djpbessems/ContainerImage.Pinniped
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
93f51c1a1d
ContainerImage.Pinniped/internal/groupsuffix/groupsuffix_test.go

689 lines
26 KiB
Go
Raw Normal View History

Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
// Copyright 2021-2022 the Pinniped contributors. All Rights Reserved.
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
// SPDX-License-Identifier: Apache-2.0
package groupsuffix
import (
"context"
"fmt"
"strings"
"testing"
"github.com/stretchr/testify/require"
corev1 "k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/runtime/schema"
Use new 'go.pinniped.dev/generated/latest' package. Signed-off-by: Matt Moyer <moyerm@vmware.com>
2021-02-16 19:00:08 +00:00
authv1alpha1 "go.pinniped.dev/generated/latest/apis/concierge/authentication/v1alpha1"
loginv1alpha1 "go.pinniped.dev/generated/latest/apis/concierge/login/v1alpha1"
configv1alpha1 "go.pinniped.dev/generated/latest/apis/supervisor/config/v1alpha1"
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
"go.pinniped.dev/internal/kubeclient"
"go.pinniped.dev/internal/testutil"
)
func ExampleReplace_loginv1alpha1() {
s, _ := Replace(loginv1alpha1.GroupName, "tuna.fish.io")
fmt.Println(s)
// Output: login.concierge.tuna.fish.io
}
func ExampleReplace_string() {
s, _ := Replace("idp.supervisor.pinniped.dev", "marlin.io")
fmt.Println(s)
// Output: idp.supervisor.marlin.io
}
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
func TestMiddlware(t *testing.T) {
const newSuffix = "some.suffix.com"
podWithoutOwner := &corev1.Pod{
TypeMeta: metav1.TypeMeta{
APIVersion: corev1.SchemeGroupVersion.String(),
Kind: "Pod",
},
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{},
},
}
nonPinnipedOwner := &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: "some-name",
UID: "some-uid",
},
}
nonPinnipedOwnerGVK := corev1.SchemeGroupVersion.WithKind("Pod")
podWithNonPinnipedOwner := &corev1.Pod{
TypeMeta: metav1.TypeMeta{
APIVersion: corev1.SchemeGroupVersion.String(),
Kind: "Pod",
},
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
*metav1.NewControllerRef(nonPinnipedOwner, nonPinnipedOwnerGVK),
},
},
}
var ok bool
pinnipedOwner := &configv1alpha1.FederationDomain{
ObjectMeta: metav1.ObjectMeta{
Name: "some-name",
UID: "some-uid",
},
}
pinnipedOwnerGVK := configv1alpha1.SchemeGroupVersion.WithKind("FederationDomain")
pinnipedOwnerWithNewGroupGVK := configv1alpha1.SchemeGroupVersion.WithKind("FederationDomain")
pinnipedOwnerWithNewGroupGVK.Group, ok = Replace(pinnipedOwnerWithNewGroupGVK.Group, newSuffix)
require.True(t, ok)
podWithPinnipedOwner := &corev1.Pod{
TypeMeta: metav1.TypeMeta{
APIVersion: corev1.SchemeGroupVersion.String(),
Kind: "Pod",
},
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
*metav1.NewControllerRef(pinnipedOwner, pinnipedOwnerGVK),
// make sure we don't update the non-pinniped owner
*metav1.NewControllerRef(nonPinnipedOwner, nonPinnipedOwnerGVK),
},
},
}
podWithPinnipedOwnerWithNewGroup := &corev1.Pod{
TypeMeta: metav1.TypeMeta{
APIVersion: corev1.SchemeGroupVersion.String(),
Kind: "Pod",
},
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
*metav1.NewControllerRef(pinnipedOwner, pinnipedOwnerWithNewGroupGVK),
// make sure we don't update the non-pinniped owner
*metav1.NewControllerRef(nonPinnipedOwner, nonPinnipedOwnerGVK),
},
},
}
federationDomainWithPinnipedOwner := &configv1alpha1.FederationDomain{
TypeMeta: metav1.TypeMeta{
APIVersion: configv1alpha1.SchemeGroupVersion.String(),
Kind: "FederationDomain",
},
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
*metav1.NewControllerRef(pinnipedOwner, pinnipedOwnerGVK),
// make sure we don't update the non-pinniped owner
*metav1.NewControllerRef(nonPinnipedOwner, nonPinnipedOwnerGVK),
},
},
}
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
federationDomainWithNewGroupAndPinnipedOwner := &configv1alpha1.FederationDomain{
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
TypeMeta: metav1.TypeMeta{
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
APIVersion: replaceGV(t, configv1alpha1.SchemeGroupVersion, newSuffix).String(),
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
Kind: "FederationDomain",
},
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
*metav1.NewControllerRef(pinnipedOwner, pinnipedOwnerGVK),
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
// make sure we don't update the non-pinniped owner
