Compaq ProLiant Server NMI Crash Dump Feature
3
11N8-0100A-WWEN
What Is NMI Crash Dump
Non-Maskable Interrupt (NMI) Crash Dump is a diagnostic mechanism. It allows for crash dump
files to be created in situations when a system is hung and not able to respond to traditional debug
mechanisms.
Crash dump analysis is an essential part of diagnosing reliability problems such as hangs in
operating systems, device drivers, and applications. Many crashes will freeze a system in such a
way that your only recourse is to do a hard reset-(i.e. power cycle the system). Since resetting the
system erases any information that would support an analysis of the problem, it is desirable for
the system to be able to perform a memory dump before a hard reset is performed. A dump
switch and the associated support in Windows NT, Windows 2000, NetWare, SCO OpenServer 5,
and SCO UnixWare 7 provide this function.
Figure 1: NMI Crash Dump Issue Resolution Events
The dump switch can be used to diagnose software failures by forcing the operating system to
invoke the non-maskable interrupt (NMI) handler and generate a crash dump log. The crash dump
log can provide critical information for root-cause analysis that may be difficult or impossible to
obtain through other means
A user initiates a Non Maskable Interrupt (NMI) event by pressing the dump switch. The NMI
can allow a hung system to become responsive enough to generate a crash dump log.
WARNING:
Using the NMI Crash Dump switch on a functioning system (using any operating
system) will cause the unit to abruptly fail. This is the designed functionality of the NMI Crash
Dump Switch. Thus, is should never be used during normal operation.
Compaq has enhanced Automatic Server Recovery (ASR) handling for situations where an
automated crash dump may be desired. ASR detects a lockup and normally provides a system
reset. ASR can be configured to generate the Crash Dump NMI instead of a reset. This will allow
for a crash dump file to be written prior to a system reset.
The Compaq health (systems management) driver must be loaded in order to properly
differentiate an NMI caused by the dump switch from a PCI SERR NMI. The dump switch
operates even if the health driver is not loaded.
The NMI Crash Dump switch may not work in all situations: after another NMI has already
occurred in the system, when the OS crash handler is incapable of running properly, following
some hardware failures, and when ASR is in progress, for example.