Permalink
Fetching contributors…
Cannot retrieve contributors at this time
70 lines (55 sloc) 4.81 KB

Sticky sessions

Typical cluster deployment consists of the load balancer (reverse proxy) and 2 or more {project_name} servers on private network. For performance purposes, it may be useful if load balancer forwards all requests related to particular browser session to the same {project_name} backend node.

The reason is, that {project_name} is using Infinispan distributed cache under the covers for save data related to current authentication session and user session. The Infinispan distributed caches are configured with one owner by default. That means that particular session is saved just on one cluster node and the other nodes need to lookup the session remotely if they want to access it.

For example if authentication session with ID 123 is saved in the Infinispan cache on node1, and then node2 needs to lookup this session, it needs to send the request to node1 over the network to return the particular session entity.

It is beneficial if particular session entity is always available locally, which can be done with the help of sticky sessions. The workflow in the cluster environment with the public frontend load balancer and two backend {project_name} nodes can be like this:

  • User sends initial request to see the {project_name} login screen

  • This request is served by the frontend load balancer, which forwards it to some random node (eg. node1). Strictly said, the node doesn’t need to be random, but can be chosen according to some other criterias (client IP address etc). It all depends on the implementation and configuration of underlying load balancer (reverse proxy).

  • {project_name} creates authentication session with random ID (eg. 123) and saves it to the Infinispan cache.

  • Infinispan distributed cache assigns the primary owner of the session based on the hash of session ID. See Infinispan documentation for more details around this. Let’s assume that Infinispan assigned node2 to be the owner of this session.

  • {project_name} creates the cookie AUTH_SESSION_ID with the format like <session-id>.<owner-node-id> . In our example case, it will be 123.node2 .

  • Response is returned to the user with the {project_name} login screen and the AUTH_SESSION_ID cookie in the browser

From this point, it is beneficial if load balancer forwards all the next requests to the node2 as this is the node, who is owner of the authentication session with ID 123 and hence Infinispan can lookup this session locally. After authentication is finished, the authentication session is converted to user session, which will be also saved on node2 because it has same ID 123 .

The sticky session is not mandatory for the cluster setup, however it is good for performance for the reasons mentioned above. You need to configure your loadbalancer to sticky over the AUTH_SESSION_ID cookie. How exactly do this is dependent on your loadbalancer.

It is recommended on the {project_name} side to use the system property jboss.node.name during startup, with the value corresponding to the name of your route. For example, -Djboss.node.name=node1 will use node1 to identify the route. This route will be used by Infinispan caches and will be attached to the AUTH_SESSION_ID cookie when the node is the owner of the particular key. Here is an example of the start up command using this system property:

cd $RHSSO_NODE1
./standalone.sh -c standalone-ha.xml -Djboss.socket.binding.port-offset=100 -Djboss.node.name=node1

Typically in production environment the route name should use the same name as your backend host, but it is not required. You can use a different route name. For example, if you want to hide the host name of your {project_name} server inside your private network.

Disable adding the route

Some load balancers can be configured to add the route information by themselves instead of relying on the back end {project_name} node. However, as described above, adding the route by the {project_name} is recommended. This is because when done this way performance improves, since {project_name} is aware of the entity that is the owner of particular session and can route to that node, which is not necessarily the local node.

You are permitted to disable adding route information to the AUTH_SESSION_ID cookie by {project_name}, if you prefer, by adding the following into your RHSSO_HOME/standalone/configuration/standalone-ha.xml file in the {project_name} subsystem configuration:

<subsystem xmlns="urn:jboss:domain:keycloak-server:1.1">
  ...
    <spi name="stickySessionEncoder">
        <provider name="infinispan" enabled="true">
            <properties>
                <property name="shouldAttachRoute" value="false"/>
            </properties>
        </provider>
    </spi>

</subsystem>