Merge pull request #109 from NearNodeFlash/api-dws-v1alpha4

Api dws v1alpha4

Author: matthew-richerson

go.mod

@@ -5,10 +5,10 @@ go 1.22.0
 toolchain go1.22.5
 
 require (
-	github.com/DataWorkflowServices/dws v0.0.1-0.20250312162549-a682dcdf49b1
+	github.com/DataWorkflowServices/dws v0.0.1-0.20250417160348-262076fecc45
 	github.com/NearNodeFlash/lustre-fs-operator v0.0.1-0.20250219194350-45aad5cbbe17
-	github.com/NearNodeFlash/nnf-ec v0.0.1-0.20250312170301-6f4fd32b0173 // indirect
-	github.com/NearNodeFlash/nnf-sos v0.0.1-0.20250326193852-54696ee478a9
+	github.com/NearNodeFlash/nnf-ec v0.0.1-0.20250416170223-aeb7554e777b // indirect
+	github.com/NearNodeFlash/nnf-sos v0.0.1-0.20250417190054-045b8b495e19
 	github.com/onsi/ginkgo/v2 v2.22.2
 	github.com/onsi/gomega v1.36.2
 	go.openly.dev/pointy v1.3.0
@@ -43,8 +43,6 @@ require (
 	github.com/imdario/mergo v0.3.16 // indirect
 	github.com/josharian/intern v1.0.0 // indirect
 	github.com/json-iterator/go v1.1.12 // indirect
-	github.com/kubeflow/common v0.4.6 // indirect
-	github.com/kubeflow/mpi-operator v0.3.1-0.20230228224311-5946ef415759 // indirect
 	github.com/mailru/easyjson v0.7.7 // indirect
 	github.com/matttproud/golang_protobuf_extensions v1.0.4 // indirect
 	github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect

