Installation guide

Saving Memory Dumps
HP-UX version 11.00.03 Administering Fault Tolerant Hardware 5-51
Saving a Dump After a System Hang
Using save_mcore from an offline CPU, you can create a core dump of the
operating system after a system hang. The Continuum system can be configured
to reboot in simplexed state after a system crash or hang (that is, one
CPU/memory module is kept offline, with its memory contents intact). You can
then obtain the dump from the offline module. If the dump is successfully
retrieved, the system will be reduplexed. If the dump is not successful, the system
will remain simplexed. You can then force the system to duplexed state by using
the ftsmaint sync command.
The following conditions must exist before save_mcore can be used to save a
dump:
■ The system is configured to use save_mcore as the default dump method.
■ All CPU/memory boards should be duplexed at the time of crash or hang.
■ The system must have been rebooted without incurring a power loss.
■ There must be sufficient space to hold the dump files.
For more information, see the adb(1). crashutil(1M), savecrash(1M) man pages.
To use save_mcore in the event of a system hang, enter
hpmc_reset
The system will start in a simplexed state. The system startup script should detect
the offline CPU/memory module and invoke save_mcore to save the dump.
Analyzing the Dumps
If you know how to analyze memory dumps, you can use a debugger to analyze
the dumps A normal crash dump contains context and other state information that
was saved when the system panicked. The save_mcore utility saves this
information special records and makes it directly available to the standard HP
debugging tools (q4, adb) because the core dump generated by save_mcore is
the same in format as what is produced by the savecrash(1M) utility, and you can
use the same debugging tools to analyze the dump.