Client connections can be either HA connections or non-HA connections. Regardless, client connections require that:
You specify large enough values for the retry and timeout options in the interfaces file. When you determine these values, allow for failover delays, such as starting Adaptive Server on the secondary node, recovery time, and for multiple node failures. Note that the Adaptive Server resource group tries to fail over until it finds a secondary node that can successfully host the resource group.
The logical hostname is accessible form the client machine.
The version of ctlib that is shipped
with Adaptive Server version 12.5.0.3 or later contains a corrected
implementation of the retry and timeout features. Re-link
your applications to take advantage of the retry and timeout features.
Non-HA connections are regular connections that do not include either the hafailover entry in the interfaces file or the CS_HAFAILOVER property set in the client connection. When these connections are lost, clients must be able to reconnect to the Adaptive Server after failure. To re-establish the connections, clients must retry enough times, or wait long enough between retries, until failover completes and Adaptive Server is running on the secondary node.
To reconnect to the server, clients can use the retry and timeout options in the interfaces file or the corresponding connection properties. In the following interfaces file example, the retry count is 10 and the timeout delay between each retry is 20 seconds:
MONEY 10 20 master tcp ether loghost 4010 query tcp ether loghost 4010
HA connections are made with:
The CS_HAFAILOVER property set at the connection or context level (equivalent to the -Q option of isql).
The hafailover entry in the interfaces file, which must point to the Adaptive Server entry to be contacted in case of failover.
In an active-passive configuration, clients must self-reference the hafailover entry because they reconnect to the same Adaptive Server after failover. That is, they must set the same server name as the hafailover server in the interfaces file because the same Adaptive Server is rebooted on the secondary node.
For example, the Adaptive Server entry in the example above can be self- referenced as:
MONEY 10 20 master tli tcp loghost 4010 query tli tcp loghost 4010 hafailover MONEY
For more information about configuring client connections with the failover property, see Appendix C, “Open Client Functionality in a Failover Configuration.”.
