Troubleshooting guide for the validating webhook. (#332)

* Troubleshooting guide for the validating webhook.

Signed-off-by: ichbinblau <theresa.shan@intel.com>

Author: Theresa

microservices-connector/README.md

@@ -151,3 +151,7 @@ make undeploy
 ### Deploy using helm chart
 
 Please refer to the [helm chart README](./helm/README.md) for deploying GMC using helm chart.
+
+### Troubleshooting guide
+
+Please refer to this [troubleshooting_guide](./troubleshooting_guide.md) for identifying GMC Custom Resource issues.

microservices-connector/api/v1alpha3/validating_webhook.go

@@ -34,6 +34,7 @@ var (
 		"TeiReranking",
 		"Tgi",
 		"TgiGaudi",
+		"TgiNvidia",
 		"Llm",
 		"DocSum",
 		"Router",
@@ -110,12 +111,16 @@ func (r *GMConnector) checkfields() field.ErrorList {
 	return allErrs
 }
 
-func checkStepName(s Step, fldRoot *field.Path, nodeName string) *field.Error {
+func checkStepName(s Step, idx int, fldRoot *field.Path, nodeName string) *field.Error {
 	if len(s.StepName) == 0 {
-		return field.Invalid(fldRoot.Child(nodeName).Child("stepName"), s, fmt.Sprintf("the step name for node %v cannot be empty", nodeName))
+		return field.Invalid(fldRoot.Child(nodeName).Child(fmt.Sprintf("steps[%d]", idx)).Child("name"),
+			s,
+			fmt.Sprintf("the step name for node %v cannot be empty", nodeName))
 	}
 	if !slices.Contains(stepNames, s.StepName) {
-		return field.Invalid(fldRoot.Child(nodeName).Child("stepName"), s, fmt.Sprintf("invalid step name: %s for node %v", s.StepName, nodeName))
+		return field.Invalid(fldRoot.Child(nodeName).Child(fmt.Sprintf("steps[%d]", idx)).Child("name"),
+			s,
+			fmt.Sprintf("invalid step name: %s for node %v", s.StepName, nodeName))
 	}
 	return nil
 }
@@ -143,20 +148,22 @@ func validateNames(nodes map[string]Router, fldPath *field.Path) field.ErrorList
 	var errs field.ErrorList
 
 	for name, router := range nodes {
-		for _, step := range router.Steps {
+		for idx, step := range router.Steps {
 			// validate step name
-			if err := checkStepName(step, fldPath, name); err != nil {
+			if err := checkStepName(step, idx, fldPath, name); err != nil {
 				errs = append(errs, err)
 			}
 
 			// check node name has been defined in the spec
 			if !nodeNameExists(step.NodeName, nodeNames) {
-				errs = append(errs, field.Invalid(fldPath.Child(name).Child("nodeName"), step, fmt.Sprintf("node name: %v in step %v does not exist", step.NodeName, step.StepName)))
+				errs = append(errs, field.Invalid(fldPath.Child(name).Child(fmt.Sprintf("steps[%d]", idx)).Child("nodeName"),
+					step,
+					fmt.Sprintf("node name: %v in step %v does not exist", step.NodeName, step.StepName)))
 			}
 
 			// check service name uniqueness
 			if len(step.InternalService.ServiceName) != 0 && slices.Contains(serviceNames, step.InternalService.ServiceName) {
-				errs = append(errs, field.Invalid(fldPath.Child(name).Child("internalService").Child("serviceName"),
+				errs = append(errs, field.Invalid(fldPath.Child(name).Child(fmt.Sprintf("steps[%d]", idx)).Child("internalService").Child("serviceName"),
 					step,
 					fmt.Sprintf("service name: %v in node %v already exists", step.InternalService.ServiceName, name)))
 			} else {