go.sum

@@ -1,11 +1,11 @@
-github.com/DataWorkflowServices/dws v0.0.1-0.20250312162549-a682dcdf49b1 h1:HrW1btnw3MKoT8yjyx/eyLvuESuF6MWcPG7CUnuvzIQ=
-github.com/DataWorkflowServices/dws v0.0.1-0.20250312162549-a682dcdf49b1/go.mod h1:i9v4K2d64a9uyd9ZazOJ85RWzVQjIgHWXkLo895KDAE=
+github.com/DataWorkflowServices/dws v0.0.1-0.20250417160348-262076fecc45 h1:VwLXgEJESKSxlez4rvL7aZoF+pVDp+Gj5GO3YtaYZp8=
+github.com/DataWorkflowServices/dws v0.0.1-0.20250417160348-262076fecc45/go.mod h1:i9v4K2d64a9uyd9ZazOJ85RWzVQjIgHWXkLo895KDAE=
 github.com/NearNodeFlash/lustre-fs-operator v0.0.1-0.20250219194350-45aad5cbbe17 h1:xYLLW3fMEKauujqkmspeWoyWhEEu+capOKUPePo4IuU=
 github.com/NearNodeFlash/lustre-fs-operator v0.0.1-0.20250219194350-45aad5cbbe17/go.mod h1:H9GBWOmFp9BGGBWDVZk0pvPWzL6jUKkieVsIIoPlzMs=
-github.com/NearNodeFlash/nnf-ec v0.0.1-0.20250312170301-6f4fd32b0173 h1:UzkNMXM2d63MztEkjLwfBxKHFz+mXS6GzGU/expEzfk=
-github.com/NearNodeFlash/nnf-ec v0.0.1-0.20250312170301-6f4fd32b0173/go.mod h1:lx13ustzE/+39fLECky+CFKkAV8GYlX9eaI6IGmHQkY=
-github.com/NearNodeFlash/nnf-sos v0.0.1-0.20250326193852-54696ee478a9 h1:eh/oIl7i+5MWm46lKFoi+v/4XtNTujUttsDxNIvojhM=
-github.com/NearNodeFlash/nnf-sos v0.0.1-0.20250326193852-54696ee478a9/go.mod h1:l+Ob6Odb8fUI4Lck1orRzIgvdDJ+YXGU96+iio9hkhI=
+github.com/NearNodeFlash/nnf-ec v0.0.1-0.20250416170223-aeb7554e777b h1:J1h0RWYCgs90lpvtfo4nzoZqMtSmOqiQpLhN+QFGH+Y=
+github.com/NearNodeFlash/nnf-ec v0.0.1-0.20250416170223-aeb7554e777b/go.mod h1:lx13ustzE/+39fLECky+CFKkAV8GYlX9eaI6IGmHQkY=
+github.com/NearNodeFlash/nnf-sos v0.0.1-0.20250417190054-045b8b495e19 h1:XrH3VpsjAyLQtFq89LbQsUb47GR7OxtHyH298ObMYyI=
+github.com/NearNodeFlash/nnf-sos v0.0.1-0.20250417190054-045b8b495e19/go.mod h1:l3Rt9mPGNqIh2S0qUOFgMCc9HEGJX6Ed5G55CB08qbg=
 github.com/benbjohnson/clock v1.1.0/go.mod h1:J11/hYXuz8f4ySSvYwY0FKfm+ezbsZBKZxNJlLklBHA=
 github.com/benbjohnson/clock v1.3.0 h1:ip6w0uFQkncKQ979AypyG0ER7mqUSBdKLOgAle/AT8A=
 github.com/benbjohnson/clock v1.3.0/go.mod h1:J11/hYXuz8f4ySSvYwY0FKfm+ezbsZBKZxNJlLklBHA=
@@ -77,10 +77,6 @@ github.com/kr/pty v1.1.1/go.mod h1:pFQYn66WHrOpPYNljwOMqo10TkYh1fy3cYio2l3bCsQ=
 github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
 github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
-github.com/kubeflow/common v0.4.6 h1:yzJf/HEdS6ginD0GlVkgbOFie0Sp66VdGjXidAGZIlk=
-github.com/kubeflow/common v0.4.6/go.mod h1:43MAof/uhpJA2C0urynqatE3oKFQc7m2HLmJty7waqY=
-github.com/kubeflow/mpi-operator v0.3.1-0.20230228224311-5946ef415759 h1:HrgG/sZ2OIbgFciVgKMzU+f4mLBa5Po5evbk9jYNoQY=
-github.com/kubeflow/mpi-operator v0.3.1-0.20230228224311-5946ef415759/go.mod h1:TRq/EvvzlWkzvz6XJs1BNXpUokLNyvHE384E0GSeS5o=
 github.com/mailru/easyjson v0.7.7 h1:UGYAvKxe3sBsEDzO8ZeWOSlIQfWFlxbzLZe7hwFURr0=
 github.com/mailru/easyjson v0.7.7/go.mod h1:xzfreul335JAWq5oZzymOObrkdz5UnU4kGfJJLY9Nlc=
 github.com/matttproud/golang_protobuf_extensions v1.0.4 h1:mmDVorXM7PCGKw94cs5zkfA9PSy5pEvNWRP0ET0TIVo=

int_test.go

@@ -28,7 +28,7 @@ import (
 	. "github.com/onsi/gomega"
 	"go.openly.dev/pointy"
 
-	dwsv1alpha3 "github.com/DataWorkflowServices/dws/api/v1alpha3"
+	dwsv1alpha4 "github.com/DataWorkflowServices/dws/api/v1alpha4"
 )
 
 var (
@@ -41,9 +41,9 @@ var (
 	highTimeout = "5m"
 
 	// Which states use the high timeout
-	highTimeoutStates = []dwsv1alpha3.WorkflowState{
-		dwsv1alpha3.StateSetup,
-		dwsv1alpha3.StateTeardown,
+	highTimeoutStates = []dwsv1alpha4.WorkflowState{
+		dwsv1alpha4.StateSetup,
+		dwsv1alpha4.StateTeardown,
 	}
 )
 
