api

The canonical location of the OpenShift API definition. This repo holds the API type definitions and serialization code used by openshift/client-go APIs in this repo ship inside OCP payloads.

Adding new FeatureGates

Add your FeatureGate to features.go. The threshold for merging a fully disabled or TechPreview FeatureGate is an open enhancement. To promote to Default on any ClusterProfile, the threshold is 99% passing tests on all platforms or QE sign off.

Adding new TechPreview FeatureGate to all ClusterProfiles (Hypershift and SelfManaged)

FeatureGateMyFeatureName = newFeatureGate("MyFeatureName").
			reportProblemsToJiraComponent("my-jira-component").
			contactPerson("my-team-lead").
			productScope(ocpSpecific).
			enableIn(TechPreviewNoUpgrade).
			mustRegister()

Adding new TechPreview FeatureGate to all only Hypershift

This will be enabled in TechPreview on Hypershift, but never enabled on SelfManaged

FeatureGateMyFeatureName = newFeatureGate("MyFeatureName").
			reportProblemsToJiraComponent("my-jira-component").
			contactPerson("my-team-lead").
			productScope(ocpSpecific).
			enableForClusterProfile(Hypershift, TechPreviewNoUpgrade).
			mustRegister()

Promoting to Default, but only on Hypershift

This will be enabled in TechPreview on all ClusterProfiles and also by Default on Hypershift. It will be disabled in Default on SelfManaged.

FeatureGateMyFeatureName = newFeatureGate("MyFeatureName").
			reportProblemsToJiraComponent("my-jira-component").
			contactPerson("my-team-lead").
			productScope([ocpSpecific|kubernetes]).
			enableIn(TechPreviewNoUpgrade).
			enableForClusterProfile(Hypershift, Default).
			mustRegister()

Promoting to Default on all ClusterProfiles

FeatureGateMyFeatureName = newFeatureGate("MyFeatureName").
			reportProblemsToJiraComponent("my-jira-component").
			contactPerson("my-team-lead").
			productScope([ocpSpecific|kubernetes]).
            enableIn(Default, TechPreviewNoUpgrade).
			mustRegister()

defining API validation tests

Tests are logically associated with FeatureGates. When adding any FeatureGated functionality a new test file is required. The test files are located in <group>/<version>/tests/<crd-name>/FeatureGate.yaml:

route/
  v1/
    tests/
      routes.route.openshift.io/
        AAA_ungated.yaml
        RouteExternalCertificate.yaml

Here's an AAA_ungated.yaml example:

apiVersion: apiextensions.k8s.io/v1 # Hack because controller-gen complains if we don't have this.
name: Route
crdName: routes.route.openshift.io
tests:

Here's an RouteExternalCertificate.yaml example:

apiVersion: apiextensions.k8s.io/v1 # Hack because controller-gen complains if we don't have this.
name: Route
crdName: routes.route.openshift.io
featureGate: RouteExternalCertificate
tests:

The integration tests use the crdName and featureGate to determine which tests apply to which manifests and automatically react to changes when the FeatureGates are enabled/disabled on various FeatureSets and ClusterProfiles.

gen-minimal-test.sh can still function to stub out files if you don't want to copy/paste an existing one.

defining FeatureGate e2e tests

In order to move an API into the Default FeatureSet, it is necessary to demonstrate completeness and reliability. E2E tests are the ONLY category of test that automatically prevents regression over time: repository presubmits do NOT provide equivalent protection. To confirm this, there is an automated verify script that runs every time a FeatureGate is added to the Default FeatureSet. The script queries our CI system (sippy/component readiness) to retrieve a list of all automated tests for a given FeatureGate and then enforces the following rules.

1. Tests must contain either [OCPFeatureGate:<FeatureGateName>] or the standard upstream [FeatureGate:<FeatureGateName>].
2. There must be at least five tests for each FeatureGate.
3. Every test must be run on every TechPreview platform we have jobs for. (Ask for an exception if your feature doesn't support a variant.)
4. Every test must run at least 14 times on every platform/variant.
5. Every test must pass at least 95% of the time on every platform/variant.
6. Test results are taken from the last 7 days if the test was run at least 14 times during that period. Otherwise, data from the last 14 days is used.
7. Test flakes (even if the test eventually passes on a retry) are considered failures and negatively impact the pass rate.

