HP (Hewlett-Packard) 9000 V2600 SCA ユーザーズマニュアル

Recovering from failures

The on-disk and file system formats of a crash dump have changed with 
HP-UX 11.0. 

libcrash

(3) is a new library provided to allow programmatic access to a 

crash dump. supports all past and current crash dump formats. By using 

libcrash

(3) under certain configurations, crash dumps no longer need 

to be copied into the file system before they can be debugged. See the 

libcrash

(3) man page for more information.

When the system crashes, HP-UX saves the image of physical memory or 
certain portions of it to predefined locations called dump devices. When 
the operator next reboots the system, a special utility copies the memory 
image from the dump devices to the HP-UX file system area. Once 
copied, the memory image can be analyzed with a debugger or saved to 
tape for later analysis.

Prior to HP-UX 11.0, dump devices had to be defined in the kernel 
configuration, and they still can be using Release 11.0. Beginning with 
Release 11.0, however, a new more-flexible method for defining dump 
devices is available using 

crashconf

.

Beginning with HP-UX Release 11.0, there are three places where dump 
devices are configured:

1. In the kernel (same as releases prior to Release 11.0)

2. During system initialization when the initialization script for 

crashconf

 runs (and reads entries from the 

/etc/fstab

 file)

3. During runtime, by the operator or administrator manually running 

the 

/sbin/crashconf

 command. 

Defining the contents and destination of the crash dump are two 
important factors to consider when preparing for the dump. The 
destination and contents are configurable without rebooting, using the 

crashconf

 interface. See the crashconf(1M) man page for more 

information.

In order to capture the memory image of the system when a crash occurs, 
the image storage location(s) must be defined in advance.