microservices-connector/api/v1alpha3/validating_webhook_test.go

@@ -119,7 +119,7 @@ func Test_checkStepName(t *testing.T) {
 				nodeName: testNode,
 			},
 			want: field.Invalid(
-				field.NewPath("spec").Child("nodes").Child("test-node").Child("stepName"),
+				field.NewPath("spec").Child("nodes").Child(testNode).Child("steps[0]").Child("name"),
 				Step{},
 				fmt.Sprintf("the step name for node %v cannot be empty", testNode),
 			),
@@ -135,7 +135,7 @@ func Test_checkStepName(t *testing.T) {
 				nodeName: testNode,
 			},
 			want: field.Invalid(
-				field.NewPath("spec").Child("nodes").Child("test-node").Child("stepName"),
+				field.NewPath("spec").Child("nodes").Child(testNode).Child("steps[0]").Child("name"),
 				Step{
 					StepName: "invalid",
 					Executor: Executor{},
@@ -158,7 +158,7 @@ func Test_checkStepName(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			if got := checkStepName(tt.args.s, tt.args.fldRoot, tt.args.nodeName); !reflect.DeepEqual(got, tt.want) {
+			if got := checkStepName(tt.args.s, 0, tt.args.fldRoot, tt.args.nodeName); !reflect.DeepEqual(got, tt.want) {
 				t.Errorf("checkStepName() = %v, want %v", got, tt.want)
 			}
 		})
@@ -323,7 +323,7 @@ func Test_validateNames(t *testing.T) {
 				fldPath: field.NewPath("spec").Child("nodes"),
 			},
 			want: append(errs, field.Invalid(
-				field.NewPath("spec").Child("nodes").Child("root").Child("internalService").Child("serviceName"),
+				field.NewPath("spec").Child("nodes").Child("root").Child("steps[1]").Child("internalService").Child("serviceName"),
 				Step{
 					StepName: "Embedding",
 					Executor: Executor{

microservices-connector/troubleshooting_guide.md

@@ -0,0 +1,36 @@
+# Troubleshooting GMC Custom Resource（CR）
+
+This doc is about identifying issues in GMC CR; A validating webhook has been configured to validate CR fields and it will report all detected errors.
+After the CR for GMC pipeline has been deployed, correct the CR if you encounter the following errors:
+
+1. Root node existence
+
+```
+The GMCConnector "chatqa" is invalid: spec.nodes: Invalid value: map[string]v1alpha3.Router{"node1":...}: a root node is required
+```
+
+In the `spec.nodes` section of the CR, a node with the name ‘root’ is required.
+
+2. StepName validation
+
+```
+The GMCConnector "chatqa" is invalid: spec.nodes.root.steps[0].name: Invalid value: v1alpha3.Step{StepName:"Embedding123", Executor:v1alpha3.Executor{NodeName:"", InternalService:v1alpha3.GMCTarget{ServiceName:"embedding-svc", NameSpace:"", Config:map[string]string{"TEI_EMBEDDING_ENDPOINT":"tei-embedding-svc", "endpoint":"/v1/embeddings"}, IsDownstreamService:false}, ExternalService:""}, Data:"", Condition:"", Dependency:"", ServiceURL:""}: invalid step name: Embedding123 for node root
+```
+
+In the CR, the value of StepName in the `spec.nodes.<nodeName>.steps[].name` field should be included in the predefined [list](./api/v1alpha3/validating_webhook.go).
+
+3. nodeName existence
+
+```
+The GMCConnector "switch" is invalid: spec.nodes.root.steps[0].nodeName: Invalid value: v1alpha3.Step{StepName:"Embedding", Executor:v1alpha3.Executor{NodeName:"node123", InternalService:v1alpha3.GMCTarget{ServiceName:"", NameSpace:"", Config:map[string]string(nil), IsDownstreamService:false}, ExternalService:""}, Data:"", Condition:"", Dependency:"", ServiceURL:""}: node name: node123 in step Embedding does not exist
+```
+
+The nodename that is referenced within the `spec.nodes.<nodeName>.steps[].nodeName` field must already be defined in the `spec.nodes` section.
+
+4. serviceName uniqueness
+
+```
+The GMCConnector "chatqa" is invalid: spec.nodes.root.steps[1].internalService.serviceName: Invalid value: v1alpha3.Step{StepName:"TeiEmbedding", Executor:v1alpha3.Executor{NodeName:"", InternalService:v1alpha3.GMCTarget{ServiceName:"tei-embedding-svc", NameSpace:"", Config:map[string]string(nil), IsDownstreamService:true}, ExternalService:""}, Data:"", Condition:"", Dependency:"", ServiceURL:""}: service name: tei-embedding-svc in node root already exists
+```
+
+The serviceName specified in the `spec.nodes.<nodeName>.steps[].internalService.serviceName` field must be unique and not duplicated with service names in other steps.