Client Reconnection Behavior
Each time that a client connects to the ICE Server, the system notifies the client of alternate points of ingress that it may use to connect to the system the next time a connection needs to be made. Clients 'save' this information to persistent storage so that once a client successfully connects to a system it can utilize these alternate sites or addresses even after logging out or quitting the application.
How the client goes about choosing which ingress address to connect to is configurable in the system’s 'High Availability' settings (on the 'Organization' settings screen in ICE Desktop). In environments where no server ingress addresses have been configured, clients will always use the 'Identity' reconnection strategy.
Reconnection Strategy | Description |
Preferred | The client will choose from the list of available ingress addresses in the order they appear (top-to-bottom) in the setting screen. |
Nearest | The client will choose from the list of available ingress addresses in the order of their physical distance from the client (calculated using the location assigned to the ingress in the Organization settings screen, and the reported location of the client). |
Identity | Default. The client will reconnect only to the server ingress address entered by the user in the login screen. All other ingress addresses specified in the Organization screen are ignored. This choice has the effect of disallowing clients to connect to any server address other than the one explicitly entered on the login screen. |
Random | The client will randomly choose from the list of available ingress addresses. |
As an example, consider a system deployed in the United States with three ingress addresses: west.demo.com, central.demo.com and east.demo.com: When the connection strategy is set to 'Preferred' clients will try to connect in the order: west, central, east, west, etc. When the strategy is 'Nearest', a client in Chicago will connect to central, then east, then west whereas a client in Los Angeles will connect to west then central, then east. When the strategy is 'Identity' all clients will ignore the configuration made in the 'High Availability' section of 'Organization' settings and connect only to the server the user entered in the login screen.
Clients connections to the ICE Server are 'sticky.' Clients do not try to reconnect to their original ingress or to a more ideal ingress once a successful connection is made. Once a client establishes a connection to an ICE Server, it maintains that connection until the user quits, logs out, or the connection is broken.
Note: Each client maintains both a single connection to the ICE Server, plus a connection to every Rallypoint needed by all active channels. Loss of ICE Server connectivity does not necessary imply loss of Rallypoint connectivity. Channels have their own Rallypoint reconnect and failover strategy (described in detail later in this document).