What is Kustomize?

Kustomize is a configuration management solution that leverages layering to preserve the base settings of your applications and components by overlaying declarative yaml artifacts (called patches) that selectively override default settings without actually changing the original files.

This approach to configuration management is incredibly powerful because most organizations rely on a combination of internally created (which Kustomize supports with bespoke) and common off-the-shelf (which Kustomize supports with COTS) applications to build their products. Overly customizing your source configuration files to satisfy individual use cases not only dramatically minimizes their reusability, it also makes ingesting upgrades either impossible or incredibly painful.

With kustomize, your team can ingest any base file updates for your underlying components while keeping use-case specific customization overrides intact. Another benefit of utilizing patch overlays is that they add dimensionality to your configuration settings, which can be isolated for troubleshooting misconfigurations or layered to create a framework of most-broad to most-specific configuration specifications.

To recap, Kustomize relies on the following system of configuration management layering to achieve reusability:

Base Layer

Specifies the most common resources

Patch Layers

Specifies use case specific resources

Some Real-World Context

Let’s say that you are using a Helm chart from a particular vendor. It’s a close fit for your use case, but not perfect, and requires some customizations. So you fork the Helm chart, make your configuration changes, and apply it to your cluster. A few months later, your vendor releases a new version of the chart you’re using that includes some important features you need. In order to leverage those new features, you have to fork the new Helm chart and re-apply your configuration changes.

At scale, re-forking and re-customizing these Helm charts becomes a large source of overhead with an increased risk of misconfigurations, threatening the stability of your product and services.

The above diagram shows a common use case of a continuous delivery pipeline which starts with a git event. The event may be a push, merge or create a new branch. In this case, Helm is used to generate the yaml files and Kustomize will patch it with environment specific values based on the events. For example: if the branch is master and tied to the production environment, then kustomize will apply the values applicable to production.

Kustomize Features

Kustomize is often used in conjunction with Helm as described above, and it's been embedded in Kubernetes since its March 2019 release of version 1.14 (invoked by the command apply -k).

Kustomize offers the following valuable attributes:

Kubectl Native

No need to install or manage as a separate dependency

Plain Yaml

No complex templating language

Declarative

Purely declarative (just like Kubectl)

Multiple Configurations

Manages any number of different configurations

Before we dive into Kustomize’s features, let’s compare Kustomize to native Helm and native Kubectl to better highlight the differentiated functionality that it offers.

Benefits of Using Kustomize

1. Reusability

Kustomize allows you to reuse one base file across all of your environments (development, staging, production) and then overlay unique specifications for each.

2. Fast Generation

Since Kustomize has no templating language, you can use standard YAML to quickly declare your configurations.

3. Easier to Debug

YAML itself is easy to understand and debug when things go wrong. Pair that with the fact that your configurations are isolated in patches, and you’ll be able to triangulate the root cause of performance issues in no time. Simply compare performance to your base configuration and any other variations that are running.

Kubernetes Example

Let’s step through how Kustomize works using a deployment scenario involving 3 different environments: dev, staging, and production. In this example we’ll use service, deployment, and horizontal pod autoscaler resources. For the dev and staging environments, there won't be any HPA involved. All of the environments will use different types of services:

Dev

ClusterIP

Staging

NodePort

Production

LoadBalancer

They each will have different HPA settings. This is how directory structure looks:

├── base
  │   ├── deployment.yaml
  │   ├── hpa.yaml
  │   ├── kustomization.yaml
  │   └── service.yaml
  └── overlays
      ├── dev
      │   ├── hpa.yaml
      │   └── kustomization.yaml
      ├── production
      │   ├── hpa.yaml
      │   ├── kustomization.yaml
      │   ├── rollout-replica.yaml
      │   └── service-loadbalancer.yaml
      └── staging
          ├── hpa.yaml
          ├── kustomization.yaml
          └── service-nodeport.yaml

1. Review Base Files

The base folder holds the common resources, such as the standard deployment.yaml, service.yaml, and hpa.yaml resource configuration files. We’ll explore each of their contents in the following sections.

