High Availability

ThinLinc has built-in functionality for obtaining High Availability for environments where the ThinLinc cluster must be available at all times.

Avoiding Single Point of Failure

A ThinLinc cluster has two main components - the VSM Server and one or several VSM Agents. The VSM Server is the central component responsible for keeping track of which sessions are currently running to allow users to reconnect to their existing sessions. It is also responsible for selecting which VSM Agent should start up a new session when a new session is requested by a user, based on load balance information.

This means that the VSM Server is a critical component in the ThinLinc cluster. It needs to be available at all times for users to be able to start new sessions or reconnect to old ones.

The VSM Server is however only critical for new session startup and reconnect. The client has no communication with the VSM Server once the client has received information about the session, which means that once the session is up and running, there is no interruption to the user owning the session should the VSM Server go down. This design greatly increases the stability of the system, since there are very few components that need to be up at all times to keep a session running.

Still, to ensure a high availability on the ability to start new, and reconnect to existing, sessions, the availability of the VSM Server must be ensured. This means that a hardware or operating system failure on the server running VSM Server must not interrupt the service, or the VSM Server will become a Single Point of Failure.

Built-in High Availability

The VSM Server comes with built-in High Availability functionality at no extra cost. The server has functionality for sharing information about the current sessions with a second node. Combined with the industry-standard Open Source Heartbeat software, this means that two physical servers can share the responsibility for keeping the VSM Server service available at all times.

If one node goes down, Heartbeat will transfer responsibility to the other node which will immediately start serving clients. Reconnects to existing sessions will work as before the switchover, since the second node is kept up to date with session information at all time.

When the first node comes up again, it will be updated with session information by the second node, and will then start serving sessions again when Heartbeat detects that it's up and running.

No Extra Hardware Requirements

The extra VSM Server can be run on a server that is also serving as VSM Agent. This means that if the primary server goes down, there will be a performance decrease when starting new and reconnecting to old sessions, but if this is acceptable, a setup of this type gives High Availability without the need for an extra server.

High Session Server Availability

With the VSM Server availability secured, the focus changes to the availability of servers running VSM Agent. The VSM Agent servers run the user sessions, which means that all programs run by the user are running on this server. If a VSM Agent server goes down, all sessions on this server will be lost.

A loss of a VSM Agent server is however no disaster if there are another VSM Agent server available. If a user is disconnected because the VSM Agent server her session was running on goes down, she can start a new session which will be started on another VSM Agent server since the load balancer in the VSM Server will detect that the previous VSM Agent is no longer available.

To summarize; Make sure there is more than one server running VSM Agent if you want high availability in your cluster.

More information about the High Availability features of ThinLinc can be found in the ThinLinc Administrator's Guide.