kubeconform: A Go repository from sklirg

Kubeconform is a Kubernetes manifests validation tool. Build it into your CI to validate your Kubernetes configuration!

It is inspired by, contains code from and is designed to stay close to Kubeval, but with the following improvements:

high performance: will validate & download manifests over multiple routines, caching downloaded files in memory
configurable list of remote, or local schemas locations, enabling validating Kubernetes custom resources (CRDs) and offline validation capabilities
uses by default a self-updating fork of the schemas registry maintained by the kubernetes-json-schema project - which guarantees up-to-date schemas for all recent versions of Kubernetes.

A small overview of Kubernetes manifest validation

Kubernetes's API is described using the OpenAPI (formerly swagger) specification, in a file checked into the main Kubernetes repository.

Because of the state of the tooling to perform validation against OpenAPI schemas, projects usually convert the OpenAPI schemas to JSON schemas first. Kubeval relies on instrumenta/OpenApi2JsonSchema to convert Kubernetes' Swagger file and break it down into multiple JSON schemas, stored in github at instrumenta/kubernetes-json-schema and published on kubernetesjsonschema.dev.

Kubeconform relies on a fork of kubernetes-json-schema that is more aggressively kept up-to-date, and contains schemas for all recent versions of Kubernetes.

Limits of Kubeconform validation

Kubeconform, similarly to kubeval, only validates manifests using the OpenAPI specifications. In some cases, the Kubernetes controllers might perform additional validation - so that manifests passing kubeval validation would still error when being deployed. See for example these bugs against kubeval: #253 #256 #257 #259. The validation logic mentioned in these bug reports is not part of Kubernetes' OpenAPI spec, and therefore kubeconform/kubeval will not detect the configuration errors.

Installation

If you are a Homebrew user, you can install by running:

$ brew install kubeconform

You can also download the latest version from the release page.

Usage

$ ./bin/kubeconform -h
Usage: ./bin/kubeconform [OPTION]... [FILE OR FOLDER]...
  -cache string
        cache schemas downloaded via HTTP to this folder
  -cpu-prof string
        debug - log CPU profiling to file
  -exit-on-error
        immediately stop execution when the first error is encountered
  -h    show help information
  -ignore-filename-pattern value
        regular expression specifying paths to ignore (can be specified multiple times)
  -ignore-missing-schemas
        skip files with missing schemas instead of failing
  -insecure-skip-tls-verify
        disable verification of the server's SSL certificate. This will make your HTTPS connections insecure
  -kubernetes-version string
        version of Kubernetes to validate against, e.g.: 1.18.0 (default "master")
  -n int
        number of goroutines to run concurrently (default 4)
  -output string
        output format - json, junit, tap, text (default "text")
  -reject string
        comma-separated list of kinds to reject
  -schema-location value
        override schemas location search path (can be specified multiple times)
  -skip string
        comma-separated list of kinds to ignore
  -strict
        disallow additional properties not in schema
  -summary
        print a summary at the end (ignored for junit output)
  -v	show version information
  -verbose
        print results for all resources (ignored for tap and junit output)

Usage examples

Validating a single, valid file

$ ./bin/kubeconform fixtures/valid.yaml
$ echo $?
0

Validating a single invalid file, setting output to json, and printing a summary

$ ./bin/kubeconform -summary -output json fixtures/invalid.yaml
{
  "resources": [
    {
      "filename": "fixtures/invalid.yaml",
      "kind": "ReplicationController",
      "version": "v1",
      "status": "INVALID",
      "msg": "Additional property templates is not allowed - Invalid type. Expected: [integer,null], given: string"
    }
  ],
  "summary": {
    "valid": 0,
    "invalid": 1,
    "errors": 0,
    "skipped": 0
  }
}
$ echo $?
1

Passing manifests via Stdin

cat fixtures/valid.yaml  | ./bin/kubeconform -summary
Summary: 1 resource found parsing stdin - Valid: 1, Invalid: 0, Errors: 0 Skipped: 0

Validating a folder, increasing the number of parallel workers

$ ./bin/kubeconform -summary -n 16 fixtures
fixtures/crd_schema.yaml - CustomResourceDefinition trainingjobs.sagemaker.aws.amazon.com failed validation: could not find schema for CustomResourceDefinition
fixtures/invalid.yaml - ReplicationController bob is invalid: Invalid type. Expected: [integer,null], given: string
[...]
Summary: 65 resources found in 34 files - Valid: 55, Invalid: 2, Errors: 8 Skipped: 0

Overriding schemas location - CRD and Openshift support

When the -schema-location parameter is not used, or set to "default", kubeconform will default to downloading schemas from https://github.com/yannh/kubernetes-json-schema. Kubeconform however supports passing one, or multiple, schemas locations - HTTP(s) URLs, or local filesystem paths, in which case it will lookup for schema definitions in each of them, in order, stopping as soon as a matching file is found.

