Versions available for this page: CUBRID 8.4.3 | CUBRID 9.0.0 |
Currently, CUBRID HA is supported by Linux only. All nodes within CUBRID HA groups must be configured on the same platforms.
CUBRID HA synchronizes data among nodes with the following method (as known as transaction log shipping): It replicates the primary key-based replication logs generated from the server of a master node to a slave node and then reflects the replication logs to the slave node.
If data of the specific table within CUBRID HA groups is not synchronized, you should check whether the appropriate primary key has specified for the table.
On the partitioned table, the table which has promoted some partitions by the PROMOTE statement replicates all data to the slave. However, since the table does not have the primary key, the data changes on the table made by the master are not applied to the slave.
Using triggers and java stored procedures in CUBRID HA can cause duplicate executions because triggers and java stored procedures executed in a master node will be executed in a slave node again. This may cause data inconsistency among nodes within CUBRID HA groups.
It is not recommended to use triggers and java stored procedures in CUBRID HA.
CUBRID HA synchronizes data among nodes within CUBRID HA groups based on replication logs. Therefore, using method that does not generate replication logs or configuring NOT NULL through CUBRID Manager may cause data inconsistency among nodes within CUBRID HA groups. Therefore, it is not recommended to use method and other menus in CUBRID HA environment except for the query processor CUBRID Manager.
The UPDATE STATISTICS statement which updates statistics does not replicate in the slave node.
The replication logs are not generated as for tasks performed in standalone mode. For this reason, data inconsistency among nodes within CUBRID HA groups may occur when performing tasks in standalone mode.
To enhance performance, a serial cache does not access Heap and does not generate replication logs when retrieving or updating serial information. Therefore, if you use a serial cache, the current values of serial caches will be inconsistent among the nodes within CUBRID HA groups.
This command is used to back up a specified database. If the -r option is used, logs that are not required for recovery will be deleted. This deletion may result in data inconsistency among nodes within CUBRID HA groups. Therefore, you must not use the -r option.
If you use INCR/DECR (click counter functions) in a slave node of HA configuration, an error is returned.
In a CUBRID HA environment, the meta data (Locator) of a LOB column is replicated and LOB data is not replicated. Therefore, if storage of a LOB type is located on the local machine, no tasks corresponding to columns are allowed in slave nodes or master nodes after failover.