Hitachi Compute Blade 2000, Руководство пользователя

Страница: 1822 / 1927

1771
1
3
S
yst
em
O
p
er
at
ion and Man
agem
ent
Useful settings for Red Hat Enterprise
Linux
Using parameters to stop the system at
hardware failure

Potential problems and the solution
At hardware failure, Compute Blade issues an interrupt, which is called NMI (*1), to
stop the system during the hardware recovery process. Red Hat Enterprise Linux,
however, has a setting by default not to stop a system when NMI is issued. As a
result, unexpected behavior may occur because the system keeps running at
hardware failure.
A solution to this problem is to set parameters for Red Hat Enterprise Linux to stop
the system when NMI is issued. This can prevent unexpected behavior at hardware
failure. It is highly recommended that you should set these parameters.
In the environment with Linux Tough Dump, a highly reliable tool for collecting
dumps, the setting stops the system when NMI is issued. Thus, you do not need to
change the setting.
(*1) NMI (Non-Maskable Interrupt): hardware interruption that cannot be prevented by
software.
Red Hat Enterprise Linux has some functions to utilize NMI,
such as monitoring CPU hang ups, profiling kernel, and
stopping a system instantly at receiving NMI, although
some functions are not supported by some OS versions.
You can use one of them as a dedicated function. When
the instant stop at NMI is enabled, CPU watchdog and
kernel profile are not available.

System that requires the setting change
Systems that satisfy the following two conditions require the setting change.
1. Linux Tough Dump is not applied.
2. The following OS is used:
Red Hat Enterprise Linux 5.3 (x86, AMD/Intel 64)

How to change the settings
1. Change “/boot/grub/grub.conf”.
Set a kernel parameter “nmi_watchdog=0” .