If the -schema-location value does not end with '.json', Kubeconform will assume filenames / a file structure identical to that of kubernetesjsonschema.dev or github.com/yannh/kubernetes-json-schema.
if the -schema-location value ends with '.json' - Kubeconform assumes the value is a Go templated string that indicates how to search for JSON schemas.
the -schema-location value of "default" is an alias for https://raw.githubusercontent.com/yannh/kubernetes-json-schema/master/{{ .NormalizedKubernetesVersion }}-standalone{{ .StrictSuffix }}/{{ .ResourceKind }}{{ .KindSuffix }}.json. Both following command lines are equivalent:

$ ./bin/kubeconform fixtures/valid.yaml
$ ./bin/kubeconform -schema-location default fixtures/valid.yaml
$ ./bin/kubeconform -schema-location 'https://raw.githubusercontent.com/yannh/kubernetes-json-schema/master/{{ .NormalizedKubernetesVersion }}-standalone{{ .StrictSuffix }}/{{ .ResourceKind }}{{ .KindSuffix }}.json' fixtures/valid.yaml

To support validating CRDs, we need to convert OpenAPI files to JSON schema, storing the JSON schemas in a local folder - for example schemas. Then we specify this folder as an additional registry to lookup:

# If the resource Kind is not found in kubernetesjsonschema.dev, also lookup in the schemas/ folder for a matching file
$ ./bin/kubeconform -schema-location default -schema-location 'schemas/{{ .ResourceKind }}{{ .KindSuffix }}.json' fixtures/custom-resource.yaml

You can validate Openshift manifests using a custom schema location. Set the OpenShift version to validate against using -kubernetes-version.

bin/kubeconform -kubernetes-version 3.8.0  -schema-location 'https://raw.githubusercontent.com/garethr/openshift-json-schema/master/{{ .NormalizedKubernetesVersion }}-standalone{{ .StrictSuffix }}/{{ .ResourceKind }}.json'  -summary fixtures/valid.yaml
Summary: 1 resource found in 1 file - Valid: 1, Invalid: 0, Errors: 0 Skipped: 0

Here are the variables you can use in -schema-location:

NormalizedKubernetesVersion - Kubernetes Version, prefixed by v
StrictSuffix - "-strict" or "" depending on whether validation is running in strict mode or not
ResourceKind - Kind of the Kubernetes Resource
ResourceAPIVersion - Version of API used for the resource - "v1" in "apiVersion: monitoring.coreos.com/v1"
KindSuffix - suffix computed from apiVersion - for compatibility with Kubeval schema registries

Converting an OpenAPI file to a JSON Schema

Kubeconform uses JSON schemas to validate Kubernetes resources. For Custom Resource, the CustomResourceDefinition first needs to be converted to JSON Schema. A script is provided to convert these CustomResourceDefinitions to JSON schema. Here is an example how to use it:

$ ./scripts/openapi2jsonschema.py https://raw.githubusercontent.com/aws/amazon-sagemaker-operator-for-k8s/master/config/crd/bases/sagemaker.aws.amazon.com_trainingjobs.yaml
JSON schema written to trainingjob_v1.json

The FILENAME_FORMAT environment variable can be used to change the output file name (Available variables: kind, group, version) (Default: {kind}_{version}).

$ export FILENAME_FORMAT='{kind}-{group}-{version}'
$ ./scripts/openapi2jsonschema.py https://raw.githubusercontent.com/aws/amazon-sagemaker-operator-for-k8s/master/config/crd/bases/sagemaker.aws.amazon.com_trainingjobs.yaml
JSON schema written to trainingjob-sagemaker-v1.json

Some CRD schemas do not have explicit validation for fields implicitly validated by the Kubernetes API like apiVersion, kind, and metadata, thus additional properties are allowed at the root of the JSON schema by default, if this is not desired the DENY_ROOT_ADDITIONAL_PROPERTIES environment variable can be set to any non-empty value.

Usage as a Github Action

Kubeconform publishes Docker Images to Github's new Container Registry, ghcr.io. These images can be used directly in a Github Action, once logged in using a Github Token.

Example:

name: kubeconform
on: push
jobs:
  kubeconform:
    runs-on: ubuntu-latest
    steps:
      - name: login to Github Packages
        run: echo "${{ github.token }}" | docker login https://ghcr.io -u ${GITHUB_ACTOR} --password-stdin
      - uses: actions/checkout@v2
      - uses: docker://ghcr.io/yannh/kubeconform:master
        with:
          entrypoint: '/kubeconform'
          args: "-summary -output json kubeconfigs/"

Note on pricing: Kubeconform relies on Github Container Registry which is currently in Beta. During that period, bandwidth is free. After that period, bandwidth costs might be applicable. Since bandwidth from Github Packages within Github Actions is free, I expect Github Container Registry to also be usable for free within Github Actions in the future. If that were not to be the case, I might publish the Docker image to a different platform.

Proxy support

Kubeconform will respect the HTTPS_PROXY variable when downloading schema files.

$ HTTPS_PROXY=proxy.local bin/kubeconform fixtures/valid.yaml

Speed comparison with Kubeval

Running on a pretty large kubeconfigs setup, on a laptop with 4 cores:

$ time kubeconform -ignore-missing-schemas -n 8 -summary  preview staging production
Summary: 50714 resources found in 35139 files - Valid: 27334, Invalid: 0, Errors: 0 Skipped: 23380

real	0m6,710s
user	0m38,701s
sys	0m1,161s

$ time kubeval -d preview,staging,production --ignore-missing-schemas --quiet
[... Skipping output]

real	0m35,336s
user	0m0,717s
sys	0m1,069s

Using kubeconform as a Go Module

Warning: This is a work-in-progress, the interface is not yet considered stable. Feedback is encouraged.

Kubeconform contains a package that can be used as a library. An example of usage can be found in examples/main.go

Additional documentation on pkg.go.dev

Credits

@garethr for the Kubeval and kubernetes-json-schema projects ❤️

sklirg/kubeconform