Most UNIX operating systems provide a method to perform a "break" or "console abort." The inherent problem when you abort a hung system is that it ceases to heartbeat in the cluster. Other cluster members may begin corrective action when they believe that the aborted node is really a failed node.
In order to preserve data integrity and to prevent the cluster from taking additional corrective actions, it is critical that the only action that you perform following an abort is to reset the system in question. Do not resume the processor as cluster membership may have changed and failover actions may already be in progress.
To remove this potential problem on Sun SPARC systems, you should alias the go
function in the OpenBoot eeprom to display a message.
To alias the go function to display a message
." Aliasing the OpenBoot 'go' command! "
: go ." It is inadvisable to use the 'go' command in a clustered environment. " cr
." Please use the 'power-off' or 'reset-all' commands instead. " cr
nvrun
command. You should see only the following text:
nvstore
command to commit your changes to the non-volatile RAM (NVRAM) for use in subsequent reboots.