@@ -143,17 +143,17 @@ var tests = []*T{
 		"#DW container name=prerun-timeout-mpi profile=example-mpi-prerun-timeout").
 		WithPermissions(1050, 1051).WithLabels("mpi", "timeout").
 		WithContainerProfile("example-mpi", &ContainerProfileOptions{PrerunTimeoutSeconds: pointy.Int(1), NoStorage: true}).
-		ExpectError(dwsv1alpha3.StatePreRun),
+		ExpectError(dwsv1alpha4.StatePreRun),
 	MakeTest("PostRun timeout on MPI containers",
 		"#DW container name=postrun-timeout-mpi profile=example-mpi-postrun-timeout").
 		WithPermissions(1050, 1051).WithLabels("mpi", "timeout").
 		WithContainerProfile("example-mpi-webserver", &ContainerProfileOptions{PostrunTimeoutSeconds: pointy.Int(1), NoStorage: true}).
-		ExpectError(dwsv1alpha3.StatePostRun),
+		ExpectError(dwsv1alpha4.StatePostRun),
 	MakeTest("Non-zero exit on MPI containers",
 		"#DW container name=mpi-container-fail profile=example-mpi-fail-noretry").
 		WithPermissions(1050, 1051).WithLabels("mpi", "fail").
 		WithContainerProfile("example-mpi-fail", &ContainerProfileOptions{RetryLimit: pointy.Int(0)}).
-		ExpectError(dwsv1alpha3.StatePostRun),
+		ExpectError(dwsv1alpha4.StatePostRun),
 
 	// Containers - Non-MPI
 	MakeTest("GFS2 with Containers",
@@ -175,27 +175,27 @@ var tests = []*T{
 		"#DW container name=prerun-timeout profile=example-prerun-timeout").
 		WithPermissions(1050, 1051).WithLabels("non-mpi", "timeout").
 		WithContainerProfile("example-forever", &ContainerProfileOptions{PrerunTimeoutSeconds: pointy.Int(1), NoStorage: true}).
-		ExpectError(dwsv1alpha3.StatePreRun),
+		ExpectError(dwsv1alpha4.StatePreRun),
 	MakeTest("PostRun timeout on non-MPI containers",
 		"#DW container name=postrun-timeout profile=example-postrun-timeout").
 		WithPermissions(1050, 1051).WithLabels("non-mpi", "timeout").
 		WithContainerProfile("example-forever", &ContainerProfileOptions{PostrunTimeoutSeconds: pointy.Int(1), NoStorage: true}).
-		ExpectError(dwsv1alpha3.StatePostRun),
+		ExpectError(dwsv1alpha4.StatePostRun),
 	MakeTest("Non-zero exit on non-MPI containers",
 		"#DW container name=container-fail profile=example-fail-noretry").
 		WithPermissions(1050, 1051).WithLabels("non-mpi", "fail").
 		WithContainerProfile("example-fail", &ContainerProfileOptions{RetryLimit: pointy.Int(0)}).
-		ExpectError(dwsv1alpha3.StatePostRun),
+		ExpectError(dwsv1alpha4.StatePostRun),
 
 	// Containers - Unsupported Filesystems. These should fail as xfs/raw filesystems are not supported for containers.
 	MakeTest("XFS with Containers",
 		"#DW jobdw type=xfs name=xfs-with-containers capacity=100GB",
 		"#DW container name=xfs-with-containers profile=example-success DW_JOB_foo_local_storage=xfs-with-containers").
-		ExpectError(dwsv1alpha3.StateProposal).WithLabels("unsupported-fs"),
+		ExpectError(dwsv1alpha4.StateProposal).WithLabels("unsupported-fs"),
 	MakeTest("Raw with Containers",
 		"#DW jobdw type=raw name=raw-with-containers capacity=100GB",
 		"#DW container name=raw-with-containers profile=example-success DW_JOB_foo_local_storage=raw-with-containers").
-		ExpectError(dwsv1alpha3.StateProposal).WithLabels("unsupported-fs"),
+		ExpectError(dwsv1alpha4.StateProposal).WithLabels("unsupported-fs"),
 
 	// Containers - Multiple Storages
 	MakeTest("GFS2 and Lustre with Containers",
@@ -250,7 +250,7 @@ var _ = Describe("NNF Integration Test", func() {
 						// TODO: Ginkgo's `--fail-fast` option still seems to execute DeferCleanup() calls
 						//       See if this is by design or if we might need to move this to an AfterEach()
 						if !context.SpecReport().Failed() {
-							t.AdvanceStateAndWaitForReady(ctx, k8sClient, workflow, dwsv1alpha3.StateTeardown)
+							t.AdvanceStateAndWaitForReady(ctx, k8sClient, workflow, dwsv1alpha4.StateTeardown)
 
