You can configure the following listener connection settings:

If an upstream service is unavailable, the service is removed from the load balancing pool until health is re-established.

For more information, see the following resources.

About

Learn more about the listener connection settings that you can modify.

About read and write buffer limits

By default, Gloo Gateway is set up with 1MiB of request read and write buffer for each gateway listener. For large requests that must be buffered and that exceed the default buffer limit, Gloo Gateway either disconnects the connection to the downstream service if headers were already sent, or returns a 500 HTTP response code. To make sure that large requests can be sent and received, you can specify the maximum number of bytes that you want to allow to be buffered between the gateway and the downstream service.

About TCP keepalive settings for downstream apps

Configure TCP keepalive for downstream services to keep a connection to the gateway open during long idle periods. A typical use case for this setting is deploying an AWS NLB instance in front of Gloo Gateway. The AWS NLB has an idle timeout of 350 seconds that cannot be changed. Without TCP keepalive, the connection from the gateway to the AWS NLB silently closes after 350 seconds of idle time. However, the connection from the client to the AWS NLB remains intact. When the client sends a request to the AWS NLB, the request cannot be forwarded to the gateway anymore and the NLB returns a reset packet (RST). Because the client cannot interpret the RST packet, the connection to the AWS NLB is closed.

To make sure that the connection from the AWS NLB to the gateway remains open, even during long idle periods, you can configure TCP keepalive for a listener.

For more information, see the following resources.

Before you begin

  1. Set up Gloo Mesh Gateway in a single cluster.
  2. Install Bookinfo and other sample apps.

  3. Configure an HTTP listener on your gateway and set up basic routing for the sample apps.

Configure listener connection policies

You can apply a connection policy at the gateway listener level. For more information, see Applying policies.

The following example sets up a maximum buffer limit of 16 megabytes for each connection to the HTTP listener on the istio-ingressgateway virtual gateway.

  apiVersion: resilience.policy.gloo.solo.io/v2
kind: ListenerConnectionPolicy
metadata:
  name: listener-connection
  namespace: httpbin
spec:
  applyToListeners:
  - virtualGateway:
      name: istio-ingressgateway
      namespace: bookinfo
      cluster: $CLUSTER_NAME
    port:
      number: 80
  config:
    perConnectionBufferLimitBytes: 16384

Review the following table to understand this configuration. For more information, see the API docs.

SettingDescription
applyToListenersThe gateway listener that you want to apply this policy to. To select a gateway listener, you must reference the Gloo virtual gateway and the port number that the listener was configured for. To learn more about gateway listeners, see Listener overview. If you do not specify a valid listener, the policy takes no effect.
perConnectionBufferLimitBytesThe maximum number of bytes that you want to allow to be buffered for each connection between the gateway and a downstream service. The default value is 1MiB.

In the following example, the connection between the gateway listener and the downstream service must be idle for 240 seconds before TCP probes are sent. After 240 seconds, the gateway starts to send TCP probes in 1 minute intervals. If the TCP keepalive packet is acknowledged by the downstream service, the connection is considered healthy. If the packet is not acknowledged, the gateway sends another probe after the 1 minute interval. If no response is received after 5 TCP keepalive probes, the connection to the downstream service is considered failed and the gateway closes connection.

  apiVersion: resilience.policy.gloo.solo.io/v2
kind: ListenerConnectionPolicy
metadata:
  annotations:
    cluster.solo.io/cluster: ""
  name: http-buffer
  namespace: bookinfo
spec:
  applyToListeners:
  - port:
      number: 443
    virtualGateway:
      cluster: cluster-1
      name: istio-ingressgateway
      namespace: bookinfo
  config:
    tcpKeepalive:
      interval: 1m
      probes: 5
      time: 240s

Review the following table to understand this configuration. For more information, see the API docs.

