Automatic switchover without a reliable network

Although automatic switchover might seem to be the best solution, it is not appropriate for all environments.

Consider what might happen if the primary database server does not actually fail, but the secondary database server registers that the primary has failed. For example, if the secondary database server does not receive responses when it signals (pings) the primary database server because of a slow or unstable network, the secondary server assumes that the primary database server failed and switches automatically to the standard type. If the primary database server also does not receive responses when it signals the secondary database server, it assumes that the secondary database server failed and turns off data replication but remains in online mode. Now the primary database server and the secondary database server (switched to the standard type) are both in online mode.

If clients can update the data on both database servers independently, the database servers in the pair reach a state in which each database server has the logical-log records that are required by the other. In this situation, you must start again and perform initial data replication with a level-0 dbspace backup of one entire database server, as described in Starting HDR for the First Time. Therefore, if your network is not entirely stable, you might not want to use automatic switchover. HDR cannot be reinstated without the risk of losing transactions on the previous secondary server.


Copyright© 2019 HCL Technologies Limited