Tune Ratelimiter config

examples/terraform/envoy-ratelimiter/ratelimit.tf

@@ -189,6 +189,14 @@ resource "kubernetes_deployment" "ratelimit" {
             name  = "STATSD_PORT"
             value = "9125"
           }
+          env {
+            name  = "GRPC_MAX_CONNECTION_AGE"
+            value = var.ratelimit_grpc_max_connection_age
+          }
+          env {
+            name  = "GRPC_MAX_CONNECTION_AGE_GRACE"
+            value = var.ratelimit_grpc_max_connection_age_grace
+          }
 
           resources {
             requests = var.ratelimit_resources.requests
@@ -214,6 +222,17 @@ resource "kubernetes_deployment" "ratelimit" {
             container_port = 9125
             protocol       = "UDP"
           }
+          # statsd-exporter does not use much resources, so setting resources to the minimum
+          resources {
+            requests = {
+              cpu    = "50m"
+              memory = "64Mi"
+            }
+            limits = {
+              cpu    = "100m"
+              memory = "128Mi"
+            }
+          }
         }
 
         volume {

examples/terraform/envoy-ratelimiter/variables.tf

@@ -89,7 +89,7 @@ variable "min_replicas" {
 variable "max_replicas" {
   description = "Maximum number of replicas for rate limit deployment"
   type        = number
-  default     = 5
+  default     = 100
 }
 
 variable "hpa_cpu_target_percentage" {
@@ -128,6 +128,19 @@ variable "ratelimit_log_level" {
   default     = "debug"
 }
 
+variable "ratelimit_grpc_max_connection_age" {
+  description = "Duration a connection may exist before it will be closed by sending a GoAway"
+  type        = string
+  default     = "5m"
+}
+
+variable "ratelimit_grpc_max_connection_age_grace" {
+  description = "Period after MaxConnectionAge after which the connection will be forcibly closed"
+  type        = string
+  default     = "1m"
+}
+
+
 variable "redis_resources" {
   description = "Compute resources for Redis container"
   type = object({
@@ -136,11 +149,11 @@ variable "redis_resources" {
   })
   default = {
     requests = {
-      cpu    = "100m"
+      cpu    = "250m"
       memory = "128Mi"
     }
     limits = {
-      cpu    = "500m"
+      cpu    = "1000m"
       memory = "512Mi"
     }
   }
@@ -154,8 +167,8 @@ variable "ratelimit_resources" {
   })
   default = {
     requests = {
-      cpu    = "250m"
-      memory = "256Mi"
+      cpu    = "150m"
+      memory = "128Mi"
     }
     limits = {
       cpu    = "500m"

sdks/java/io/components/src/main/java/org/apache/beam/sdk/io/components/ratelimiter/EnvoyRateLimiterFactory.java

@@ -39,8 +39,9 @@
 /** A {@link RateLimiterFactory} for Envoy Rate Limit Service. */
 public class EnvoyRateLimiterFactory implements RateLimiterFactory {
   private static final Logger LOG = LoggerFactory.getLogger(EnvoyRateLimiterFactory.class);
-  private static final int RPC_RETRY_COUNT = 5;
-  private static final long RPC_RETRY_DELAY_MILLIS = 1000;
+  // Retry up to 1 min rather than failing fast and asking runner to retry full bundle.
+  private static final int RPC_RETRY_COUNT = 10;
+  private static final long RPC_RETRY_DELAY_MILLIS = 6000;
 
   private final RateLimiterOptions options;