feat(translator): Implement Retry in BTP

api/v1alpha1/retry_types.go

@@ -37,6 +37,7 @@ type RetryOn struct {
 	Triggers []TriggerEnum `json:"triggers,omitempty"`
 
 	// HttpStatusCodes specifies the http status codes to be retried.
+	// The retriable-status-codes trigger must also be configured for these status codes to trigger a retry.
 	//
 	// +optional
 	HTTPStatusCodes []HTTPStatus `json:"httpStatusCodes,omitempty"`

charts/gateway-helm/crds/generated/gateway.envoyproxy.io_backendtrafficpolicies.yaml

@@ -786,7 +786,8 @@ spec:
                     properties:
                       httpStatusCodes:
                         description: HttpStatusCodes specifies the http status codes
-                          to be retried.
+                          to be retried. The retriable-status-codes trigger must also
+                          be configured for these status codes to trigger a retry.
                         items:
                           description: HTTPStatus defines the http status code.
                           exclusiveMaximum: true

internal/gatewayapi/backendtrafficpolicy.go

@@ -251,6 +251,7 @@ func (t *Translator) translateBackendTrafficPolicyForRoute(policy *egv1a1.Backen
 		fi *ir.FaultInjection
 		to *ir.Timeout
 		ka *ir.TCPKeepalive
+		rt *ir.Retry
 	)
 
 	// Build IR
@@ -276,6 +277,9 @@ func (t *Translator) translateBackendTrafficPolicyForRoute(policy *egv1a1.Backen
 	if policy.Spec.TCPKeepalive != nil {
 		ka = t.buildTCPKeepAlive(policy)
 	}
+	if policy.Spec.Retry != nil {
+		rt = t.buildRetry(policy)
+	}
 	// Apply IR to all relevant routes
 	prefix := irRoutePrefix(route)
 	for _, ir := range xdsIR {
@@ -290,6 +294,7 @@ func (t *Translator) translateBackendTrafficPolicyForRoute(policy *egv1a1.Backen
 					r.CircuitBreaker = cb
 					r.FaultInjection = fi
 					r.TCPKeepalive = ka
+					r.Retry = rt
 
 					// some timeout setting originate from the route
 					if policy.Spec.Timeout != nil {
@@ -313,6 +318,7 @@ func (t *Translator) translateBackendTrafficPolicyForGateway(policy *egv1a1.Back
 		fi *ir.FaultInjection
 		ct *ir.Timeout
 		ka *ir.TCPKeepalive
+		rt *ir.Retry
 	)
 
 	// Build IR
@@ -337,6 +343,9 @@ func (t *Translator) translateBackendTrafficPolicyForGateway(policy *egv1a1.Back
 	if policy.Spec.TCPKeepalive != nil {
 		ka = t.buildTCPKeepAlive(policy)
 	}
+	if policy.Spec.Retry != nil {
+		rt = t.buildRetry(policy)
+	}
 
 	// Apply IR to all the routes within the specific Gateway
 	// If the feature is already set, then skip it, since it must be have
@@ -369,6 +378,9 @@ func (t *Translator) translateBackendTrafficPolicyForGateway(policy *egv1a1.Back
 			if r.TCPKeepalive == nil {
 				r.TCPKeepalive = ka
 			}
+			if r.Retry == nil {
+				r.Retry = rt
+			}
 
 			if policy.Spec.Timeout != nil {
 				ct = t.buildTimeout(policy, r)
@@ -1016,3 +1028,90 @@ func (t *Translator) buildTCPKeepAlive(policy *egv1a1.BackendTrafficPolicy) *ir.
 	}
 	return ka
 }
+
+func (t *Translator) buildRetry(policy *egv1a1.BackendTrafficPolicy) *ir.Retry {
+	var rt *ir.Retry
+	if policy.Spec.Retry != nil {
+		prt := policy.Spec.Retry
+		rt = &ir.Retry{}
+
+		if prt.NumRetries != nil {
+			rt.NumRetries = ptr.To(uint32(*prt.NumRetries))
+		}
+
+		if prt.RetryOn != nil {
+			ro := &ir.RetryOn{}
+			bro := false
+			if prt.RetryOn.HTTPStatusCodes != nil {
+				ro.HTTPStatusCodes = makeIrStatusSet(prt.RetryOn.HTTPStatusCodes)
+				bro = true
+			}
+
+			if prt.RetryOn.Triggers != nil {
+				ro.Triggers = makeIrTriggerSet(prt.RetryOn.Triggers)
+				bro = true
+			}
+
+			if bro {
+				rt.RetryOn = ro
+			}
+		}
+
+		if prt.PerRetry != nil {
+			pr := &ir.PerRetryPolicy{}
+			bpr := false
+
+			if prt.PerRetry.Timeout != nil {
+				pr.Timeout = prt.PerRetry.Timeout
+				bpr = true
+			}
+
+			if prt.PerRetry.BackOff != nil {
+				if prt.PerRetry.BackOff.MaxInterval != nil || prt.PerRetry.BackOff.BaseInterval != nil {
+					bop := &ir.BackOffPolicy{}
+					if prt.PerRetry.BackOff.MaxInterval != nil {
+						bop.MaxInterval = prt.PerRetry.BackOff.MaxInterval
+					}
+
+					if prt.PerRetry.BackOff.BaseInterval != nil {
+						bop.BaseInterval = prt.PerRetry.BackOff.BaseInterval
+					}
+					pr.BackOff = bop
+					bpr = true
+				}
+			}
+
+			if bpr {
+				rt.PerRetry = pr
+			}
+		}
+	}
+
+	return rt
+}
+
+func makeIrStatusSet(in []egv1a1.HTTPStatus) []ir.HTTPStatus {
+	var irStatuses []ir.HTTPStatus
+	// deduplicate http statuses
+	statusSet := make(map[egv1a1.HTTPStatus]bool, len(in))
+	for _, r := range in {
+		if _, ok := statusSet[r]; !ok {
+			statusSet[r] = true
+			irStatuses = append(irStatuses, ir.HTTPStatus(r))
+		}
+	}
+	return irStatuses
+}
+
+func makeIrTriggerSet(in []egv1a1.TriggerEnum) []ir.TriggerEnum {
+	var irTriggers []ir.TriggerEnum
+	// deduplicate http statuses
+	triggerSet := make(map[egv1a1.TriggerEnum]bool, len(in))
+	for _, r := range in {
+		if _, ok := triggerSet[r]; !ok {
+			triggerSet[r] = true
+			irTriggers = append(irTriggers, ir.TriggerEnum(r))
+		}
+	}
+	return irTriggers
+}

