TransportServer resources

This document is reference material for the TransportServer resource used by F5 NGINX Ingress Controller.

The TransportServer resource allows you to configure TCP, UDP, and TLS Passthrough load balancing. The resource is implemented as a Custom Resource.

The GitHub repository has examples of the resources for specific use cases.

Prerequisites

  • For TCP and UDP, the TransportServer resource must be used in conjunction with the GlobalConfiguration resource, which must be created separately.
  • For TLS Passthrough, make sure to enable the -enable-tls-passthrough command-line argument of NGINX Ingress Controller.

TransportServer Specification

The TransportServer resource defines load balancing configuration for TCP, UDP, or TLS Passthrough traffic. Below are a few examples:

  • TCP load balancing:

    apiVersion: k8s.nginx.org/v1
    kind: TransportServer
    metadata:
      name: dns-tcp
    spec:
      host: dns.example.com
      listener:
        name: dns-tcp
        protocol: TCP
      tls:
        secret: cafe-secret
      upstreams:
      - name: dns-app
        service: dns-service
        port: 5353
      action:
        pass: dns-app
    
  • UDP load balancing:

    apiVersion: k8s.nginx.org/v1
    kind: TransportServer
    metadata:
      name: dns-udp
    spec:
      listener:
        name: dns-udp
        protocol: UDP
      upstreams:
      - name: dns-app
        service: dns-service
        port: 5353
      upstreamParameters:
        udpRequests: 1
        udpResponses: 1
      action:
        pass: dns-app
    
  • TLS passthrough load balancing:

    apiVersion: k8s.nginx.org/v1
    kind: TransportServer
    metadata:
      name: secure-app
    spec:
      listener:
        name: tls-passthrough
        protocol: TLS_PASSTHROUGH
      host: app.example.com
      upstreams:
      - name: secure-app
        service: secure-app
        port: 8443
      action:
        pass: secure-app
    
Field Description Type Required
listener The listener on NGINX that will accept incoming connections/datagrams. listener Yes
host The host (domain name) of the server. Must be a valid subdomain as defined in RFC 1123, such as my-app or hello.example.com. Wildcard domains like *.example.com are not allowed. When specified, NGINX will use this host for SNI-based routing. For TLS Passthrough, this field is required. For TCP with TLS termination, specifying the host enables SNI routing and requires specifying a TLS secret. string No
tls The TLS termination configuration. Not supported for TLS Passthrough load balancing. tls No
upstreams A list of upstreams. []upstream Yes
upstreamParameters The upstream parameters. upstreamParameters No
action The action to perform for a client connection/datagram. action Yes
ingressClassName Specifies which Ingress Controller must handle the TransportServer resource. string No
streamSnippets Sets a custom snippet in the stream context. string No
serverSnippets Sets a custom snippet in the server context. string No

* – Required for TLS Passthrough load balancing.

Listener

The listener field references a listener that NGINX will use to accept incoming traffic for the TransportServer. For TCP and UDP, the listener must be defined in the GlobalConfiguration resource. When referencing a listener, both the name and the protocol must match. For TLS Passthrough, use the built-in listener with the name tls-passthrough and the protocol TLS_PASSTHROUGH.

The combination of spec.listener.name and spec.host must be unique among all TransportServers. If two TransportServers specify the same combination of spec.listener.name and spec.host, one of them will be rejected to prevent conflicts. In the case where no host is specified, it is considered an empty string.

An example:

listener:
  name: dns-udp
  protocol: UDP
Field Description Type Required
name The name of the listener. string Yes
protocol The protocol of the listener. string Yes

TLS

The tls field defines TLS configuration for a TransportServer. When using TLS termination (not TLS Passthrough), you can specify the host field to enable SNI-based routing, allowing multiple applications to share the same listener port and be distinguished by the TLS SNI hostname. Each application can use its own TLS certificate and key specified via the secret field.

secret: cafe-secret
Field Description Type Required
secret The name of a secret with a TLS certificate and key. The secret must belong to the same namespace as the TransportServer. The secret must be of the type kubernetes.io/tls and contain keys named tls.crt and tls.key that contain the certificate and private key as described here. string No