*metav1.NewControllerRef(nonPinnipedOwner, nonPinnipedOwnerGVK),
},
},
}
federationDomainWithNewGroupAndPinnipedOwnerWithNewGroup := &configv1alpha1.FederationDomain{
TypeMeta: metav1.TypeMeta{
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
APIVersion: replaceGV(t, configv1alpha1.SchemeGroupVersion, newSuffix).String(),
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
Kind: "FederationDomain",
},
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
*metav1.NewControllerRef(pinnipedOwner, pinnipedOwnerWithNewGroupGVK),
// make sure we don't update the non-pinniped owner
*metav1.NewControllerRef(nonPinnipedOwner, nonPinnipedOwnerGVK),
},
},
}
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
tokenCredentialRequest := with(
&loginv1alpha1.TokenCredentialRequest{},
gvk(replaceGV(t, loginv1alpha1.SchemeGroupVersion, newSuffix).WithKind("TokenCredentialRequest")),
)
tokenCredentialRequestWithPinnipedAuthenticator := with(
tokenCredentialRequest,
authenticatorAPIGroup(authv1alpha1.SchemeGroupVersion.Group),
)
tokenCredentialRequestWithCustomAPIGroupAuthenticator := with(
tokenCredentialRequest,
authenticatorAPIGroup(replaceGV(t, authv1alpha1.SchemeGroupVersion, newSuffix).Group),
)
tokenCredentialRequestWithNewGroup := with(
tokenCredentialRequest,
gvk(replaceGV(t, loginv1alpha1.SchemeGroupVersion, newSuffix).WithKind("TokenCredentialRequest")),
)
tokenCredentialRequestWithNewGroupAndPinnipedAuthenticator := with(
tokenCredentialRequestWithNewGroup,
authenticatorAPIGroup(authv1alpha1.SchemeGroupVersion.Group),
)
tokenCredentialRequestWithNewGroupAndCustomAPIGroupAuthenticator := with(
tokenCredentialRequestWithNewGroup,
authenticatorAPIGroup(replaceGV(t, authv1alpha1.SchemeGroupVersion, newSuffix).Group),
)
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
tests := []struct {
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
name string
apiGroupSuffix string
rt *testutil.RoundTrip
requestObj, responseObj kubeclient.Object
wantNilMiddleware bool
wantMutateRequests, wantMutateResponses int
wantMutateRequestErrors, wantMutateResponseErrors []string
wantRequestObj, wantResponseObj kubeclient.Object
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
}{
{
name: "api group suffix is empty",
apiGroupSuffix: "",
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbGet).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
wantNilMiddleware: true,
},
{
name: "api group suffix is default",
apiGroupSuffix: "pinniped.dev",
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbGet).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
wantNilMiddleware: true,
},
{
name: "get resource without pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbGet).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
responseObj: podWithoutOwner,
wantMutateResponses: 1,
wantResponseObj: podWithoutOwner,
},
{
name: "get resource without pinniped.dev with non-pinniped.dev owner ref",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbGet).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
responseObj: podWithNonPinnipedOwner,
wantMutateResponses: 1,
wantResponseObj: podWithNonPinnipedOwner,
},
{
name: "get resource without pinniped.dev with pinniped.dev owner ref",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbGet).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
responseObj: podWithPinnipedOwnerWithNewGroup,
wantMutateResponses: 1,
wantResponseObj: podWithPinnipedOwner,
},
{
name: "get resource with pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbGet).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
requestObj: with(&metav1.PartialObjectMetadata{}, gvk(loginv1alpha1.SchemeGroupVersion.WithKind("TokenCredentialRequest"))),
responseObj: tokenCredentialRequestWithNewGroup,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantMutateRequests: 1,
wantMutateResponses: 1,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantRequestObj: with(&metav1.PartialObjectMetadata{}, gvk(replaceGV(t, loginv1alpha1.SchemeGroupVersion, newSuffix).WithKind("TokenCredentialRequest"))),
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
name: "create resource without pinniped.dev and without owner ref",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
requestObj: podWithoutOwner,
responseObj: podWithoutOwner,
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: podWithoutOwner,
wantResponseObj: podWithoutOwner,
},
{