If your FeatureGate lacks automated testing, there is an exception process that allows QE to sign off on the promotion by commenting on the PR.

defining new APIs

When defining a new API, please follow the OpenShift API conventions, and then follow the instructions below to regenerate CRDs (if necessary) and submit a pull request with your new API definitions and generated files.

New APIs (new CRDs) must be added first as an unstable API (v1alpha1). Once the feature is more developed, and ready to be promoted to stable, the API can be promoted to v1.

Why do we start with v1alpha1?

By starting an API as a v1alpha1, we can iterate on the API with the ability to make breaking changes. We can make changes to the schema, change validations, change entire types and even serialization without worry.

When changes are made to an API, any existing client code will need to be updated to match. If there are breaking changes (such as changing the serialization), then this requires a new version of the API.

If we did not bump the API version for each breaking change, a client, generated prior to the breaking change, would panic when it tried to deserialize the new serialization of the API.

If, during development of a feature, we need to make a breaking change, we should move the feature to v1alpha2 (or v1alpha3, etc), until we reach a version that we are happy to promote to v1.

Do not make changes to the API when promoting the feature to v1.

Adding a new stable API (v1)

When copying, it matters which // +foo markers are two comments blocks up and which are one comment block up.

// +genclient
// +genclient:nonNamespaced
// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
// the next line of whitespace matters

// MyAPI is amazing, let me describe it!
//
// Compatibility level 1: Stable within a major release for a minimum of 12 months or 3 minor releases (whichever is longer).
// +openshift:compatibility-gen:level=1
// +openshift:file-pattern=cvoRunLevel=0000_50,operatorName=my-operator,operatorOrdering=01
// +kubebuilder:object:root=true
// +kubebuilder:subresource:status
// +kubebuilder:resource:path=myapis,scope=Cluster
// +openshift:api-approved.openshift.io=https://github.com/openshift/api/pull/<this PR number>
// +openshift:capability=IfYouHaveOne
// +kubebuilder:printcolumn:name=Column Name,JSONPath=.status.something,type=string,description=how users should interpret this.
// +kubebuilder:metadata:annotations=key=value
// +kubebuilder:metadata:labels=key=value
// +kubebuilder:validation:XValidation:rule=
type MyAPI struct {
	metav1.TypeMeta `json:",inline"`

	// metadata is the standard object's metadata.
	// More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#metadata
	metav1.ObjectMeta `json:"metadata,omitempty"`

	// spec is the desired state of the cluster version - the operator will work
	// to ensure that the desired version is applied to the cluster.
	// +kubebuilder:validation:Required
	Spec MyAPISpec `json:"spec"`
	// status contains information about the available updates and any in-progress
	// updates.
	// +optional
	Status MyAPIStatus `json:"status"`
}

Adding a new unstable API (v1alpha)

First, add a FeatureGate as described above.

Like above, but there's an additional

// +kubebuilder:validation:XValidation:rule=
// +openshift:enable:FeatureGate=MyFeatureGate
type MyAPI struct {
	...
}

Adding new fields

Here are few other use-cases for convenience, but have a look in ./example for other possibilities.

// +openshift:validation:FeatureGateAwareXValidation:featureGate=MyFeatureGate,rule="has(oldSelf.coolNewField) ? has(self.coolNewField) : true",message="coolNewField may not be removed once set"
type MyAPI struct {
    // +openshift:enable:FeatureGate=MyFeatureGate
    // +optional
    CoolNewField string `json:"coolNewField"`
}

// EvolvingDiscriminator defines the audit policy profile type.
// +openshift:validation:FeatureGateAwareEnum:featureGate="",enum="";StableValue
// +openshift:validation:FeatureGateAwareEnum:featureGate=MyFeatureGate,enum="";StableValue;TechPreviewOnlyValue
type EvolvingDiscriminator string

const (
  // "StableValue" is always present.
  StableValue EvolvingDiscriminator = "StableValue"

  // "TechPreviewOnlyValue" should only be allowed when TechPreviewNoUpgrade is set in the cluster
  TechPreviewOnlyValue EvolvingDiscriminator = "TechPreviewOnlyValue"
)

required labels

In addition to the standard lgtm and approved labels this repository requires either:

bugzilla/valid-bug - applied if your PR references a valid bugzilla bug

OR

qe-approved, docs-approved, and px-approved - these labels can be applied by anyone in the openshift org via the /label command.

