Symmetrical Sessions

Symmetrical sessions happen with dumb load balancers like DNS round-robin. A single session may bounce from machine A to machine B and back to machine B. For JDBC sessions, the symmetrical session case needs the always-load-session attribute described below. Each request must load the most up-to-date version of the session.

Distributed sessions in a symmetrical environment are required to make sessions work at all. Otherwise the state will end up spread across the JVMs. However, because each request must update its session information, it is less efficient than sticky sessions.

Sticky Sessions

Sticky sessions require more intelligence on the load-balancer, but are easier for the JVM. Once a session starts, the load-balancer will always send it to the same JVM. Resin's load balancing, for example, encodes the session id as 'aaaXXX' and 'baaXXX'. The 'aaa' session will always go to JVM-a and 'baa' will always go to JVM-b.

Distributed sessions with a sticky session environment add reliability. If JVM-a goes down, JVM-b can pick up the session without the user noticing any change. In addition, distributed sticky sessions are more efficient. The distributor only needs to update sessions when they change. So if you update the session once when the user logs in, the distributed sessions can be very efficient.

always-load-session

Symmetrical sessions must use the 'always-load-session' flag to update each session data on each request. always-load-session is only needed for jdbc-store sessions. tcp-store sessions use a more-sophisticated protocol that eliminates the need for always-load-session, so tcp-store ignores the always-load-session flag.

The always-load-session attribute forces sessions to check the store for each request. By default, sessions are only loaded from persistent store when they are created. In a configuration with multiple symmetric web servers, sessions can be loaded on each request to ensure consistency.

always-save-session

By default, Resin only saves session data when you add new values to the session object, i.e. if the request calls setAttribute. This may be insufficient when storing large objects. For example, if you change an internal field of a large object, Resin will not automatically detect that change and will not save the session object.

With always-save-session Resin will always write the session to the store at the end of each request. Although this is less efficient, it guarantees that updates will get stored in the backup after each request.

always-save-session

Resin's distributed sessions needs to know when a session has changed in order to save the new session value. Although Resin can detect when an application calls HttpSession.setAttribute, it can't tell if an internal session value has changed. The following Counter class shows the issue:

package test;

public class Counter implements java.io.Serializable {
  private int _count;

  public int nextCount() { return _count++; }
}

Assuming a copy of the Counter is saved as a session attribute, Resin doesn't know if the application has called nextCount. If it can't detect a change, Resin will not backup the new session, unless always-save-session is set. When always-save-session is true, Resin will back up the session on every request.

...
<web-app id="/foo">
...
<session-config>
  <use-persistent-store/>
  <always-save-session/>
</session-config>
...
</web-app>

Serialization

Resin's distributed sessions relies on Hessian serialization to save and restore sessions. Application object must implement java.io.Serializable for distributed sessions to work.

Session Request

To see how cluster sessions work, consider a case where the load balancer sends the request to a random host. Server C owns the session but the load balancer gives the request to Server A. In the following figure, the request modifies the session so it must be saved as well as loaded.

The session id encodes the owning host. The example session id, ca8MbyA, decodes to an server index of 3, mapping to Server C. Resin determines the backup host from the cookie as well. Server A must know the owning host for every cookie so it can communicate with the owning srun. The example configuration defines all the sruns Server A needs to know about. If Server C is unavailable, Server A can use its configuration knowledge to use Server D as a backup for ca8MbyA instead..

When the request first accesses the session, Server A asks Server C for the serialized session data (2:load). Since Server A doesn't cache the session data, it must ask Server C for an update on each request. For requests that only read the session, this TCP load is the only extra overhead, i.e. they can skip 3-5. The always-save-session flag, in contrast, will always force a write.

At the end of the request, Server A writes any session updates to Server C (3:store). If always-save-session is false and the session doesn't change, this step can be skipped. Server A sends the new serialized session contents to Server C. Server C saves the session on its local disk (4:save) and saves a backup to Server D (5:backup).

Sticky Session Request

Smart load balancers that implement sticky sessions can improve cluster performance. In the previous request, Resin's cluster sessions maintain consistency for dumb load balancers or twisted clients like the AOL browsers. The cost is the additional network traffic for 2:load and 3:store. Smart load-balancers can avoid the network traffic of 2 and 3.

Server C decodes the session id, caaMbyA. Since it owns the session, Server C gives the session to the servlet with no work and no network traffic. For a read-only request, there's zero overhead for cluster sessions. So even a semi-intelligent load balancer will gain a performance advantage. Normal browsers will have zero overhead, and bogus AOL browsers will have the non-sticky session overhead.

A session write saves the new serialized session to disk (2:save) and to Server D (3:backup). always-save-session will determine if Resin can take advantage of read-only sessions or must save the session on each request.

Disk copy

Resin stores a disk copy of the session information, in the location specified by the path. The disk copy serves two purposes. The first is that it allows Resin to keep session information for a large number of sessions. An efficient memory cache keeps the most active sessions in memory and the disk holds all of the sessions without requiring large amounts of memory. The second purpose of the disk copy is that the sessions are recovered from disk when the server is restarted.

Failover

Since the session always has a current copy on two servers, the load balancer can direct requests to the next server in the ring. The backup server is always ready to take control. The failover will succeed even for dumb load balancers, as in the non-sticky-session case, because the srun hosts will use the backup as the new owning server.

In the example, either Server C or Server D can stop and the sessions will use the backup. Of course, the failover will work for scheduled downtime as well as server crashes. A site could upgrade one server at a time with no observable downtime.

Recovery

When Server C restarts, possibly with an upgraded version of Resin, it needs to use the most up-to-date version of the session; its file-saved session will probably be obsolete. When a "new" session arrives, Server C loads the saved session from both the file and from Server D. It will use the newest session as the current value. Once it's loaded the "new" session, it will remain consistent as if the server had never stopped.

No Distributed Locking

Resin's cluster sessions does not lock sessions. For browser-based sessions, only one request will execute at a time. Since browser sessions have no concurrently, there's no need for distributed locking. However, it's a good idea to be aware of the lack of distributed locking.