Why did VERITAS Cluster Server (VCS) not initiate a failover when the system was shut down?
Some important notes to remember while testing VCS failovers with versions before VCS 1.3.0.
A machine shutdown is *not* viewed as a failure and will not initiate a failover. This is true for all "init", "halt", "reboot", "shutdown" and "uadmin 2 0" commands. These commands are viewed as regular administrative operations and not as failures so no other node in the cluster will attempt to takeover the applications. To make sure that the applications are always available, the applications must be manually switched over prior to shutting down the node.
Alternatively, the command "/sbin/gabsync" can be issued to start off a 2 minute timer during which all shutdowns will be treated as failures. The applications will be failed over successfully should a shutdown be initiated in this 2 minute window.
The best way to test if the applications are indeed failing over is to issue a Stop-A on the node, or to kill the application process through a kill -9 command.
Starting with VCS 1.3.0 administrative shutdowns will cause a failover.
Imported Document Id
This is machine translated content
Login to Subscribe
Please login to set up your subscription.
Didn't find the article you were looking for? Try these resources.