanism
Make use of CRDs
Domain/App specific knowledge
Control Loop Mechanism
It continuously watches for an application changes.
If the replica died then it creates a new one.
If the application configuration changes, then it will apply those configurations.
If the image version is updated, it will restart it with another version.
Make use of CRDs
A CRD stands for Custom Defined Resources. CRDs simply let you store and retrieve structured data.
A CRD is an extension of the Kubernetes API. These are some custom resources like Deployment, StatefulSet, and ConfigMap but you can create your own custom components also.
Domain/App specific knowledge
The operator takes the custom resources and applies the control loop mechanism in order to match the desired state with the actual state.
The data of the stateful application that is stored in the CRD will be taken and the domain-specific knowledge will be applied to these resources. As a result, it will automate the entire lifecycle of an application.
Operator Framework
The Operator Framework is an open-source project that enables developers to develop an operator. It does this with the help of an Operator software development kit (Operator SDK).
Who creates these Operators?
Those who are experts in the business logic of installing, running, and updating the application can create these operators.
Let's say a team wants to create a cluster of MySQL replicas and wants to deploy and run them in Kubernetes then that operator will contain this knowledge.
How to create the MySQL cluster?
How to run it?
How to synchronize it?
How update it?
How to take backups?
This information will be combined and packaged and as a result of this information, an operator will be created.
For each application, there is a specific operator.
How many users are expecting operators to perform at a higher level?
If you read the DoK Community 2022 report and see the graph, you will understand that:
4 percent of users are expecting the basic installation of operators to be at a higher level.
13 percent of users are expecting seamless upgrades of operators to be at a higher level.
36 percent of users are expecting the full lifecycle of operators to be at a higher level.
40 percent of users are expecting deep insights of operators to be at a higher level.
7 percent of users are expecting the autopilot of operators to be at a higher level.
it shows that there is still a lack of quality operators that need to be developed. You can check out the OperatorHub to know which operators are created and which one needs to be developed.
That's it for now. Did you like this blog? Please let me know.
You can follow me on YouTube, Twitter, and LinkedIn also.
Happy Learning!