name: "create resource without pinniped.dev and with owner ref that has no pinniped.dev owner",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
requestObj: podWithNonPinnipedOwner,
responseObj: podWithNonPinnipedOwner,
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: podWithNonPinnipedOwner,
wantResponseObj: podWithNonPinnipedOwner,
},
{
name: "create resource without pinniped.dev and with owner ref that has pinniped.dev owner",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
requestObj: podWithPinnipedOwner,
responseObj: podWithPinnipedOwnerWithNewGroup,
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: podWithPinnipedOwnerWithNewGroup,
wantResponseObj: podWithPinnipedOwner,
},
{
name: "create subresource without pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")).
WithSubresource("some-subresource"),
responseObj: podWithPinnipedOwner,
wantMutateResponses: 1,
wantResponseObj: podWithPinnipedOwner,
},
{
Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
// test that multiple of our middleware request mutations play nicely with each other on create
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
name: "create resource with pinniped.dev and with owner ref that has pinniped.dev owner",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithNamespace("some-namespace").
WithResource(configv1alpha1.SchemeGroupVersion.WithResource("federationdomains")),
requestObj: federationDomainWithPinnipedOwner,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
responseObj: federationDomainWithNewGroupAndPinnipedOwnerWithNewGroup,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantMutateRequests: 2,
wantMutateResponses: 1,
wantRequestObj: federationDomainWithNewGroupAndPinnipedOwnerWithNewGroup,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantResponseObj: federationDomainWithNewGroupAndPinnipedOwner, // the middleware will reset object GVK for us
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
// test that multiple of our middleware request mutations play nicely with each other on update
name: "update resource with pinniped.dev and with owner ref that has pinniped.dev owner",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbUpdate).
WithNamespace("some-namespace").
WithResource(configv1alpha1.SchemeGroupVersion.WithResource("federationdomains")),
requestObj: federationDomainWithPinnipedOwner,
responseObj: federationDomainWithNewGroupAndPinnipedOwnerWithNewGroup,
wantMutateRequests: 2,
wantMutateResponses: 1,
wantRequestObj: federationDomainWithNewGroupAndPinnipedOwnerWithNewGroup,
wantResponseObj: federationDomainWithNewGroupAndPinnipedOwner, // the middleware will reset object GVK for us
},
{
name: "update resource without pinniped.dev and without owner ref",
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbUpdate).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
requestObj: podWithoutOwner,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
responseObj: podWithoutOwner,
Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
wantMutateRequests: 1,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantMutateResponses: 1,
Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
wantRequestObj: podWithoutOwner,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantResponseObj: podWithoutOwner,
},
{
Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
name: "update resource without pinniped.dev and with owner ref that has pinniped.dev owner",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbUpdate).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
requestObj: podWithPinnipedOwner,
responseObj: podWithPinnipedOwnerWithNewGroup,
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: podWithPinnipedOwnerWithNewGroup,
wantResponseObj: podWithPinnipedOwner,
},
{
name: "update resource with pinniped.dev and without owner ref",
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbUpdate).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequest,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
responseObj: tokenCredentialRequestWithNewGroup,
Enhance Kube middleware to rewrite API group of ownerRefs on update verb When oidcclientsecretstorage.Set() wants to update the contents of the storage Secret, it also wants to keep the original ownerRef of the storage Secret, so it needs the middleware to rewrite the API group of the ownerRef again during the update (just like it had initially done during the create of the Secret).