base/deployment.yaml

apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: frontend-deployment
  spec:
    selector:
    matchLabels:
      app: frontend-deployment
    template:
    metadata:
      labels:
        app: frontend-deployment
    spec:
      containers:
      - name: app
        image: foo/bar:latest
        ports:
        - name: http
          containerPort: 8080
          protocol: TCP

base/service.yaml

apiVersion: v1
  kind: Service
  metadata:
    name: frontend-service
  spec:
    ports:
    - name: http
      port: 8080
    selector:
    app: frontend-deployment

base/hpa.yaml

apiVersion: autoscaling/v2beta2
  kind: HorizontalPodAutoscaler
  metadata:
    name: frontend-deployment-hpa
  spec:
    scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: frontend-deployment
    minReplicas: 1
    maxReplicas: 5
    metrics:
    - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50

base/kustomization.yaml

The kustmization.yaml file is the most important file in the base folder and it describes what resources you use.

apiVersion: kustomize.config.k8s.io/v1beta1
  kind: Kustomization
  
  resources:
    - service.yaml
    - deployment.yaml
    - hpa.yaml

2. Define Dev Overlay Files

The overlays folder houses environment-specific overlays. It has 3 sub-folders (one for each environment).

dev/kustomization.yaml

This file defines which base configuration to reference and patch using patchesStrategicMerge, which allows partial YAML files to be defined and overlaid on top of the base.

apiVersion: kustomize.config.k8s.io/v1beta1
  kind: Kustomization
  bases:
  - ../../base
  patchesStrategicMerge:
  - hpa.yaml

dev/hpa.yaml

This file has the same resource name as the one located in the base file. This helps in matching the file for patching. This file also contains important values, such as min/max replicas, for the dev environment.

apiVersion: autoscaling/v2beta2
  kind: HorizontalPodAutoscaler
  metadata:
    name: frontend-deployment-hpa
  spec:
    minReplicas: 1
    maxReplicas: 2
    metrics:
    - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 90

If you compare the previous hpa.yaml file with base/hpa.yaml, you’ll notice differences in minReplicas, maxReplicas, and averageUtilization values.

3. Review Patches

To confirm that your patch config file changes are correct before applying to the cluster, you can run kustomize build overlays/dev:

apiVersion: v1
  kind: Service
  metadata:
    name: frontend-service
  spec:
    ports:
    - name: http
      port: 8080
    selector:
      app: frontend-deployment
  ---
  apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: frontend-deployment
  spec:
    selector:
      matchLabels:
        app: frontend-deployment
    template:
      metadata:
        labels:
          app: frontend-deployment
      spec:
        containers:
        - image: foo/bar:latest
          name: app
          ports:
          - containerPort: 8080
            name: http
            protocol: TCP
  ---
  apiVersion: autoscaling/v2beta2
  kind: HorizontalPodAutoscaler
  metadata:
    name: frontend-deployment-hpa
  spec:
    maxReplicas: 2
    metrics:
    - resource:
        name: cpu
        target:
          averageUtilization: 90
          type: Utilization
      type: Resource
    minReplicas: 1
    scaleTargetRef:
      apiVersion: apps/v1
      kind: Deployment

4. Apply Patches

Once you have confirmed that your overlays are correct, use the kubectl apply -k overlays/dev command to apply the the settings to your cluster:

> kubectl apply -k  overlays/dev 
  service/frontend-service created
  deployment.apps/frontend-deployment created
  horizontalpodautoscaler.autoscaling/frontend-deployment-hpa created

After handling the dev environment, we will demo the production environment as in our case it’s superset if staging(in terms of k8s resources).

