External WebLogic clients

Contents

Overview

When WebLogic clients and servers are external to Kubernetes, and you want to give them access to Kubernetes hosted WebLogic cluster EJBs and JMS, there are two supported approaches:

Conversely, if giving a Kubernetes hosted WebLogic Server access to external WebLogic Server EJBs, JMS, or JTA, then consider the following:

  • You may need to enable unknown host access on the external WebLogic Servers.
  • Plus, if the target server can be accessed only through a load balancer using HTTP:
    • Set up an HTTP tunneling-enabled custom channel on the external WebLogic Servers.
    • Specify URLs on the source server that resolve to the load balancer’s address and that start with http instead of t3.
    • Ensure the load balancer configures the HTTP flow to be ‘sticky’.

The operator does not currently support external WebLogic JTA access to a Kubernetes hosted WebLogic cluster. This is because external JTA access requires each server in a cluster to be individually addressable, but this conflicts with the current operator requirement that a network channel in a cluster have the same port across all servers in the cluster.

Load balancer tunneling

Load balancer tunneling is the preferred approach for giving external clients and servers access to a Kubernetes hosted WebLogic cluster EJB and JMS. This approach involves configuring a network channel on the desired WebLogic cluster that accepts T3 protocol traffic that’s tunneled over HTTP, deploying a load balancer that redirects external HTTP network traffic to the desired WebLogic network channel, and ensuring that EJB and JMS clients specify a URL that resolves the load balancer’s network address.

Here are the steps:

  • In WebLogic, configure a custom channel for the T3 protocol that enables HTTP tunneling, and specifies an external address and port that correspond to the address and port remote clients will use to access the load balancer. See Adding a WebLogic custom channel for samples and details.

  • Set up a load balancer that redirects HTTP traffic to the custom channel. For more information on load balancers, see Ingress. If you’re using OKE/OCI to host your Kubernetes cluster, also see Using an OCI Load Balancer.

  • Important: Ensure that the load balancer configures the HTTP flow to be ‘sticky’ - for example, a Traefik load balancer has a sticky sessions option. This ensures that all of the packets of a tunneling client connection flow to the same pod, otherwise the connection will stall when its packets are load balanced to a different pod.

  • If you are adding access for remote WebLogic Servers, then the Kubernetes hosted servers may need to enable unknown host access.

  • Remote clients and servers can then access the custom channel using an http:// URL instead of a t3:// URL.

  • Review the Security notes.

Kubernetes NodePorts

Kubernetes NodePorts provide an alternative approach for giving external WebLogic EJB or JMS clients access to a Kubernetes hosted WebLogic cluster. This approach involves configuring a network channel on the desired WebLogic cluster that accepts T3 protocol traffic, and exposing a Kubernetes NodePort that redirects external network traffic on the Kubernetes Nodes to the network channel.

The NodePort approach is available only when worker nodes are accessible by the clients, for example, when they have public IP addresses. If private worker nodes are used and access to them is possible only through a load balancer or bastion, then the NodePort approach is not a valid option to provide access to external clients.

Here are the steps:

Adding a WebLogic custom channel

When is a WebLogic custom channel needed?

WebLogic implicitly creates a multi-protocol default channel that spans the Listen Address and Port fields specified on each server in the cluster, but this channel is usually unsuitable for external network traffic from EJB and JMS clients. Instead, you may need to configure an additional dedicated WebLogic custom channel to handle remote EJB or JMS client network traffic.

A custom channel provides a way to configure an external listen address and port for use by external clients, unlike a default channel. External listen address or port configuration is needed when a channel’s configured listen address or port would not work if used to form a URL in the remote client. This is because remote EJB and JMS clients internally use their client’s channel’s configured network information to reconnect to WebLogic when needed. (The EJB and JMS clients do not always use the initial URL specified in the client’s JNDI context.)

A custom channel can be locked down using two-way SSL as a way to prevent access by unauthorized external JMS and EJB clients, only accepts protocols that are explicitly enabled for the channel, and can be configured to be the only channel that accepts EJB/JMS clients that tunnel over HTTP. A default channel may often be deliberately unencrypted for convenient internal use, or, if used externally, is used for web traffic (not tunneling traffic) only. In addition, a default channel supports several protocols but it’s a best practice to limit the protocols that can be accessed by external clients. Finally, external clients may require access using HTTP tunneling in order to make connections, but it’s often inadvisable to enable tunneling for an unsecured default channel that’s already servicing external HTTP traffic. This is because enabling HTTP tunneling would potentially allow unauthorized external JMS and EJB clients unsecured access to the WebLogic cluster through the same HTTP path.

Configuring a WebLogic custom channel

The basic requirements for configuring a custom channel for remote EJB and JMS access are:

  • Configure a T3 protocol network access point (NAP) with the same name and port on each server (the operator will set the listen address for you).

  • Configure the external listen address and port on each NAP to match the address and port component of a URL your clients can use. For example, if you are providing access to remote clients using a load balancer, then these should match the address and port of the load balancer.

