In Symfoware Server, the following failover operation modes are provided in each system:
Shared disk mode
Log shipping mode
This operation mode requires integration with the PRIMECLUSTER cluster software, to use the shared disk (GDS) it provides. A server in the cluster system receives the application processing, acting as the active server (active node).
For example, if the active node fails, another server (standby node) inherits the shared disk content and becomes the active node. Using the PRIMECLUSTER feature, the applications to be run on the active node for the database can also be linked with the database and switched. There is also no need to send the updated content of the database on the active node to the standby node, so the processing performance will be the same as of a non-cluster system.
Note
When integrating PRIMECLUSTER with Symfoware Server, the disk is not referenced by the standby node, and therefore the shared disk mode cannot be used for load distribution.
This operation mode is based on PostgreSQL streaming replication. Other software such as cluster software is not required.
This mode replicates the database on all servers that comprise the cluster system. It achieves this by sending and reflecting the updated transaction log of the database from the server that receives the update (primary server) to another server (standby server).
In addition, the client driver automatically distinguishes between the primary and standby servers, so applications can be connected transparently regardless of the physical server. Accordingly, a flexible system can easily be built, for example a system that offloads connections to the standby server if the connections only perform referencing.
Note
Symfoware Server Enterprise Edition and Standard Edition support only configurations comprising one standby server per primary server.
Log shipping mode has the following three features:
Streaming replication feature
The PostgreSQL streaming replication feature can be used as is.
The database will be copied in asynchronous mode, that is, the transaction log will be asynchronously sent to the standby server. This will result in performance equivalent to that of a non-cluster system.
In this case, note that, strictly speaking, a slight delay will occur in the copying of the database of the standby server.
Database multiplexing feature
Consists of a feature that detects faults in elements that are essential for the continuity of the database operation (such as the database process, disk, and network), as well as a switchover feature to manually switch the primary server and a disconnection feature for the standby server. Note that SQL statements cannot be run on the standby server.
The database will be copied in synchronous mode.
Mirroring Controller option
This option provides failover functionality to automatically switch and disconnect the primary server when the database multiplexing feature is used. Reference type SQL statements can also be run on the standby server.
See
Refer to "Part 2 Shared Disk Type Failover Mode" for information on the shared disk mode.
The streaming replication feature is not described in this manual.
Refer to "High Availability, Load Balancing, and Replication" in the PostgreSQL Documentation for information on the streaming replication feature.
Refer to "Part 3 Database Multiplexing Mode" for information on the database multiplexing feature.
Also refer to "Part 3 Database Multiplexing Mode" when using the Mirroring Controller option.
Additionally, refer to "When the Mirroring Controller option is used" in the various chapters and sections for the description and procedure required specifically for the Mirroring Controller option.
