Replication Server stores messages on disk to make sure that they can be delivered following a failure. When you install a Replication Server, you allocate an initial disk partition that Replication Server uses for its disk storage. You can add additional partitions when you have finished installing the Replication Server.
The partition is either a raw disk device or operating system file. Because UNIX operating systems buffer file I/O, you may not be able to completely recover data following a failure. On such a system, use operating system files for partitions only in a test environment. Use raw disk partitions for production environments. See the Replication Server Reference Manual for more information about adding partitions.
Replication Server allocates stable queues from its disk partitions for the routes and connections it serves. Messages are saved in the stable queues at least until the messages are confirmed as received at their destination.
The amount of disk space you should allocate for Replication Server partitions depends on the size of transactions and the transaction rate for your application. Stable queues act as buffers for data as it flows through your replication system. If a remote site’s Replication Server is unreachable during a network failure, the primary Replication Server stores transactions in a stable queue until communication is restored. The more space allocated for disk partitions, the longer the Replication Server can queue data without interrupting operations in the primary database.
Appendix A, “Capacity Planning” explains in detail how to calculate partition space for a Replication Server.