2022-09-22 04:30:44 +00:00
wantMutateRequests: 2,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantMutateResponses: 1,
wantRequestObj: tokenCredentialRequestWithNewGroup,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
name: "list resource without pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbList).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
responseObj: podWithoutOwner,
wantMutateResponses: 1,
wantResponseObj: podWithoutOwner,
},
{
name: "list resource with pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbList).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequest,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
responseObj: tokenCredentialRequestWithNewGroup,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: tokenCredentialRequestWithNewGroup,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
name: "watch resource without pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbWatch).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
responseObj: podWithoutOwner,
wantMutateResponses: 1,
wantResponseObj: podWithoutOwner,
},
{
name: "watch resource with pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbList).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequest,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
responseObj: tokenCredentialRequestWithNewGroup,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: tokenCredentialRequestWithNewGroup,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
name: "patch resource without pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbPatch).
WithNamespace("some-namespace").
WithResource(corev1.SchemeGroupVersion.WithResource("pods")),
responseObj: podWithoutOwner,
wantMutateResponses: 1,
wantResponseObj: podWithoutOwner,
},
{
name: "patch resource with pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
WithVerb(kubeclient.VerbPatch).
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequest,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
responseObj: tokenCredentialRequestWithNewGroup,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: tokenCredentialRequestWithNewGroup,
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
},
{
name: "create tokencredentialrequest with pinniped.dev authenticator api group",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequestWithPinnipedAuthenticator,
responseObj: tokenCredentialRequestWithNewGroup, // a token credential response does not contain a spec
wantMutateRequests: 3,
wantMutateResponses: 1,
wantRequestObj: tokenCredentialRequestWithNewGroupAndCustomAPIGroupAuthenticator,
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
},
{
name: "create tokencredentialrequest with custom authenticator api group fails because api group is expected to be pinniped.dev",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequestWithCustomAPIGroupAuthenticator,
responseObj: tokenCredentialRequestWithNewGroup, // a token credential response does not contain a spec
wantMutateRequests: 3,
wantMutateResponses: 1,
Switch to go.uber.org/zap for JSON formatted logging Signed-off-by: Monis Khan <mok@vmware.com>
2022-04-16 02:43:53 +00:00
wantMutateRequestErrors: []string{`cannot replace token credential request / authenticator API group "authentication.concierge.some.suffix.com" with group suffix "some.suffix.com"`},
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
},
{
name: "create tokencredentialrequest with pinniped.dev authenticator api group and subresource",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")).
WithSubresource("some-subresource"),
requestObj: tokenCredentialRequestWithPinnipedAuthenticator,
responseObj: tokenCredentialRequestWithNewGroup, // a token credential response does not contain a spec
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: tokenCredentialRequestWithNewGroupAndPinnipedAuthenticator,
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
},
{
name: "non-create tokencredentialrequest with pinniped.dev authenticator api group",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbList).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequestWithPinnipedAuthenticator,
responseObj: tokenCredentialRequestWithNewGroup, // a token credential response does not contain a spec
wantMutateRequests: 1,
wantMutateResponses: 1,
wantRequestObj: tokenCredentialRequestWithNewGroupAndPinnipedAuthenticator,
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
},
{
name: "create tokencredentialrequest with nil authenticator api group",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: tokenCredentialRequest,
responseObj: tokenCredentialRequestWithNewGroup, // a token credential response does not contain a spec
wantMutateRequests: 3,
wantMutateResponses: 1,
Switch to go.uber.org/zap for JSON formatted logging Signed-off-by: Monis Khan <mok@vmware.com>
2022-04-16 02:43:53 +00:00
wantMutateRequestErrors: []string{`cannot replace token credential request / without authenticator API group`},
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
wantResponseObj: tokenCredentialRequestWithNewGroup, // the middleware will reset object GVK for us
},
{
name: "create tokencredentialrequest with non-*loginv1alpha1.TokenCredentialRequest",
apiGroupSuffix: newSuffix,
rt: (&testutil.RoundTrip{}).