Upstream

The upstream defines a destination for the TransportServer. For example:

name: secure-app
service: secure-app
port: 8443
maxFails: 3
maxConns: 100
failTimeout: 30s
loadBalancingMethod: least_conn
Field Description Type Required
name The name of the upstream. Must be a valid DNS label as defined in RFC 1035. For example, hello and upstream-123 are valid. The name must be unique among all upstreams of the resource. string Yes
service The name of a service. The service must belong to the same namespace as the resource. If the service doesn’t exist, NGINX will assume the service has zero endpoints and close client connections/ignore datagrams. string Yes
port The port of the service. If the service doesn’t define that port, NGINX will assume the service has zero endpoints and close client connections/ignore datagrams. The port must fall into the range 1..65535. int Yes
maxFails Sets the number of unsuccessful attempts to communicate with the server that should happen in the duration set by the failTimeout parameter to consider the server unavailable. The default 1. int No
maxConns Sets the number of maximum connections to the proxied server. Default value is zero, meaning there is no limit. The default is 0. int No
failTimeout Sets the time during which the specified number of unsuccessful attempts to communicate with the server should happen to consider the server unavailable and the period of time the server will be considered unavailable. The default is 10s. string No
healthCheck The health check configuration for the Upstream. See the health_check directive. Note: this feature is supported only in NGINX Plus. healthcheck No
loadBalancingMethod The method used to load balance the upstream servers. By default, connections are distributed between the servers using a weighted round-robin balancing method. See the upstream section for available methods and their details. string No
backup The name of the backup service of type ExternalName. This will be used when the primary servers are unavailable. Note: The parameter cannot be used along with the random , hash or ip_hash load balancing methods. string No
backupPort The port of the backup service. The backup port is required if the backup service name is provided. The port must fall into the range 1..65535. uint16 No

Upstream.Healthcheck

The Healthcheck defines an active health check. In the example below we enable a health check for an upstream and configure all the available parameters:

name: secure-app
service: secure-app
port: 8443
healthCheck:
  enable: true
  interval: 20s
  timeout: 30s
  jitter: 3s
  fails: 5
  passes: 5
  port: 8080
Note:
This feature is only supported with NGINX Plus.
Field Description Type Required
enable Enables a health check for an upstream server. The default is false. boolean No
interval The interval between two consecutive health checks. The default is 5s. string No
timeout This overrides the timeout set by proxy_timeout which is set in SessionParameters for health checks. The default value is 5s. string No
jitter The time within which each health check will be randomly delayed. By default, there is no delay. string No
fails The number of consecutive failed health checks of a particular upstream server after which this server will be considered unhealthy. The default is 1. integer No
passes The number of consecutive passed health checks of a particular upstream server after which the server will be considered healthy. The default is 1. integer No
port The port used for health check requests. By default, the server port is used. Note: in contrast with the port of the upstream, this port is not a service port, but a port of a pod. integer No
match Controls the data to send and the response to expect for the healthcheck. match No

Upstream.Healthcheck.Match

The match controls the data to send and the response to expect for the healthcheck:

match:
  send: 'GET / HTTP/1.0\r\nHost: localhost\r\n\r\n'
  expect: "~200 OK"

Both send and expect fields can contain hexadecimal literals with the prefix \x followed by two hex digits, for example, \x80.

See the match directive for details.

Field Description Type Required
send A string to send to an upstream server. string No
expect A literal string or a regular expression that the data obtained from the server should match. The regular expression is specified with the preceding ~* modifier (for case-insensitive matching), or the ~ modifier (for case-sensitive matching). NGINX Ingress Controller validates a regular expression using the RE2 syntax. string No

UpstreamParameters

The upstream parameters define various parameters for the upstreams:

upstreamParameters:
  udpRequests: 1
  udpResponses: 1
  connectTimeout: 60s
  nextUpstream: true
  nextUpstreamTimeout: 50s
  nextUpstreamTries: 1
Field Description Type Required
udpRequests The number of datagrams, after receiving which, the next datagram from the same client starts a new session. See the proxy_requests directive. The default is 0. int No
udpResponses The number of datagrams expected from the proxied server in response to a client datagram. See the proxy_responses directive. By default, the number of datagrams is not limited. int No
connectTimeout The timeout for establishing a connection with a proxied server. See the proxy_connect_timeout directive. The default is 60s. string No
nextUpstream If a connection to the proxied server cannot be established, determines whether a client connection will be passed to the next server. See the proxy_next_upstream directive. The default is true. bool No
nextUpstreamTries The number of tries for passing a connection to the next server. See the proxy_next_upstream_tries directive. The default is 0. int No
nextUpstreamTimeout The time allowed to pass a connection to the next server. See the proxy_next_upstream_timeout directive. The default us 0. string No

SessionParameters

The session parameters define various parameters for TCP connections and UDP sessions.

sessionParameters:
  timeout: 50s
Field Description Type Required
timeout The timeout between two successive read or write operations on client or proxied server connections. See proxy_timeout directive. The default is 10m. string No

Action

The action defines an action to perform for a client connection/datagram.

