HP-UX 11i Version 1.5 System Crash Dump

Post-Reboot Dump Processing

Chapter 1 17

Post-Reboot Dump Processing

Prior to HP-UX 11.0, one of the ﬁrst user space tasks in the boot process

was running savecore (now obsolete), which would check for a crash

dump and, if found, copy it from the dump device(s) to the ﬁle system.

This same program had many other features used by system

administrators after the system was up.

In HP-UX release 11.0 and later releases, savecrash(1M) and

crashutil(1M) handle boot-time tasks and post-boot tasks, respectively.

The savecrash(1M) command runs at boot time; it preserves any and all

crash-related information which could be lost as system activity

continues. It has many options, which may be set in

/etc/rc.config.d/savecrash. One thing worth noting is that

savecrash can be told to save only that portion of the crash dump

endangered by swap activity, rather than all of it (the default). This is

useful for debugging on the system that crashed: the debuggers can

debug straight out of the dump devices. (See “Debugging of Crash

Dumps” on page 18.) If care has been taken to conﬁgure dump devices

that are not shared by swap activity (see “CrashDump Conﬁguration” on

page 13), this means that virtually all of the time it would normally take

to copy the dump onto the ﬁle system can be eliminated.

The crashutil(1M) command handles all of the post-boot crash dump

manipulation. Currently it has two main functions. First, it completes

saving those portions of the dump that reside on non-swap dump devices.

The crashutil command copies the remaining portions of the dump

onto ﬁle system so that it can be transported to another system for

debugging. Second, it converts dumps from one dump format to another.

For example, an 11.0 dump can be converted to an older format so that

old debugging tools can access the dump.

Note: The concept of a special tape format for dumps has gone away. If

dumps are needed on tape:

• run crashutil(1M) (to make sure that parts of it aren’t still residing

on dump devices)

• save the dump from the dump device to a directory on a ﬁle system

(typically /var/adm/crash)

• use tar or any similar command to package and ship the crash dump