feat: add ingress-agnostic mode for improved compatibility with various ingress controllers

api/v1beta2/sparkapplication_types.go

@@ -307,6 +307,12 @@ type SparkUIConfiguration struct {
 	// TlsHosts is useful If we need to declare SSL certificates to the ingress object
 	// +optional
 	IngressTLS []networkingv1.IngressTLS `json:"ingressTLS,omitempty"`
+	// UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+	// (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+	// This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+	// When false (default), the legacy nginx-specific behavior is used for backward compatibility.
+	// +optional
+	UseIngressAgnosticMode *bool `json:"useIngressAgnosticMode,omitempty"`
 }
 
 // DriverIngressConfiguration is for driver ingress specific configuration parameters.
@@ -333,6 +339,12 @@ type DriverIngressConfiguration struct {
 	// TlsHosts is useful If we need to declare SSL certificates to the ingress object
 	// +optional
 	IngressTLS []networkingv1.IngressTLS `json:"ingressTLS,omitempty"`
+	// UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+	// (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+	// This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+	// When false (default), the legacy nginx-specific behavior is used for backward compatibility.
+	// +optional
+	UseIngressAgnosticMode *bool `json:"useIngressAgnosticMode,omitempty"`
 }
 
 // ApplicationStateType represents the type of the current state of an application.

charts/spark-operator-chart/templates/controller/deployment.yaml

@@ -85,6 +85,9 @@ spec:
         {{- with .Values.controller.uiIngress.annotations }}
         - --ingress-annotations={{ . | toJson }}
         {{- end }}
+        {{- if .Values.controller.uiIngress.useIngressAgnosticMode }}
+        - --use-ingress-agnostic-mode=true
+        {{- end }}
         {{- end }}
         {{- if .Values.controller.batchScheduler.enable }}
         - --enable-batch-scheduler=true

charts/spark-operator-chart/values.yaml

@@ -70,10 +70,10 @@ controller:
 
   # -- Feature gates to enable or disable specific features.
   featureGates:
-  - name: PartialRestart
-    enabled: false
-  - name: LoadSparkDefaults
-    enabled: false
+    - name: PartialRestart
+      enabled: false
+    - name: LoadSparkDefaults
+      enabled: false
 
   # -- The number of old history to retain to allow rollback.
   revisionHistoryLimit: 10
@@ -125,6 +125,11 @@ controller:
     annotations: {}
       # key1: value1
       # key2: value2
+    # -- When enabled, disables nginx-specific ingress path modifications (regex capture groups like `(/|$)(.*)`)
+    # and annotation injection (`nginx.ingress.kubernetes.io/rewrite-target`).
+    # This allows the ingress to work with any ingress controller like Traefik, HAProxy, Istio, OpenShift Routes, etc.
+    # Can be overridden per SparkApplication via `sparkUIOptions.useIngressAgnosticMode` or `driverIngressOptions[].useIngressAgnosticMode`.
+    useIngressAgnosticMode: false
 
   batchScheduler:
     # -- Specifies whether to enable batch scheduler for spark jobs scheduling.
@@ -166,10 +171,10 @@ controller:
 
   # -- Volumes for controller pods.
   volumes:
-  # Create a tmp directory to write Spark artifacts to for deployed Spark apps.
-  - name: tmp
-    emptyDir:
-      sizeLimit: 1Gi
+    # Create a tmp directory to write Spark artifacts to for deployed Spark apps.
+    - name: tmp
+      emptyDir:
+        sizeLimit: 1Gi
 
   # -- Node selector for controller pods.
   nodeSelector: {}
@@ -211,10 +216,10 @@ controller:
 
   # -- Volume mounts for controller containers.
   volumeMounts:
-  # Mount a tmp directory to write Spark artifacts to for deployed Spark apps.
-  - name: tmp
-    mountPath: "/tmp"
-    readOnly: false
+    # Mount a tmp directory to write Spark artifacts to for deployed Spark apps.
+    - name: tmp
+      mountPath: "/tmp"
+      readOnly: false
 
