fix: cherry picking appeng fixes for 2.18 (#367)

* fix: added accept header to nim manifest pull requests accepting oci (#362) * fix: added accept header to nim manifest pull requests accepting oci Signed-off-by: Tomer Figenblat <[email protected]> * chore: removed left-over debugging message Signed-off-by: Tomer Figenblat <[email protected]> --------- Signed-off-by: Tomer Figenblat <[email protected]> (cherry picked from commit f7fdf0a) * Fix: Ensure odh-model-controller Deployment Waits for ConfigMap (#361) * initial commit to fix NIM and deployment issues Signed-off-by: Marcus Trujillo <[email protected]> * update to set default on nimState Signed-off-by: Marcus Trujillo <[email protected]> * updated optinal to false Signed-off-by: Marcus Trujillo <[email protected]> --------- Signed-off-by: Marcus Trujillo <[email protected]> (cherry picked from commit 7806732) * fix: nim containers terminated prematurely (#363) Signed-off-by: Tomer Figenblat <[email protected]> Co-authored-by: Mikhail Mikhailitchenko <[email protected]> (cherry picked from commit c92b19f) * chore: added gocyclo no linting comment on the main function Signed-off-by: Tomer Figenblat <[email protected]> --------- Signed-off-by: Tomer Figenblat <[email protected]> Co-authored-by: Marcus Trujillo <[email protected]>
opendatahub-io · Feb 7, 2025 · e705bf9 · e705bf9
1 parent 7307576
commit e705bf9
Show file tree

Hide file tree

Showing 4 changed files with 14 additions and 7 deletions.
diff --git a/cmd/main.go b/cmd/main.go
@@ -21,7 +21,6 @@ import (
 	"crypto/tls"
 	"flag"
 	"os"
-	"slices"
 	"strconv"
 
 	// Import all Kubernetes client auth plugins (e.g. Azure, GCP, OIDC, etc.)
@@ -78,6 +77,7 @@ func getEnvAsBool(name string, defaultValue bool) bool {
 	return defaultValue
 }
 
+// nolint:gocyclo
 func main() {
 	var metricsAddr string
 	var enableLeaderElection bool
@@ -244,7 +244,10 @@ func main() {
 
 	nimState := os.Getenv("NIM_STATE")
 	signalHandlerCtx := ctrl.SetupSignalHandler()
-	if !slices.Contains([]string{"removed", ""}, nimState) {
+	if nimState == "" {
+		nimState = "managed"
+	}
+	if nimState != "removed" {
 		if err = (&nim.AccountReconciler{
 			Client:  mgr.GetClient(),
 			Scheme:  mgr.GetScheme(),

diff --git a/config/manager/manager.yaml b/config/manager/manager.yaml
@@ -105,7 +105,7 @@ spec:
               configMapKeyRef:
                 name: odh-model-controller-parameters
                 key: nim-state
-                optional: true
+                optional: false
         livenessProbe:
           httpGet:
             path: /healthz

diff --git a/internal/controller/testdata/nvidia_api_mock.go b/internal/controller/testdata/nvidia_api_mock.go
@@ -70,8 +70,10 @@ func (r *NimHttpClientMock) Do(req *http.Request) (*http.Response, error) {
 		// check testdata/ngc_catalog_response_page_0.json
 		authHeaderParts := strings.Split(req.Header.Get("Authorization"), " ")
 		if authHeaderParts[0] == "Bearer" && authHeaderParts[1] == "this-is-my-fake-token-please-dont-share-it-with-anyone" {
-			// the token is returned by the nvcr.io/proxy-auth endpoint (stubbed), check testdata/runtime_token_response.json
-			return &http.Response{StatusCode: 200}, nil
+			if req.Header.Get("Accept") == "application/vnd.oci.image.index.v1+json" {
+				// the token is returned by the nvcr.io/proxy-auth endpoint (stubbed), check testdata/runtime_token_response.json
+				return &http.Response{StatusCode: 200}, nil
+			}
 		}
 	}
 

diff --git a/internal/controller/utils/nim.go b/internal/controller/utils/nim.go
@@ -273,6 +273,7 @@ func attemptToPullManifest(runtime NimRuntime, tokenResp *NimTokenResponse) erro
 	}
 
 	req.Header.Add("Authorization", fmt.Sprintf("Bearer %s", tokenResp.Token))
+	req.Header.Add("Accept", "application/vnd.oci.image.index.v1+json")
 
 	resp, respErr := NimHttpClient.Do(req)
 	if respErr != nil {
@@ -331,8 +332,9 @@ func GetNimServingRuntimeTemplate(scheme *runtime.Scheme) (*v1alpha1.ServingRunt
 		Spec: v1alpha1.ServingRuntimeSpec{
 			ServingRuntimePodSpec: v1alpha1.ServingRuntimePodSpec{
 				Annotations: map[string]string{
-					"prometheus.io/path": "/metrics",
-					"prometheus.io/port": "8000",
+					"prometheus.io/path":                    "/metrics",
+					"prometheus.io/port":                    "8000",
+					"serving.knative.dev/progress-deadline": "30m",
 				},
 				Containers: []corev1.Container{
 					{Env: []corev1.EnvVar{