internal/gatewayapi/route.go

@@ -657,6 +657,7 @@ func (t *Translator) processHTTPRouteParentRefListener(route RouteContext, route
 					Mirrors:               routeRoute.Mirrors,
 					ExtensionRefs:         routeRoute.ExtensionRefs,
 					Timeout:               routeRoute.Timeout,
+					Retry:                 routeRoute.Retry,
 				}
 				// Don't bother copying over the weights unless the route has invalid backends.
 				if routeRoute.BackendWeights.Invalid > 0 {

internal/gatewayapi/testdata/backendtrafficpolicy-with-retries.in.yaml

@@ -0,0 +1,110 @@
+gateways:
+- apiVersion: gateway.networking.k8s.io/v1
+  kind: Gateway
+  metadata:
+    namespace: envoy-gateway
+    name: gateway-1
+  spec:
+    gatewayClassName: envoy-gateway-class
+    listeners:
+    - name: http
+      protocol: HTTP
+      port: 80
+      allowedRoutes:
+        namespaces:
+          from: All
+- apiVersion: gateway.networking.k8s.io/v1
+  kind: Gateway
+  metadata:
+    namespace: envoy-gateway
+    name: gateway-2
+  spec:
+    gatewayClassName: envoy-gateway-class
+    listeners:
+    - name: http
+      protocol: HTTP
+      port: 80
+      allowedRoutes:
+        namespaces:
+          from: All
+grpcRoutes:
+- apiVersion: gateway.networking.k8s.io/v1alpha2
+  kind: GRPCRoute
+  metadata:
+    namespace: default
+    name: grpcroute-1
+  spec:
+    parentRefs:
+    - namespace: envoy-gateway
+      name: gateway-1
+      sectionName: http
+    rules:
+    - backendRefs:
+      - name: service-1
+        port: 8080
+httpRoutes:
+- apiVersion: gateway.networking.k8s.io/v1
+  kind: HTTPRoute
+  metadata:
+    namespace: default
+    name: httproute-1
+  spec:
+    hostnames:
+    - gateway.envoyproxy.io
+    parentRefs:
+    - namespace: envoy-gateway
+      name: gateway-2
+      sectionName: http
+    rules:
+    - matches:
+      - path:
+          value: "/"
+      backendRefs:
+      - name: service-1
+        port: 8080
+backendTrafficPolicies:
+- apiVersion: gateway.envoyproxy.io/v1alpha1
+  kind: BackendTrafficPolicy
+  metadata:
+    namespace: envoy-gateway
+    name: policy-for-gateway
+  spec:
+    targetRef:
+      group: gateway.networking.k8s.io
+      kind: Gateway
+      name: gateway-1
+      namespace: envoy-gateway
+    retry:
+      retryOn:
+        triggers:
+        - cancelled
+      perRetry:
+        timeout: 250ms
+        backoff:
+          baseInterval: 100ms
+          maxInterval: 10s
+- apiVersion: gateway.envoyproxy.io/v1alpha1
+  kind: BackendTrafficPolicy
+  metadata:
+    namespace: default
+    name: policy-for-route
+  spec:
+    targetRef:
+      group: gateway.networking.k8s.io
+      kind: HTTPRoute
+      name: httproute-1
+      namespace: default
+    retry:
+      numRetries: 5
+      retryOn:
+        httpStatusCodes:
+        - 429
+        - 503
+        triggers:
+        - connect-failure
+        - retriable-status-codes
+      perRetry:
+        timeout: 250ms
+        backoff:
+          baseInterval: 100ms
+          maxInterval: 10s