5. Define Prod Overlay Files

prod/hpa.yaml

In our production hpa.yaml, let’s say we want to allow up to 10 replicas, with new replicas triggered by a resource utilization threshold of 70% avg CPU usage. This is how that would look:

apiVersion: autoscaling/v2beta2
  kind: HorizontalPodAutoscaler
  metadata:
    name: frontend-deployment-hpa
  spec:
    minReplicas: 1
    maxReplicas: 10
    metrics:
    - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 70

prod/rollout-replicas.yaml

There's also a rollout-replicas.yaml file in our production directory which specifies our rolling strategy:

apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: frontend-deployment
  spec:
    replicas: 10
    strategy:
    rollingUpdate:
      maxSurge: 1
      maxUnavailable: 1
    type: RollingUpdate

prod/service-loadbalancer.yaml

We use this file to change the service type to LoadBalancer (whereas in staging/service-nodeport.yaml, it is being patched as NodePort).

apiVersion: v1
  kind: Service
  metadata:
    name: frontend-service
  spec:
    type: LoadBalancer

prod/kustomization.yaml

This file operates the same way in the production folder as it does in your base folder: it defines which base file to reference and which patches to apply for your production environment. In this case, it includes two more files: rollout-replica.yaml and service-loadbalancer.yaml.

apiVersion: kustomize.config.k8s.io/v1beta1
  kind: Kustomization
  bases:
  - ../../base
  patchesStrategicMerge:
  - rollout-replica.yaml
  - hpa.yaml
  - service-loadbalancer.yaml

6. Review Prod Patches

1. Lets see if production values are being applied by running kustomize build overlays/production
2. Once you have reviewed, apply your overlays to the cluster with kubectl apply -k overlays/production

   > kubectl apply -k overlays/production
     service/frontend-service created
     deployment.apps/frontend-deployment created
     horizontalpodautoscaler.autoscaling/frontend-deployment-hpa created

Kustomize Best Practices

1. Keep your custom resources and their instances in separate packages, otherwise you will encounter race conditions and your creation will get stuck. For example, many people keep both the CertManager CRD and CertManager’s resources in the same package, which can cause problems. Most of the time, reapplying the YAML fixes the issue. But it's good practice to keep them separately.
2. Try to keep the common values like namespace, common metadata in the base file.
3. Organize your resources by kind, using the following naming convention: lowercase-hypenated.yaml (e.g., horizontal-pod-autoscaler.yaml). Place services in the service.yaml file.
4. Follow standard directory structure, using bases/ for base files and patches/ or overlays/ for environment-specific files.
5. While developing or before pushing to git, run kubectl kustomize cfg fmt file_name to format the file and set the indentation right.

Install Kustomize

Kustomize comes pre bundled with kubectl version >= 1.14. You can check your version using kubectl version. If version is 1.14 or greater there's no need to take any steps.

For a stand alone Kustomize installation(aka Kustomize cli) , use the following to set it up.

1. Run the following:

   curl -s "https://raw.githubusercontent.com/
     kubernetes-sigs/kustomize/master/hack/install_kustomize.sh"  | bash

2. Move Kustomize to your path, so that it can be accessed system wide:

   sudo mv kustomize /usr/local/bin

3. Open a new terminal and run kustomize -h to verify:

   > kustomize -h
     
     Manages declarative configuration of Kubernetes.
     See https://sigs.k8s.io/kustomize
     
     Usage:
       kustomize [command]
     
     Available Commands:
       build                     Build a kustomization target from a directory or URL.
       cfg                       Commands for reading and writing configuration.
       completion                Generate shell completion script
       create                    Create a new kustomization in the current directory
       edit                      Edits a kustomization file
       fn                        Commands for running functions against configuration.
       help                      Help about any command
       version                   Prints the kustomize version
     
     Flags:
       -h, --help          help for kustomize
           --stack-trace   print a stack-trace on error
     
     Additional help topics:
       kustomize docs-fn                   [Alpha] Documentation for developing and invoking Configuration Functions.
       kustomize docs-fn-spec              [Alpha] Documentation for Configuration Functions Specification.
       kustomize docs-io-annotations       [Alpha] Documentation for annotations used by io.
       kustomize docs-merge                [Alpha] Documentation for merging Resources (2-way merge).
       kustomize docs-merge3               [Alpha] Documentation for merging Resources (3-way merge).
       kustomize tutorials-command-basics  [Alpha] Tutorials for using basic config commands.
       kustomize tutorials-function-basics [Alpha] Tutorials for using functions.
     
     Use "kustomize [command] --help" for more information about a command.

For more installation options, see the Kubectl documentation.