WithVerb(kubeclient.VerbCreate).
WithResource(loginv1alpha1.SchemeGroupVersion.WithResource("tokencredentialrequests")),
requestObj: podWithoutOwner,
responseObj: podWithoutOwner,
wantMutateRequests: 3,
wantMutateResponses: 1,
wantMutateRequestErrors: []string{`cannot cast obj of type *v1.Pod to *loginv1alpha1.TokenCredentialRequest`},
wantResponseObj: podWithoutOwner,
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
}
for _, test := range tests {
test := test
t.Run(test.name, func(t *testing.T) {
m := New(test.apiGroupSuffix)
if test.wantNilMiddleware {
require.Nil(t, m, "wanted nil middleware")
return
}
m.Handle(context.Background(), test.rt)
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
require.Len(t, test.rt.MutateRequests, test.wantMutateRequests, "undesired request mutation count")
require.Len(t, test.rt.MutateResponses, test.wantMutateResponses, "undesired response mutation count")
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
if test.wantMutateRequests != 0 {
require.NotNil(t, test.requestObj, "expected test.requestObj to be set")
objMutated := test.requestObj.DeepCopyObject().(kubeclient.Object)
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
var mutateRequestErrors []string
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
for _, mutateRequest := range test.rt.MutateRequests {
mutateRequest := mutateRequest
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
if err := mutateRequest(objMutated); err != nil {
mutateRequestErrors = append(mutateRequestErrors, err.Error())
}
}
if len(test.wantMutateRequestErrors) > 0 {
require.Equal(t, test.wantMutateRequestErrors, mutateRequestErrors, "mutate request errors did not match")
} else {
require.Equal(t, test.wantRequestObj, objMutated, "request obj did not match")
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
}
}
if test.wantMutateResponses != 0 {
require.NotNil(t, test.responseObj, "expected test.responseObj to be set")
objMutated := test.responseObj.DeepCopyObject().(kubeclient.Object)
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
var mutateResponseErrors []string
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
for _, mutateResponse := range test.rt.MutateResponses {
mutateResponse := mutateResponse
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
if err := mutateResponse(objMutated); err != nil {
mutateResponseErrors = append(mutateResponseErrors, err.Error())
}
}
if len(test.wantMutateRequestErrors) > 0 {
require.Equal(t, test.wantMutateResponseErrors, mutateResponseErrors, "mutate response errors did not match")
} else {
require.Equal(t, test.wantResponseObj, objMutated, "response obj did not match")
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
}
}
})
}
}
func TestReplaceError(t *testing.T) {
Use custom suffix in `Spec.Authenticator.APIGroup` of `TokenCredentialRequest` When the Pinniped server has been installed with the `api_group_suffix` option, for example using `mysuffix.com`, then clients who would like to submit a `TokenCredentialRequest` to the server should set the `Spec.Authenticator.APIGroup` field as `authentication.concierge.mysuffix.com`. This makes more sense from the client's point of view than using the default `authentication.concierge.pinniped.dev` because `authentication.concierge.mysuffix.com` is the name of the API group that they can observe their cluster and `authentication.concierge.pinniped.dev` does not exist as an API group on their cluster. This commit includes both the client and server-side changes to make this work, as well as integration test updates. Co-authored-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com> Co-authored-by: Margo Crawford <margaretc@vmware.com>
2021-02-03 23:49:15 +00:00
s, ok := Replace("bad-suffix-that-doesnt-end-in-pinniped-dot-dev", "shouldnt-matter.com")
require.Equal(t, "", s)
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
require.False(t, ok)