  • If you want WebLogic T3 clients to tunnel through HTTP, then enable HTTP tunneling on each NAP. This is often necessary for load balancers.

  • Do NOT set outbound-enabled to true on the network access point (the default is false), because this may cause internal network traffic to stall in an attempt to route through the network access point.

  • For operator controlled WebLogic clusters, ensure you haven’t enabled calculated-listen-ports for WebLogic dynamic cluster servers. The operator requires that a channel have the same port on each server in a cluster, but calculated-listen-ports causes the port to be different on each server.

  • For clusters that are not operator controlled, minimally ensure that the server’s default channel ListenAddress is configured. Oracle strongly recommends configuring a ListenAddress on all WebLogic Servers. Note that if a NAP’s ListenAddress is left blank, then it will use the default channel’s ListenAddress. (This is not a concern for operator controlled clusters as the operator sets the listen addresses on every WebLogic Server.)

For example, here is a snippet of a WebLogic domain config.xml file for channel MyChannel defined for an operator controlled WebLogic dynamic cluster named cluster-1:

<server-template>
  <name>cluster-1-template</name>
  <listen-port>8001</listen-port>
  <cluster>cluster-1</cluster>
  <network-access-point>
    <name>MyChannel</name>
    <protocol>t3</protocol>
    <public-address>some.public.address.com</public-address>
    <listen-port>7999</listen-port>
    <public-port>30999</public-port>
    <http-enabled-for-this-protocol>true</http-enabled-for-this-protocol>
    <tunneling-enabled>true</tunneling-enabled>
    <outbound-enabled>false</outbound-enabled>
    <enabled>true</enabled>
    <two-way-ssl-enabled>false</two-way-ssl-enabled>
    <client-certificate-enforced>false</client-certificate-enforced>
  </network-access-point>
</server-template>
<cluster>
  <name>cluster-1</name>
  <cluster-messaging-mode>unicast</cluster-messaging-mode>
  <dynamic-servers>
    <name>cluster-1</name>
    <server-template>cluster-1-template</server-template>
    <maximum-dynamic-server-count>5</maximum-dynamic-server-count>
    <calculated-listen-ports>false</calculated-listen-ports>
    <server-name-prefix>managed-server</server-name-prefix>
    <dynamic-cluster-size>5</dynamic-cluster-size>
    <max-dynamic-cluster-size>5</max-dynamic-cluster-size>
  </dynamic-servers>
</cluster>

And, here is a snippet of offline WLST code that corresponds to the above config.xml file snippet:

  templateName = "cluster-1-template"
  cd('/ServerTemplates/%s' % templateName)
  templateChannelName = "MyChannel"
  create(templateChannelName, 'NetworkAccessPoint')
  cd('NetworkAccessPoints/%s' % templateChannelName)
  set('Protocol', 't3')
  set('ListenPort', 7999)
  set('PublicPort', 30999)
  set('PublicAddress', 'some.public.address.com')
  set('HttpEnabledForThisProtocol', true)
  set('TunnelingEnabled', true)
  set('OutboundEnabled', false)
  set('Enabled', true)
  set('TwoWaySslEnabled', false)
  set('ClientCertificateEnforced', false)

In this example:

  • WebLogic binds the custom network channel to port 7999 and the default network channel to 8001.

  • The operator will automatically create a Kubernetes Service named DOMAIN_UID-cluster-cluster-1 for both the custom and default channel.

  • The operator will automatically set the ListenAddress on each WebLogic Server for each of its channels.

  • Internal clients running in the same Kubernetes cluster as the channel can access the cluster using t3://DOMAIN_UID-cluster-cluster-1:8001.

  • External clients would be expected to access the cluster using the custom channel with URLs like t3://some.public.address.com:30999 or, if using tunneling, http://some.public.address.com:30999.

WebLogic custom channel notes

  • Channel configuration for a configured cluster requires configuring the same network access point on each server. The operator currently doesn’t test or support network channels that have a different configuration on each server in the cluster.

  • Additional steps are required for external clients beyond configuring the custom channel - see Overview.

Setting up a NodePort

Getting started

A Kubernetes NodePort exposes a port on each worker node in the Kubernetes cluster (they are not typically exposed on masters), where the port is accessible from outside of a Kubernetes cluster. This port redirects network traffic to pods within the Kubernetes cluster. Setting up a Kubernetes NodePort is one approach for giving external WebLogic clients access to JMS or EJBs.

If an EJB or JMS service is running on an Administration Server, then you can skip the rest of this section and use the spec.adminServer.adminService.channels Domain field to have the operator create a NodePort for you. See Reference - Domain. Otherwise, if the EJB or JMS service is running in a WebLogic cluster or standalone WebLogic Server Managed Server, and you desire to provide access to the service using a NodePort, then the NodePort must be exposed ‘manually’ - see the following sample and table.

Setting up a NodePort usually also requires setting up a custom network channel. See Adding a WebLogic Custom Channel.

