HP (Hewlett-Packard) 9000 V2500 SCA ユーザーズマニュアル

Recovering from failures

/etc/rc.config.d/savecrash

) reduces system recovery time. After 

the system recovery, run 

savecrash

 manually to copy the memory 

image from the dump area to the HP-UX file system area.

If a memory dump resides partially on dedicated dump devices and 
partially on devices that are also used for paging, only those pages that 
are endangered by paging activity can be saved. 

Pages residing on the dedicated dump devices can remain there. It is 
possible to analyze memory dumps directly from the dedicated dump 
devices using a debugger that supports this feature. If, however, there is 
a need to send the memory dump to someone else for analysis, move the 
pages on the dedicated dump devices to the HP-UX file system area. 
Then use a utility such as 

tar

 to bundle them for shipment. To do that, 

use the command 

/usr/sbin/crashutil

 instead of 

savecrash

 to 

complete the copy.

This section discusses how to make sure the part of memory that 
contains the instruction or piece of data that caused the crash is 
captured. The factors that must be considered are:

• Full dump vs. selective dump

• Dump definitions built into the kernel vs. defined at runtime

• Using a device for both paging and as a dump device

The only way to guarantee capturing the specific instruction or data that 
caused the crash is to capture everything. This means selecting a full 
dump of memory. 

Be aware, however, that this can be costly in terms of time and disk 
space. A large amount of time and disk space is needed to dump the 
entire contents of memory in a system with a large memory 
configuration or to copy a large memory image to the HP-UX file system 
area during the reboot process.

The amount of dump area should at least be equal to the amount of 
memory in the system; depending on a number of factors, additional disk 
space greater than the amount of physical memory in the system may be 
needed