Who should apply these qe/docs/px labels?

• For a no-FF team who is merging a feature before code freeze, they need to get those labels applied to their api repo PR by the appropriate teams (i.e. qe, docs, px)
• For a FF(traditional) team who is merging a feature before FF, they can self-apply the labels(via /label commands), they are basically irrelevant for those teams
• For a FF team who is merging a feature after FF, the PR should be rejected barring an exception

Why are these labels needed?

We need a way for no-FF teams to be able to merge post-FF that does not require a BZ. For non-shared repos that mechanism is the qe/docs/px-approved labels. We are expanding that mechanism to shared repos because the alternative would be that no-FF teams would put a dummy bugzilla/valid-bug label on their feature PRs in order to be able to merge them after feature freeze. Since most individuals can't apply a bugzilla/valid-bug label to a PR, this introduces additional obstacles on those PRs. Conversely, anyone can apply the docs/qe/px-approved labels, so "FF" teams that need to apply these labels to merge can do so w/o needing to involve anyone additional.

Does this mean feature-freeze teams can use the no-FF process to merge code?

No, signing a team up to be a no-FF team includes some basic education on the process and includes ensuring the associated QE+Docs participants are aware the team is moving to that model. If you'd like to sign your team up, please speak with Gina Hargan who will be happy to help on-board your team.

vendoring generated manifests into other repositories

If your repository relies on vendoring and copying CRD manifests (good job!), you'll need have an import line that depends on the package that contains the CRD manifests. For example, adding

import (
	_ "github.com/openshift/api/operatoringress/v1/zz_generated.crd-manifests"
)

to any .go file will work, but some commonly chosen files are tools/tools.go or pkg/dependencymagnet/doc.go. Once added, a go mod vendor will pick up the package containing the manifests for you to copy.

generating CRD schemas

Since Kubernetes 1.16, every CRD created in apiextensions.k8s.io/v1 is required to have a structural OpenAPIV3 schema. The schemas provide server-side validation for fields, as well as providing the descriptions for oc explain. Moreover, schemas ensure structural consistency of data in etcd. Without it anything can be stored in a resource which can have security implications. As we host many of our CRDs in this repo along with their corresponding Go types we also require them to have schemas. However, the following instructions apply for CRDs that are not hosted here as well.

These schemas are often very long and complex, and should not be written by hand. For OpenShift, we provide Makefile targets in build-machinery-go which generate the schema, built on upstream's controller-gen tool.

If you make a change to a CRD type in this repo, simply calling make update-codegen-crds should regenerate all CRDs and update the manifests. If yours is not updated, ensure that the path to its API is included in our calls to the Makefile targets, if this doesn't help try calling make generate-with-container for executing the generators in a controlled environment.

To add this generator to another repo:

1. Vendor github.com/openshift/build-machinery-go

2. Update your Makefile to include the following:

include $(addprefix ./vendor/github.com/openshift/build-machinery-go/make/, \
  targets/openshift/crd-schema-gen.mk \
)

$(call add-crd-gen,<TARGET_NAME>,<API_DIRECTORY>,<CRD_MANIFESTS>,<MANIFEST_OUTPUT>)

The parameters for the call are:

1. TARGET_NAME: The name of your generated Make target. This can be anything, as long as it does not conflict with another make target. Recommended to be your api name.
2. API_DIRECTORY: The location of your API. For example if your Go types are located under pkg/apis/myoperator/v1/types.go, this should be ./pkg/apis/myoperator/v1.
3. CRD_MANIFESTS: The directory your CRDs are located in. For example, if that is manifests/my_operator.crd.yaml then it should be ./manifests
4. MANIFEST_OUTPUT: This should most likely be the same as CRD_MANIFESTS, and is only provided for flexibility to output generated code to a different directory.

You can include as many calls to different APIs as necessary, or if you have multiple APIs under the same directory (eg, v1 and v2beta1) you can use 1 call to the parent directory pointing to your API.

After this, calling make update-codegen-crds should generate a new structural OpenAPIV3 schema for your CRDs.