Sample NodePort resource

The following NodePort YAML file exposes an external node port of 30999 and internal port 7999 for a domain UID of DOMAIN_UID, a domain name of DOMAIN_NAME, and a cluster name of CLUSTER_NAME. It assumes that 7999 corresponds to a T3 protocol port of a channel that’s configured on your WebLogic cluster.

apiVersion: v1
kind: Service
metadata:
  namespace: default
  name: DOMAIN_UID-cluster-CLUSTER_NAME-ext
  labels:
    weblogic.domainUID: DOMAIN_UID
spec:
  type: NodePort
  externalTrafficPolicy: Cluster
  sessionAffinity: ClientIP
  selector:
    weblogic.domainUID: DOMAIN_UID
    weblogic.clusterName: CLUSTER_NAME
  ports:
  - name: myclustert3channel
    nodePort: 30999
    port: 7999
    protocol: TCP
    targetPort: 7999
Table of NodePort attributes
Attribute Description
metadata.name For this particular use case, the NodePort name can be arbitrary as long as it is DNS compatible. But, as a convention, it’s recommended to use DOMAIN_UID-cluster-CLUSTER_NAME-ext. To ensure the name is DNS compatible, use all lower case and convert any underscores (_) to dashes (-).
metadata.namespace Must match the namespace of your WebLogic cluster.
metadata.labels Optional. It’s helpful to set a weblogic.domainUid label so that cleanup scripts can locate all Kubernetes resources associated with a particular domain UID.
spec.type Must be NodePort.
spec.externalTrafficPolicy Set to Cluster for most use cases. This may lower performance, but ensures that a client that attaches to a node without any pods that match the spec.selector will be rerouted to a node with pods that do match. If set to Local, then connections to a particular node will route only to that node’s pods and will fail if the node doesn’t host any pods with the given spec.selector. It’s recommended for clients of a spec.externalTrafficPolicy: Local NodePort to use a URL that resolves to a list of all nodes, such as t3://mynode1,mynode2:30999, so that a client connect attempt will implicitly try mynode2 if mynode1 fails (alternatively, use a round-robin DNS address in place of mynode1,mynode2).
spec.sessionAffinity Set to ClientIP to ensure an HTTP tunneling connection always routes to the same pod, otherwise the connection may hang and fail.
spec.selector Specifies a weblogic.domainUID and weblogic.clusterName to associate the NodePort resource with your cluster’s pods. The operator automatically sets these labels on the WebLogic cluster pods that it deploys for you.
spec.ports.name This name is arbitrary.
spec.ports.nodePort The external port that clients will use. This must match the external port that’s configured on the WebLogic configured channels/network access points. By default, Kubernetes requires that this value range from 30000 to 32767.
spec.ports.port and spec.targetPort These must match the port that’s configured on the WebLogic configured channel/network access points.

Enabling unknown host access

When is it necessary to enable unknown host access?

If a source WebLogic Server attempts to initiate an EJB, JMS, or JTA connection with a target WebLogic Server, then the target WebLogic Server will reject the connection by default if it cannot find the source server’s listen address in its DNS. Such a failed connection attempt can yield log messages or exceptions like "...RJVM has already been shutdown..." or "...address was valid earlier, but now we get...".

This means that it’s usually necessary to enable unknown host access on an external WebLogic Server so that it can support EJB, JMS, or JTA communication that is initated by an operator hosted WebLogic Server. For example, if an operator hosted WebLogic Server with service address mydomainuid-myservername initiates a JMS connection to a remote WebLogic Server, then the remote server will implicitly attempt to lookup mydomainuid-myservername in its DNS as part of the connection setup, and this lookup will typically fail.

Similarly, this also means that it’s necessary to enable unknown host access on an operator hosted WebLogic Server that accepts EJB or JMS connection requests from external WebLogic Servers when the external WebLogic Server’s listen addresses cannot be resolved by the DNS running in the Kuberneters cluster.

How to enable unknown host access

To enable an ‘unknown host’ source WebLogic Server to initiate EJB, JMS, or JTA communication with a target WebLogic Server:

  • Set the weblogic.rjvm.allowUnknownHost Java system property to true on each target WebLogic Server.
    • For operator hosted WebLogic Servers, you can set this property by including -Dweblogic.rjvm.allowUnknownHost=true in the JAVA_OPTIONS Domain environment variable defined in the domain resource’s spec.serverPod.env attribute.
  • Also apply patch 30656708 on each target WebLogic Server for versions 12.2.1.4 (PS4) or earlier.

Security notes

  • With some cloud providers, a load balancer or NodePort may implicitly expose a port to the public Internet.

  • If such a port supports a protocol suitable for WebLogic clients, note that WebLogic allows access to JNDI entries, EJB/RMI applications, and JMS by anonymous users by default.

  • You can configure a custom channel with a secure protocol and two-way SSL to help prevent external access by unwanted clients. See When is a WebLogic custom channel needed?.

Optional reading