[fix] Resolve v2alpha API exceptions (kubeflow#2317)

Resolve v2alpha API exceptions by adding necessary listType validations. Signed-off-by: Varsha Prasad Narsing <[email protected]> Signed-off-by: sailesh duddupudi <[email protected]>
saileshd1402 · Dec 2, 2024 · 94b8414 · 94b8414
1 parent 3f5c458
commit 94b8414
Show file tree

Hide file tree

Showing 10 changed files with 315 additions and 83 deletions.
diff --git a/api.v2/openapi-spec/swagger.json b/api.v2/openapi-spec/swagger.json
@@ -72,31 +72,38 @@
           "items": {
             "type": "string",
             "default": ""
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "command": {
           "description": "Entrypoint commands for the training container.",
           "type": "array",
           "items": {
             "type": "string",
             "default": ""
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "env": {
           "description": "List of environment variables to set in the container. These values will be merged with the TrainingRuntime's environments.",
           "type": "array",
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.EnvVar"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         },
         "envFrom": {
           "description": "List of sources to populate environment variables in the container. These   values will be merged with the TrainingRuntime's environments.",
           "type": "array",
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.EnvFromSource"
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "name": {
           "description": "Name for the container. TrainingRuntime must have this container.",
@@ -109,7 +116,11 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.VolumeMount"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         }
       }
     },
@@ -134,7 +145,11 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.EnvVar"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         },
         "secretRef": {
           "description": "Reference to the secret with credentials to download dataset. Secret must be created in the TrainJob's namespace.",
@@ -156,7 +171,11 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.EnvVar"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         },
         "secretRef": {
           "description": "Reference to the secret with credentials to download model. Secret must be created in the TrainJob's namespace.",
@@ -269,10 +288,6 @@
       "description": "MPIMLPolicySource represents a MPI runtime configuration.",
       "type": "object",
       "properties": {
-        "SSHAuthMountPath": {
-          "description": "Directory where SSH keys are mounted.",
-          "type": "string"
-        },
         "mpiImplementation": {
           "description": "Implementation name for the MPI to create the appropriate hostfile. Defaults to OpenMPI.",
           "type": "string"
@@ -285,6 +300,10 @@
         "runLauncherAsNode": {
           "description": "Whether to run training process on the launcher Job. Defaults to false.",
           "type": "boolean"
+        },
+        "sshAuthMountPath": {
+          "description": "Directory where SSH keys are mounted.",
+          "type": "string"
         }
       }
     },
@@ -312,7 +331,11 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.EnvVar"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         },
         "secretRef": {
           "description": "Reference to the secret with credentials to export model. Secret must be created in the TrainJob's namespace.",
@@ -357,15 +380,23 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/kubeflow.org.v2alpha1.ContainerOverride"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         },
         "initContainers": {
           "description": "Overrides for the init container in the desired job templates.",
           "type": "array",
           "items": {
             "default": {},
             "$ref": "#/definitions/kubeflow.org.v2alpha1.ContainerOverride"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         },
         "nodeSelector": {
           "description": "Override for the node selector to place Pod on the specific mode.",
@@ -385,23 +416,29 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/kubeflow.org.v2alpha1.PodSpecOverrideTargetJob"
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "tolerations": {
           "description": "Override for the Pod's tolerations.",
           "type": "array",
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.Toleration"
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "volumes": {
           "description": "Overrides for the Pod volume configuration.",
           "type": "array",
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.Volume"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         }
       }
     },
@@ -460,7 +497,8 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/k8s.io.api.autoscaling.v2.MetricSpec"
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "minNodes": {
           "description": "Lower limit for the number of nodes to which training job can scale down.",
@@ -583,7 +621,8 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/kubeflow.org.v2alpha1.PodSpecOverride"
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "runtimeRef": {
           "description": "Reference to the training runtime. The field is immutable.",
@@ -624,7 +663,11 @@
           "items": {
             "default": {},
             "$ref": "#/definitions/kubeflow.org.v2alpha1.JobStatus"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         }
       }
     },
@@ -638,23 +681,29 @@
           "items": {
             "type": "string",
             "default": ""
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "command": {
           "description": "Entrypoint commands for the training container.",
           "type": "array",
           "items": {
             "type": "string",
             "default": ""
-          }
+          },
+          "x-kubernetes-list-type": "atomic"
         },
         "env": {
           "description": "List of environment variables to set in the training container. These values will be merged with the TrainingRuntime's trainer environments.",
           "type": "array",
           "items": {
             "default": {},
             "$ref": "#/definitions/v1.EnvVar"
-          }
+          },
+          "x-kubernetes-list-map-keys": [
+            "name"
+          ],
+          "x-kubernetes-list-type": "map"
         },
         "image": {
           "description": "Docker image for the training container.",