Use custom suffix in `Spec.Authenticator.APIGroup` of `TokenCredentialRequest` When the Pinniped server has been installed with the `api_group_suffix` option, for example using `mysuffix.com`, then clients who would like to submit a `TokenCredentialRequest` to the server should set the `Spec.Authenticator.APIGroup` field as `authentication.concierge.mysuffix.com`. This makes more sense from the client's point of view than using the default `authentication.concierge.pinniped.dev` because `authentication.concierge.mysuffix.com` is the name of the API group that they can observe their cluster and `authentication.concierge.pinniped.dev` does not exist as an API group on their cluster. This commit includes both the client and server-side changes to make this work, as well as integration test updates. Co-authored-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com> Co-authored-by: Margo Crawford <margaretc@vmware.com>
2021-02-03 23:49:15 +00:00
s, ok = Replace("bad-suffix-that-end-in.prefixed-pinniped.dev", "shouldnt-matter.com")
require.Equal(t, "", s)
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
require.False(t, ok)
}
func TestReplaceSuffix(t *testing.T) {
s, ok := Replace("something.pinniped.dev.something-else.pinniped.dev", "tuna.io")
require.Equal(t, "something.pinniped.dev.something-else.tuna.io", s)
require.True(t, ok)
Use custom suffix in `Spec.Authenticator.APIGroup` of `TokenCredentialRequest` When the Pinniped server has been installed with the `api_group_suffix` option, for example using `mysuffix.com`, then clients who would like to submit a `TokenCredentialRequest` to the server should set the `Spec.Authenticator.APIGroup` field as `authentication.concierge.mysuffix.com`. This makes more sense from the client's point of view than using the default `authentication.concierge.pinniped.dev` because `authentication.concierge.mysuffix.com` is the name of the API group that they can observe their cluster and `authentication.concierge.pinniped.dev` does not exist as an API group on their cluster. This commit includes both the client and server-side changes to make this work, as well as integration test updates. Co-authored-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com> Co-authored-by: Margo Crawford <margaretc@vmware.com>
2021-02-03 23:49:15 +00:00
// When the replace wasn't actually needed, it still returns true.
s, ok = Unreplace("something.pinniped.dev", "pinniped.dev")
require.Equal(t, "something.pinniped.dev", s)
require.True(t, ok)
}
func TestUnreplaceSuffix(t *testing.T) {
s, ok := Unreplace("something.pinniped.dev.something-else.tuna.io", "tuna.io")
require.Equal(t, "something.pinniped.dev.something-else.pinniped.dev", s)
require.True(t, ok)
// When the unreplace wasn't actually needed, it still returns true.
s, ok = Unreplace("something.pinniped.dev", "pinniped.dev")
require.Equal(t, "something.pinniped.dev", s)
require.True(t, ok)
// When the unreplace was needed but did not work, return false.
s, ok = Unreplace("something.pinniped.dev.something-else.tuna.io", "salmon.io")
require.Equal(t, "", s)
require.False(t, ok)
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
}
func TestValidate(t *testing.T) {
tests := []struct {
apiGroupSuffix string
wantErrorPrefix string
}{
{
apiGroupSuffix: "happy.suffix.com",
},
{
apiGroupSuffix: "no-dots",
Migrate callers to k8s.io/apimachinery/pkg/util/errors.NewAggregate Signed-off-by: Monis Khan <mok@vmware.com>
2021-02-05 17:56:05 +00:00
wantErrorPrefix: "must contain '.'",
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
apiGroupSuffix: ".starts.with.dot",
Migrate callers to k8s.io/apimachinery/pkg/util/errors.NewAggregate Signed-off-by: Monis Khan <mok@vmware.com>
2021-02-05 17:56:05 +00:00
wantErrorPrefix: "a lowercase RFC 1123 subdomain must consist",
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
apiGroupSuffix: "ends.with.dot.",
Migrate callers to k8s.io/apimachinery/pkg/util/errors.NewAggregate Signed-off-by: Monis Khan <mok@vmware.com>
2021-02-05 17:56:05 +00:00