 							Expect(k8sClient.Delete(ctx, workflow)).To(Succeed())
 						}

internal/internal.go

@@ -29,7 +29,7 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/intstr"
 
-	dwsv1alpha3 "github.com/DataWorkflowServices/dws/api/v1alpha3"
+	dwsv1alpha4 "github.com/DataWorkflowServices/dws/api/v1alpha4"
 	"github.com/DataWorkflowServices/dws/utils/dwdparse"
 )
 
@@ -68,7 +68,7 @@ type T struct {
 	decorators []interface{}
 
 	// Workflow defines the DWS Workflow resource that is the target of the test.
-	workflow *dwsv1alpha3.Workflow
+	workflow *dwsv1alpha4.Workflow
 
 	// User Id to use for the workflow
 	userId int
@@ -84,7 +84,7 @@ type T struct {
 	helperPods []*corev1.Pod
 
 	// Compute nodes that were assigned to the test. This is determined at test runtime.
-	computes *dwsv1alpha3.Computes
+	computes *dwsv1alpha4.Computes
 }
 
 func MakeTest(name string, directives ...string) *T {
@@ -118,13 +118,13 @@ func MakeTest(name string, directives ...string) *T {
 		t.directives = []string{}
 	}
 
-	t.workflow = &dwsv1alpha3.Workflow{
+	t.workflow = &dwsv1alpha4.Workflow{
 		ObjectMeta: metav1.ObjectMeta{
 			Name:      t.WorkflowName(),
 			Namespace: corev1.NamespaceDefault,
 		},
-		Spec: dwsv1alpha3.WorkflowSpec{
-			DesiredState: dwsv1alpha3.StateProposal,
+		Spec: dwsv1alpha4.WorkflowSpec{
+			DesiredState: dwsv1alpha4.StateProposal,
 			DWDirectives: t.WorkflowDirectives(),
 			JobID:        intstr.FromInt(GinkgoParallelProcess()),
 			WLMID:        strconv.Itoa(GinkgoParallelProcess()),
@@ -145,7 +145,7 @@ func (t *T) WorkflowDirectives() []string {
 	return t.directives
 }
 
-func (t *T) Workflow() *dwsv1alpha3.Workflow {
+func (t *T) Workflow() *dwsv1alpha4.Workflow {
 	return t.workflow
 }
 

internal/options.go

@@ -35,7 +35,7 @@ import (
 	corev1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
-	dwsv1alpha3 "github.com/DataWorkflowServices/dws/api/v1alpha3"
+	dwsv1alpha4 "github.com/DataWorkflowServices/dws/api/v1alpha4"
 	lusv1alpha1 "github.com/NearNodeFlash/lustre-fs-operator/api/v1alpha1"
 	nnfv1alpha7 "github.com/NearNodeFlash/nnf-sos/api/v1alpha7"
 
@@ -57,7 +57,7 @@ type TOptions struct {
 	hardwareRequired    bool
 	lowTimeout          time.Duration
 	highTimeout         time.Duration
-	highTimeoutStates   []dwsv1alpha3.WorkflowState
+	highTimeoutStates   []dwsv1alpha4.WorkflowState
 	useExternalComputes bool
 }
 
@@ -67,21 +67,21 @@ func (o *TOptions) hasComplexOptions() bool {
 }
 
 type TStopAfter struct {
-	state dwsv1alpha3.WorkflowState
+	state dwsv1alpha4.WorkflowState
 }
 
 // Stop after lets you stop a test after a given state is reached
-func (t *T) StopAfter(state dwsv1alpha3.WorkflowState) *T {
+func (t *T) StopAfter(state dwsv1alpha4.WorkflowState) *T {
 	t.options.stopAfter = &TStopAfter{state: state}
 	return t
 }
 
 type TExpectError struct {
-	state dwsv1alpha3.WorkflowState
+	state dwsv1alpha4.WorkflowState
 }
 
 // Expect an error at the designed state; Proceed to teardown
-func (t *T) ExpectError(state dwsv1alpha3.WorkflowState) *T {
+func (t *T) ExpectError(state dwsv1alpha4.WorkflowState) *T {
 	t.options.expectError = &TExpectError{state: state}
 	t.options.stopAfter = &TStopAfter{state: state}
 	return t.WithLabels("error")