Failover with ISV cluster management software

You can use independent software vendor (ISV) cluster management software instead of the Connection Manager to manage failover processing in high-availability cluster environments.

If the primary server in a high-availability cluster encounters a problem that requires a secondary server to assume the role of the primary, it is important that, before performing the actual failover, disk I/O is prohibited on the failed primary server and is allowed on the new primary server. In addition, network access to the failed primary server must be prevented. This is especially true for SD secondary servers, where disk corruption can occur if these steps are not done correctly.

The mechanism for enabling disk I/O operations from a server in a high-availability cluster environment is known as I/O Fencing. I/O Fencing is configured using a callback script. When a failure of the primary server occurs, the failover process executes a callback script on the secondary server before the secondary server assumes the role of the primary server. The script calls any I/O specific commands to enable or disable disk access. The script enables write access to the shared disk on the server that is to become the primary server, and disables write access to the shared disk on the failed server.

Use the FAILOVER_CALLBACK configuration parameter to specify the name of the script to run when a database server transitions from a secondary server to a primary server, or from a secondary server to a standard server. A template script named ifx_failover_callback.sh (UNIX™) or ifx_failover_callback.bat (Windows™) is provided in the $ONEDB_HOME/etc directory. When configured, the script specified by FAILOVER_CALLBACK is executed before the secondary server is switched to a primary or standard server.

You can test the failover script by performing one of the following actions, depending on your type of high-availability cluster:
  • Converting an SD secondary server into a primary server.
  • If the DRAUTO configuration parameter is set to 0, shutting down the primary server and convert the HDR secondary server to standard mode.
  • If the DRAUTO configuration parameter is set to 1, shutting down the primary server in an HDR pair.
  • Shutting down the primary server in a remote stand-alone cluster and converting the RS secondary server to standard mode.

An Invoking Failover Callback message is in the online.log listing the path and file name of the failover script after it is run.

See the information about the FAILOVER_CALLBACK configuration parameter in the HCL OneDB™ Administrator's Reference.

If the script specified by FAILOVER_CALLBACK fails (that is, if it returns a non-zero exit code), the failover of the secondary to the primary (or standard) server also fails. In this case, the DBA must manually perform the failover.