wantErrorPrefix: "a lowercase RFC 1123 subdomain must consist",
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
{
apiGroupSuffix: ".multiple-issues.because-this-string-is-longer-than-the-253-character-limit-of-a-dns-1123-identifier-" + chars(253),
Migrate callers to k8s.io/apimachinery/pkg/util/errors.NewAggregate Signed-off-by: Monis Khan <mok@vmware.com>
2021-02-05 17:56:05 +00:00
wantErrorPrefix: "[must be no more than 253 characters, a lowercase RFC 1123 subdomain must consist",
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
},
}
for _, test := range tests {
test := test
t.Run(test.apiGroupSuffix, func(t *testing.T) {
err := Validate(test.apiGroupSuffix)
if test.wantErrorPrefix != "" {
require.Error(t, err)
require.Truef(
t,
strings.HasPrefix(err.Error(), test.wantErrorPrefix),
"%q does not start with %q", err.Error(), test.wantErrorPrefix)
} else {
require.NoError(t, err)
}
})
}
}
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
type withFunc func(obj kubeclient.Object)
func with(obj kubeclient.Object, withFuncs ...withFunc) kubeclient.Object {
obj = obj.DeepCopyObject().(kubeclient.Object)
for _, withFunc := range withFuncs {
withFunc(obj)
}
return obj
}
func gvk(gvk schema.GroupVersionKind) withFunc {
return func(obj kubeclient.Object) {
obj.GetObjectKind().SetGroupVersionKind(gvk)
}
}
func authenticatorAPIGroup(apiGroup string) withFunc {
return func(obj kubeclient.Object) {
tokenCredentialRequest := obj.(*loginv1alpha1.TokenCredentialRequest)
tokenCredentialRequest.Spec.Authenticator.APIGroup = &apiGroup
}
}
func replaceGV(t *testing.T, baseGV schema.GroupVersion, apiGroupSuffix string) schema.GroupVersion {
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
t.Helper()
groupName, ok := Replace(baseGV.Group, apiGroupSuffix)
require.True(t, ok, "expected to be able to replace %q's suffix with %q", baseGV.Group, apiGroupSuffix)
internal/groupsuffix: mutate TokenCredentialRequest's Authenticator This is a partial revert of 288d9c999efe. For some reason it didn't occur to me that we could do it this way earlier. Whoops. This also contains a middleware update: mutation funcs can return an error now and short-circuit the rest of the request/response flow. The idea here is that if someone is configuring their kubeclient to use middleware, they are agreeing to a narrow-er client contract by doing so (e.g., their TokenCredentialRequest's must have an Spec.Authenticator.APIGroup set). I also updated some internal/groupsuffix tests to be more realistic. Signed-off-by: Andrew Keesler <akeesler@vmware.com>
2021-02-05 01:02:59 +00:00
return schema.GroupVersion{Group: groupName, Version: baseGV.Version}
Allow multiple Pinnipeds to work on same cluster Yes, this is a huge commit. The middleware allows you to customize the API groups of all of the *.pinniped.dev API groups. Some notes about other small things in this commit: - We removed the internal/client package in favor of pkg/conciergeclient. The two packages do basically the same thing. I don't think we use the former anymore. - We re-enabled cluster-scoped owner assertions in the integration tests. This code was added in internal/ownerref. See a0546942 for when this assertion was removed. - Note: the middlware code is in charge of restoring the GV of a request object, so we should never need to write mutations that do that. - We updated the supervisor secret generation to no longer manually set an owner reference to the deployment since the middleware code now does this. I think we still need some way to make an initial event for the secret generator controller, which involves knowing the namespace and the name of the generated secret, so I still wired the deployment through. We could use a namespace/name tuple here, but I was lazy. Signed-off-by: Andrew Keesler <akeesler@vmware.com> Co-authored-by: Ryan Richard <richardry@vmware.com>
2021-01-13 01:27:41 +00:00
}
func chars(count int) string {
return strings.Repeat("a", count)
}
Reference in New Issue Copy Permalink