Notes

• This will not generate entire CRDs, only their OpenAPIV3 schemas. If you do not already have a CRD, you will get no output from the generator.
• Ensure that your API is correctly declared for the generator to pick it up. That means, in your doc.go, include the following:
  1. // +groupName=<API_GROUP_NAME>, this should match the group in your CRD spec
  2. // +kubebuilder:validation:Optional, this tells the operator that fields should be optional unless explicitly marked with // +kubebuilder:validation:Required

For more information on the API markers to add to your Go types, see the Kubebuilder book

Order of generation

make update-codegen-crds does roughly this:

1. Run the empty-partial-schema tool. This creates empty CRD manifests in zz_generated.featuregated-crd-manifests for each FeatureGate.
2. Run the schemapatch tool. This fills in the schema for each per-FeatureGate CRD manifest.
3. Run the manifest-merge tool. This combines all the per-FeatureGate CRD manifests and manual-overrides

empty-partial-schema

This tool is gengo based and scans all types for a // +kubebuilder:object:root=true marker. For each type match, the type is navigated and all tags that include a featureGate (// +openshift:enable:FeatureGate, // +openshift:validation:FeatureGateAwareEnum, and // +openshift:validation:FeatureGateAwareXValidation) are tracked. For each type, for each FeatureGate, a file CRD manifest is created in zz_generated.featuregated-crd-manifests. The most common kube-builder tags are re-implemented in this stage to fill in the non-schema portion of the CRD manifests. This includes things like metadata, resource, and some custom openshift tags as well.

The generator ignores the schema when doing verify, so it doesn't fail on needing to run schemapatch. The generator should clean up old FeatureGated manifests when the gate is removed. Ungated files are created for resources that are sometimes ungated. Annotations are injected to indicate which FeatureGate a manifest is for: this is later read by schemapatch and manifest-merge.

schemapatch

This tool is kubebuilder based with patches to handle FeatureGated types, members, and validation. It reads the injected annotation from empty-partial-schema to decide which FeatureGate should be considered enabled when creating the schema that needs to be injected. It has no knowledge of whether the FeatureGate is enabled or disabled in particular ClusterProfile,FeatureSet tuples. It only needs a single pass over all the FeatureGated partial manifests.

If the schema generation isn't doing what you want, manual-override-crd-manifests allows partially overlaying bits of the CRD manifest. yamlpatch is no longer supported. The format is just "write the CRD you want and delete the stuff the generator sets properly". More specifically, it is the partial manifest that server-side-apply (structured merge diff) would properly merge on top of the CRD that is generated otherwise. Caveat, you cannot test this with a kube-apiserver because the CRD schema uses atomic lists and we had to patch that schema to indicate map lists keyed by version.

manifest-merge

This tool is gengo based and it combines the files in zz_generated.featuregated-crd-manifests and manual-override-crd-manifests on a per ClusterProfile,FeatureSet tuple. This tool takes as input all possible ClusterProfiles and all possible FeatureSets. It then maps from ClusterProfile,FeatureSet tuple to the set of enabled and disabled FeatureGates. Then for each CRD,ClusterProfile,Feature tuple, it merges the pertinent input using structured-merge-diff (SSA) logic based on the CRD schema plus a patch to make atomic fields map-lists. Pertinence is determined based on

1. does this manifest have preferred ClusterProfile annotations: if so, honor them; if not, include everywhere.
2. does this manifest have FeatureGate annotations: if so, match against the enabled set for the ClusterProfile,FeatureSet tuple. Note that CustomNoUpgrade selects everything

Once we have CRD for each ClusterProfile,FeatureSet tuple we choose what to serialize. This roughly follows:

1. if all the CRDs are the same, write a single file and annotate with no FeatureSet and every ClusterProfile. Done.
2. if all the CRDs are the same across all ClusterProfiles for each FeatureSet, create one file per FeatureSet and annotate with one FeatureSet and all ClusterProfiles. Done.
3. if all the CRDs are the same across all FeatureSets for one ClusterProfile, create one file and annotate with no FeatureSet and one ClusterProfile. Continue to 4.
4. for all remaining ClusterProfile,FeatureSet tuples, serialize a file with one FeatureSet and one ClusterProfile.