SettingDescription
applyToListenersThe gateway listener that you want to apply this policy to. To select a gateway listener, you must reference the Gloo virtual gateway and the port number that the listener was configured for. To learn more about gateway listeners, see Listener overview. If you do not specify a valid listener, the policy takes no effect.
tcpKeepalive.intervalThe time duration between TCP keepalive probes. Example formats include 1ms, 1s, 1m, or 1h. For more information about the time value format, see the Google protocol buffer documentation.
tcpKeepalive.probesThe maximum number of TCP keepalive packets that are sent before a connection between the gateway and the downstream service is considered failed.
tcpKeepalive.timeThe time duration a connection between the gateway and the downstream service must to be idle before TCP keepalive probes are sent. Example formats include 1ms, 1s, 1m, or 1h. For more information about the time value format, see the Google protocol buffer documentation.

Verify listener connection policies

Verify that the listener connection settings are updated in the Envoy filter.

  1. Apply the listener connection policy in your cluster.

      kubectl apply -f- <<EOF 
apiVersion: resilience.policy.gloo.solo.io/v2
kind: ListenerConnectionPolicy
metadata:
  name: listener-connection
  namespace: bookinfo
spec:
  config:
    perConnectionBufferLimitBytes: 16384
    tcpKeepalive:
      interval: 1m
      probes: 5
      time: 240s
  applyToListeners:
  - virtualGateway:
      name: istio-ingressgateway
      namespace: bookinfo
      cluster: $CLUSTER_NAME
    port:
      number: 80
EOF

  2. Check that the configuration is applied in the Envoy filter.

      kubectl get envoyfilter istio-ingressgateway-listener-8080-listener-connection -n gloo-mesh-gateways -o yaml

    Example output:

       apiVersion: networking.istio.io/v1alpha3
   kind: EnvoyFilter
   metadata:
     annotations:
       cluster.solo.io/cluster: gloo-gateway-docs-mgt
     creationTimestamp: "2023-06-05T18:05:26Z"
     generation: 1
     labels:
       context.mesh.gloo.solo.io/cluster: gloo-gateway-docs-mgt
       context.mesh.gloo.solo.io/namespace: gloo-mesh-gateways
       context.mesh.gloo.solo.io/workspace: gloo-gateway-docs-mgt
       gloo.solo.io/parent_cluster: gloo-gateway-docs-mgt
       gloo.solo.io/parent_group: networking.gloo.solo.io
       gloo.solo.io/parent_kind: VirtualGateway
       gloo.solo.io/parent_name: istio-ingressgateway
       gloo.solo.io/parent_namespace: bookinfo
       gloo.solo.io/parent_version: v2
       reconciler.mesh.gloo.solo.io/name: translator
     name: istio-ingressgateway-listener-8080-listener-connection
     namespace: gloo-mesh-gateways
     resourceVersion: "18104008"
     uid: 962d88bf-7373-45f9-98bc-510de612d3ec
   spec:
     configPatches:
     - applyTo: LISTENER
       match:
         listener:
           portNumber: 8080
       patch:
         operation: MERGE
         value:
           per_connection_buffer_limit_bytes: 16384
     - applyTo: LISTENER
       match:
         listener:
           portNumber: 8080
       patch:
         operation: MERGE
         value:
           socket_options:
           - description: enable keep-alive
             int_value: 1
             level: 1
             name: 9
             state: STATE_PREBIND
           - description: idle time before first keep-alive probe is sent
             int_value: 240
             level: 6
             name: 4
             state: STATE_PREBIND
           - description: keep-alive interval
             int_value: 60
             level: 6
             name: 5
             state: STATE_PREBIND
           - description: keep-alive probes count
             int_value: 5
             level: 6
             name: 6
             state: STATE_PREBIND
     workloadSelector:
       labels:
         istio: ingressgateway

Cleanup

You can optionally remove the resources that you set up as part of this guide.
  kubectl delete listenerconnectionpolicy listener-connection -n bookinfo