Kubernetes Configuration Management Methods
There are three different modes of Kubernetes configuration management:
- Imperative Commands
- Imperative Configuration files
- Declarative Configuration files
When creating a k8s cluster from scratch, it’s often easiest to create deployments using kubectl commands, e.g.
kubctl create deployment ai_analysis --image=store.madaq.com/ai_analsys:v1.2.1
This is the Imperative Commands option. It lets you quickly focus on creating the infrastructure you need, and sets up a configuration with reasonable defaults. It’s easy, quick and efficient.
However, there is no audit trail, no review process, no version control, and no repeatable script.
The Imperative Configuration file approach allows you to create and manage specification files, and use your typical development/review/release process (e.g. github, pull requests, CI)
kubectl create -f deployments/ai_analysis.yml
In this case, the specification for your deployment is stored in a file. This file can be altered and redeployed:
kubectl replace -f deployments/ai_analysis.yml
You can also edit these files on-the-fly using edit
kubectl edit deployments blog -n blog
This will open k8s’ configuration file directly into an editor, where you can update settings.
The third option is Declarative Configuration. Here, your complete configuration can be specified on your file system. You can apply this configuration, and kubectl will automatically compute the required operations to get the existing, running configuration to match the new specification.
You can preview the changes before you make them:
kubectl diff -f -R configs/
kubectl apply -f -R configs/
This method is good because your repository can act like the single source of truth for your k8s configuration, but it does come at the cost of complexity and difficulty in debugging when things go wrong. See Declarative Management of Kubernetes Objects Using Configuration Files | Kubernetes for more fiddly details.
Retrieving an existing configuration
You might want to rapidly develop a new cluster using imperative commands, but then migrate to a more formal and controlled method later. This is straightforward enough, since you can easily extract your live object configuration and save it to a file.
kubectl get deployment ai_analysis -o yaml > deployments/ai_analysis.yml
This can then be modified (also, the status section of the live object
configuration needs to be removed).
Extracting Everything
If you want migrate everything to a file configuration method, you need to retrieve the existing live object configurations, and save them all to files.
kubectl get all -n namespace
This will get full configurations for all pods, services and deployments for a particular namespace, but not custom resources. See How to List all Resources in a Kubernetes Namespace - Studytonight for more useful details.