   # -- Pod resource requests and limits for controller containers.
   # Note, that each job submission will spawn a JVM within the controller pods using "/usr/local/openjdk-11/bin/java -Xmx128m".
@@ -236,7 +241,7 @@ controller:
     runAsNonRoot: true
     capabilities:
       drop:
-      - ALL
+        - ALL
     seccompProfile:
       type: RuntimeDefault
 
@@ -341,10 +346,10 @@ webhook:
 
   # -- Volumes for webhook pods.
   volumes:
-  # Create a dir for the webhook to generate its certificates in.
-  - name: serving-certs
-    emptyDir:
-      sizeLimit: 500Mi
+    # Create a dir for the webhook to generate its certificates in.
+    - name: serving-certs
+      emptyDir:
+        sizeLimit: 500Mi
 
   # -- Node selector for webhook pods.
   nodeSelector: {}
@@ -386,12 +391,11 @@ webhook:
 
   # -- Volume mounts for webhook containers.
   volumeMounts:
-  # Mount a dir for the webhook to generate its certificates in.
-  - name: serving-certs
-    mountPath: /etc/k8s-webhook-server/serving-certs
-    subPath: serving-certs
-    readOnly: false
-
+    # Mount a dir for the webhook to generate its certificates in.
+    - name: serving-certs
+      mountPath: /etc/k8s-webhook-server/serving-certs
+      subPath: serving-certs
+      readOnly: false
 
   # -- Pod resource requests and limits for webhook pods.
   resources: {}
@@ -410,7 +414,7 @@ webhook:
     runAsNonRoot: true
     capabilities:
       drop:
-      - ALL
+        - ALL
     seccompProfile:
       type: RuntimeDefault
 
@@ -429,7 +433,7 @@ spark:
   # Namespaces specified here will be watched in addition to those matching jobNamespaceSelector.
   # Make sure the namespaces have already existed.
   jobNamespaces:
-  - default
+    - default
 
   # -- Label selector to filter namespaces to watch.
   # Supports standard Kubernetes label selector syntax (e.g., 'spark-operator=enabled,env in (prod,staging)').

cmd/operator/controller/start.go

@@ -94,11 +94,12 @@ var (
 	defaultBatchScheduler string
 
 	// Spark web UI service and ingress
-	enableUIService    bool
-	ingressClassName   string
-	ingressURLFormat   string
-	ingressTLS         []networkingv1.IngressTLS
-	ingressAnnotations map[string]string
+	enableUIService        bool
+	ingressClassName       string
+	ingressURLFormat       string
+	ingressTLS             []networkingv1.IngressTLS
+	ingressAnnotations     map[string]string
+	useIngressAgnosticMode bool
 
 	// Leader election
 	enableLeaderElection        bool
@@ -172,6 +173,7 @@ func NewStartCommand() *cobra.Command {
 	command.Flags().StringVar(&ingressURLFormat, "ingress-url-format", "", "Ingress URL format.")
 	command.Flags().StringVar(&ingressTLSstring, "ingress-tls", "", "JSON format string for the default TLS config on the Spark UI ingresses. e.g. '[{\"hosts\":[\"*.example.com\"],\"secretName\":\"example-secret\"}]'. `ingressTLS` in the SparkApplication spec will override this value.")
 	command.Flags().StringVar(&ingressAnnotationsString, "ingress-annotations", "", "JSON format string for the default ingress annotations for the Spark UI ingresses. e.g. '[{\"cert-manager.io/cluster-issuer\": \"letsencrypt\"}]'. `ingressAnnotations` in the SparkApplication spec will override this value.")
+	command.Flags().BoolVar(&useIngressAgnosticMode, "use-ingress-agnostic-mode", false, "When enabled, disables nginx-specific ingress path modifications (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target). This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.")
 
