Split Server Configurations
If you configure the master Security server and the CDS initial server on different machines, the cell is said to have a "split server" configuration. If you choose a split server configuration, you must take some extra steps in the initial cell configuration phase. This is because every DCE machine must be configured as a DCE client, including machines that are also configured as some kind of DCE server machine. However, a DCE client configuration cannot succeed unless a CDS server is available. Therefore, the initial Security server machine cannot immediately be configured as a DCE client, but must wait until a CDS server machine is configured.
This is not a problem when the master Security server and CDS server are configured on one machine. However, in a split server configuration, you must first configure the Security server, then configure the CDS server on another machine, and then come back to the Security server machine and configure it as a DCE client or as an additional server. The following steps detail these actions:
· Configure Machine 1 as a Security server machine.
· Go to Machine 2 and configure it as a CDS server machine.
· Go back to Machine 1 and configure it as a DCE client.
Note that if you configure the Security server machine as some other kind of server machine (for example, as a DTS server), then it will automatically be configured as a DCE client. So the following scenario also works:
· Configure Machine 1 as a Security server machine. The dce_config script will prompt you with the clock synchronization question discussed in the following topic. When it does this, answer n.
· Go to Machine 2 and configure it as a CDS server machine.
· Go back to Machine 1 and configure it as a DTS server.
In this scenario, Step 3 causes Machine 1 to be configured as a DCE client as well as a DTS server.
Consult the OSF DCE Release Notes for the current information on timing and sequence constraints when bringing up split server cells.