diff --git a/hack/violation_exception_v2alpha1.list b/hack/violation_exception_v2alpha1.list
@@ -1,20 +0,0 @@
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,ContainerOverride,Args
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,ContainerOverride,Command
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,ContainerOverride,Env
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,ContainerOverride,EnvFrom
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,ContainerOverride,VolumeMounts
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,DatasetConfig,Env
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,InputModel,Env
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,OutputModel,Env
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,PodSpecOverride,Containers
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,PodSpecOverride,InitContainers
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,PodSpecOverride,TargetJobs
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,PodSpecOverride,Tolerations
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,PodSpecOverride,Volumes
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,TorchElasticPolicy,Metrics
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,TrainJobSpec,PodSpecOverrides
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,TrainJobStatus,JobsStatus
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,Trainer,Args
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,Trainer,Command
-API rule violation: list_type_missing,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,Trainer,Env
-API rule violation: names_match,github.com/kubeflow/training-operator/pkg/apis/kubeflow.org/v2alpha1,MPIMLPolicySource,SSHAuthMountPath

diff --git a/manifests/v2/base/crds/kubeflow.org_clustertrainingruntimes.yaml b/manifests/v2/base/crds/kubeflow.org_clustertrainingruntimes.yaml
@@ -49,9 +49,6 @@ spec:
                   mpi:
                     description: Configuration for the MPI Runtime.
                     properties:
-                      SSHAuthMountPath:
-                        description: Directory where SSH keys are mounted.
-                        type: string
                       mpiImplementation:
                         description: |-
                           Implementation name for the MPI to create the appropriate hostfile.
@@ -68,6 +65,9 @@ spec:
                           Whether to run training process on the launcher Job.
                           Defaults to false.
                         type: boolean
+                      sshAuthMountPath:
+                        description: Directory where SSH keys are mounted.
+                        type: string
                     type: object
                   numNodes:
                     description: |-
@@ -569,6 +569,7 @@ spec:
                               - type
                               type: object
                             type: array
+                            x-kubernetes-list-type: atomic
                           minNodes:
                             description: Lower limit for the number of nodes to which
                               training job can scale down.

diff --git a/manifests/v2/base/crds/kubeflow.org_trainingruntimes.yaml b/manifests/v2/base/crds/kubeflow.org_trainingruntimes.yaml
@@ -49,9 +49,6 @@ spec:
                   mpi:
                     description: Configuration for the MPI Runtime.
                     properties:
-                      SSHAuthMountPath:
-                        description: Directory where SSH keys are mounted.
-                        type: string
                       mpiImplementation:
                         description: |-
                           Implementation name for the MPI to create the appropriate hostfile.
@@ -68,6 +65,9 @@ spec:
                           Whether to run training process on the launcher Job.
                           Defaults to false.
                         type: boolean
+                      sshAuthMountPath:
+                        description: Directory where SSH keys are mounted.
+                        type: string
                     type: object
                   numNodes:
                     description: |-
@@ -569,6 +569,7 @@ spec:
                               - type
                               type: object
                             type: array
+                            x-kubernetes-list-type: atomic
                           minNodes:
                             description: Lower limit for the number of nodes to which
                               training job can scale down.