 	command.Flags().BoolVar(&enableLeaderElection, "leader-election", false, "Enable leader election for controller manager. "+
 		"Enabling this will ensure there is only one active controller manager.")
@@ -450,6 +452,7 @@ func newSparkApplicationReconcilerOptions() sparkapplication.Options {
 		SparkApplicationMetrics:      sparkApplicationMetrics,
 		SparkExecutorMetrics:         sparkExecutorMetrics,
 		MaxTrackedExecutorPerApp:     maxTrackedExecutorPerApp,
+		UseIngressAgnosticMode:       useIngressAgnosticMode,
 	}
 	if enableBatchScheduler {
 		options.KubeSchedulerNames = kubeSchedulerNames

config/crd/bases/sparkoperator.k8s.io_scheduledsparkapplications.yaml

@@ -5266,6 +5266,13 @@ spec:
                           description: ServiceType allows configuring the type of
                             the service. Defaults to ClusterIP.
                           type: string
+                        useIngressAgnosticMode:
+                          description: |-
+                            UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+                            (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+                            This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+                            When false (default), the legacy nginx-specific behavior is used for backward compatibility.
+                          type: boolean
                       required:
                       - servicePort
                       - servicePortName
@@ -10565,6 +10572,13 @@ spec:
                         description: ServiceType allows configuring the type of the
                           service. Defaults to ClusterIP.
                         type: string
+                      useIngressAgnosticMode:
+                        description: |-
+                          UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+                          (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+                          This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+                          When false (default), the legacy nginx-specific behavior is used for backward compatibility.
+                        type: boolean
                     type: object
                   sparkVersion:
                     description: SparkVersion is the version of Spark the application

config/crd/bases/sparkoperator.k8s.io_sparkapplications.yaml

@@ -5211,6 +5211,13 @@ spec:
                       description: ServiceType allows configuring the type of the
                         service. Defaults to ClusterIP.
                       type: string
+                    useIngressAgnosticMode:
+                      description: |-
+                        UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+                        (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+                        This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+                        When false (default), the legacy nginx-specific behavior is used for backward compatibility.
+                      type: boolean
                   required:
                   - servicePort
                   - servicePortName
@@ -10486,6 +10493,13 @@ spec:
                     description: ServiceType allows configuring the type of the service.
                       Defaults to ClusterIP.
                     type: string
+                  useIngressAgnosticMode:
+                    description: |-
+                      UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+                      (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+                      This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+                      When false (default), the legacy nginx-specific behavior is used for backward compatibility.
+                    type: boolean
                 type: object
               sparkVersion:
                 description: SparkVersion is the version of Spark the application

docs/api-docs.md

@@ -1168,6 +1168,21 @@ map[string]string
 <p>TlsHosts is useful If we need to declare SSL certificates to the ingress object</p>
 </td>
 </tr>
+<tr>
+<td>
+<code>useIngressAgnosticMode</code><br/>
+<em>
+bool
+</em>
+</td>
+<td>
+<em>(Optional)</em>
+<p>UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+(regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+When false (default), the legacy nginx-specific behavior is used for backward compatibility.</p>
+</td>
+</tr>
 </tbody>
 </table>
 <h3 id="sparkoperator.k8s.io/v1beta2.DriverSpec">DriverSpec
@@ -3465,6 +3480,21 @@ map[string]string
 <p>TlsHosts is useful If we need to declare SSL certificates to the ingress object</p>
 </td>
 </tr>
+<tr>
+<td>
+<code>useIngressAgnosticMode</code><br/>
+<em>
+bool
+</em>
+</td>
+<td>
+<em>(Optional)</em>
+<p>UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+(regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+When false (default), the legacy nginx-specific behavior is used for backward compatibility.</p>
+</td>
+</tr>
 </tbody>
 </table>
 <hr/>

internal/controller/sparkapplication/controller.go

@@ -72,6 +72,12 @@ type Options struct {
 	SparkExecutorMetrics    *metrics.SparkExecutorMetrics
 
 	MaxTrackedExecutorPerApp int
+
+	// UseIngressAgnosticMode when set to true, disables nginx-specific ingress path modifications
+	// (regex capture groups) and annotation injection (nginx.ingress.kubernetes.io/rewrite-target).
+	// This allows the ingress to work with any ingress controller like Traefik, HAProxy, etc.
+	// Can be overridden per SparkApplication via sparkUIOptions.useIngressAgnosticMode.
+	UseIngressAgnosticMode bool
 }
 
 // Reconciler reconciles a SparkApplication object.