In the example below, client connections/datagrams are passed to an upstream dns-app:

action:
  pass: dns-app
Field Description Type Required
pass Passes connections/datagrams to an upstream. The upstream with that name must be defined in the resource. string Yes

Using TransportServer

You can use the usual kubectl commands to work with TransportServer resources, similar to Ingress resources.

For example, the following command creates a TransportServer resource defined in transport-server-passthrough.yaml with the name secure-app:

kubectl apply -f transport-server-passthrough.yaml
transportserver.k8s.nginx.org/secure-app created

You can get the resource by running:

kubectl get transportserver secure-app
NAME         AGE
secure-app   46sm

In the kubectl get and similar commands, you can also use the short name ts instead of transportserver.

Using Snippets

Snippets allow you to insert raw NGINX config into different contexts of NGINX configuration. In the example below, we use snippets to configure access control in a TransportServer:

apiVersion: k8s.nginx.org/v1
kind: TransportServer
metadata:
  name: cafe
spec:
  host: cafe.example.com
  serverSnippets: |
    deny  192.168.1.1;
    allow 192.168.1.0/24;    
  upstreams:
  - name: tea
    service: tea-svc
    port: 80

Snippets can also be specified for a stream. In the example below, we use snippets to limit the number of connections:

apiVersion: k8s.nginx.org/v1
kind: TransportServer
metadata:
  name: cafe
spec:
  host: cafe.example.com
  streamSnippets: limit_conn_zone $binary_remote_addr zone=addr:10m;
  serverSnippets: limit_conn addr 1;
  upstreams:
  - name: tea
    service: tea-svc
    port: 80
Note:
To configure snippets in the stream context, use stream-snippets ConfigMap key.

For additional information, view the Advanced configuration with Snippets topic.

Validation

Two types of validation are available for the TransportServer resource:

  • Structural validation by the kubectl and Kubernetes API server.
  • Comprehensive validation by NGINX Ingress Controller.

Structural Validation

The custom resource definition for the TransportServer includes structural OpenAPI schema which describes the type of every field of the resource.

If you try to create (or update) a resource that violates the structural schema (for example, you use a string value for the port field of an upstream), kubectl and Kubernetes API server will reject such a resource:

  • Example of kubectl validation:

    kubectl apply -f transport-server-passthrough.yaml
    
    error: error validating "transport-server-passthrough.yaml": error validating data: ValidationError(TransportServer.spec.upstreams[0].port): invalid type for org.nginx.k8s.v1.TransportServer.spec.upstreams.port: got "string", expected "integer"; if you choose to ignore these errors, turn validation off with --validate=false
    
  • Example of Kubernetes API server validation:

    kubectl apply -f transport-server-passthrough.yaml --validate=false
    
    The TransportServer "secure-app" is invalid: []: Invalid value: map[string]interface {}{ ... }: validation failure list:
    spec.upstreams.port in body must be of type integer: "string"
    

If a resource is not rejected (it doesn’t violate the structural schema), NGINX Ingress Controller will validate it further.

Comprehensive Validation

NGINX Ingress Controller validates the fields of a TransportServer resource. If a resource is invalid, NGINX Ingress Controller will reject it: the resource will continue to exist in the cluster, but NGINX Ingress Controller will ignore it.

You can check if NGINX Ingress Controller successfully applied the configuration for a TransportServer. For our example secure-app TransportServer, we can run:

kubectl describe ts secure-app
...
Events:
  Type    Reason          Age   From                      Message
  ----    ------          ----  ----                      -------
  Normal  AddedOrUpdated  3s    nginx-ingress-controller  Configuration for default/secure-app was added or updated

Note how the events section includes a Normal event with the AddedOrUpdated reason that informs us that the configuration was successfully applied.

If you create an invalid resource, NGINX Ingress Controller will reject it and emit a Rejected event. For example, if you create a TransportServer secure-app with a pass action that references a non-existing upstream, you will get :

kubectl describe ts secure-app
...
Events:
  Type     Reason    Age   From                      Message
  ----     ------    ----  ----                      -------
  Warning  Rejected  2s    nginx-ingress-controller  TransportServer default/secure-app is invalid and was rejected: spec.action.pass: Not found: "some-app"

Note how the events section includes a Warning event with the Rejected reason.

Note: If you make an existing resource invalid, NGINX Ingress Controller will reject it and remove the corresponding configuration from NGINX.

Customization via ConfigMap

The ConfigMap keys (except for stream-snippets, stream-log-format, resolver-addresses, resolver-ipv6, resolver-valid and resolver-timeout) do not affect TransportServer resources.


Last modified October 22, 2024