Hello to everyone, this is my first post here so I hope someone can help me.
I'm running several Linux (different infrastructures) and recently one of them failed, when we try to find anything to diagnose the failure we couldn't because there was no memory dump in the system, we checked with novell and they told us to use this feature:
http://www.novell.com/coolsolutions/feature/15284.html
I've been trying to set it up with no luck at all, can someone point me out were is my error?
I'm trying out in a SLES 9 SP 3
And this is my config file:
## Path: System/Kernel/LKCD
## Description: Linux Kernel Crash Dump (LKCD) options
#
# Copyright 1999-2005 Silicon Graphics, Inc. All rights reserved.
#
# This file contains the configuration variables for Linux kernel crash
# dumps. The file should reside in /etc/sysconfig/dump, permissions
# set to 0644.
## Type: list(0,1)
## Default: 1
# DUMP_ACTIVE indicates whether the dump process is active or not. If this
# variable is 0, the dump kernel process will not be activated.
#
DUMP_ACTIVE="1"
## Type: string
## Default: eth0
#
# DUMPDEV represents the name of the dump device.
#
# For local disk dumps, it is typically the primary swap partition on
susetest:~ # cat /etc/sysconfig/dump
## Path: System/Kernel/LKCD
## Description: Linux Kernel Crash Dump (LKCD) options
#
# Copyright 1999-2005 Silicon Graphics, Inc. All rights reserved.
#
# This file contains the configuration variables for Linux kernel crash
# dumps. The file should reside in /etc/sysconfig/dump, permissions
# set to 0644.
## Type: list(0,1)
## Default: 1
# DUMP_ACTIVE indicates whether the dump process is active or not. If this
# variable is 0, the dump kernel process will not be activated.
#
DUMP_ACTIVE="1"
## Type: string
## Default: eth0
#
# DUMPDEV represents the name of the dump device.
#
# For local disk dumps, it is typically the primary swap partition on
# the local system, although any disk device can be used. Please be
# EXTRA careful when defining this value, as one false slip can cause
# all kinds of problems. Currently, a link is created from /dev/vmdump
# to the first swap device; rename this to the exact device to dump to
# if that's what you want.
#
# For network dumps, it is the name of the interface used for sending
# the crash dump. Note that this needs to support polled operation.
#
DUMPDEV="/dev/vgroot/lvswap"
## Type: string
## Default: /var/log/dump
#
# DUMPDIR is the location where crash dumps are saved. In that
# directory, a file called 'bounds' will be created, which is
# the current index of the last crash dump saved. The 'bounds'
# file will be updated with an incremented once a new crash dump or
# crash report is saved.
#
DUMPDIR="/var/log/dump"
## Type: list(0,1,2,4,8)
## Default: 2
#
# DUMP_LEVEL has a number of possible values:
#
# DUMP_NONE (0): Do nothing, just return if called.
# DUMP_HEADER (1): Dump the dump header and first 128K bytes out.
# DUMP_KERN (2): Everything in DUMP_HEADER and kernel pages only.
# DUMP_USED (4): Everything except kernel free pages.
# DUMP_ALL (8): All memory.
#
# NOTE: You must use the numeric value, not the name of the variable.
DUMP_LEVEL="2"
## Type: integer(0:2)
## Default: 2
#
# DUMP_COMPRESS indicates which compression mechanism the kernel should
# attempt to use for compression -- the default is not to use dump
# compression unless someone specifically asks for it.
#
# The values for DUMP_COMPRESS are currently:
#
# DUMP_COMPRESS_NONE(0): Don't compress this dump.
# DUMP_COMPRESS_RLE(1): Use RLE compression.
# DUMP_COMPRESS_GZIP(2): Use GZIP compression.
#
DUMP_COMPRESS="2"
## Type: list(0x80000000,0x80000001,0x80000004,0x80000005,0x40000000,0x40000008)
## Default: 0x40000000
#
# DUMP_FLAGS are the flag parameters to use when configuring system dumps.
#
# DUMP_FLAGS_NONE(0x00000000): No flags are required.
# DUMP_FLAGS_NONDISRUPT(0x00000001): Do not reboot after dumping;
# continue running.
# DUMP_FLAGS_POLL(0x00000004): Use polling I/O if available.
# DUMP_FLAGS_SECURE(0x00000008): Use secure mode if available.
# DUMP_FLAGS_DISKDUMP(0x80000000): Dump target is a local block device.
# DUMP_FLAGS_NETDUMP(0x40000000): Dump target is a network device.
#
# DUMP_FLAGS_POLL uses polling I/O for disk dumps if available.
#
# DUMP_FLAGS_SECURE switches to secure mode for netdumps.
#
# Note: DUMP_FLAGS must be specified as a hexadecimal number
#
DUMP_FLAGS="0x40000000"
## Type: list(0,1)
## Default: 1
#
# DUMP_SAVE defines whether to save the memory image to disk or not.
# If the value is 1, the vmcore image is stored, and a crash report
# is created from the saved dump. If it is not set to 1, only a crash
# report will be created, and the dump will not be saved. This option
# can be used on systems that do not want their disk space consumed
# by large crash dump images.
#
DUMP_SAVE="1"
## Type: integer
## Default: 5
#
# PANIC_TIMEOUT represents the timeout (in seconds) before reboot after a
# panic occurs. Typically this is set to 0 on the system, which means the
# kernel sits and spins until someone resets the machine. This is not the
# preferred action if we want to recover the dump after the reboot.
#
PANIC_TIMEOUT="5"
## Type: integer
## Default: 10
#
# BOUNDS_LIMIT places a limit on the number of dumps that can be retained
# in the /var/log/dump directory. Default value is 10. This implies after
# dump.9, the next dump is stored as dump.0, replacing the existing file.
# Set this value to 0 if you do not want to limit the number of dumps.
#
BOUNDS_LIMIT=10
## Type: string
## Default: /boot/vmlinuz
#
# Kexec based softboot dump specific configuration variables (currently
# valid only for lkcd on 2.5 kernels)
#
# KEXEC_IMAGE is the kernel boot image in the /boot directory which
# needs to be booted after the dump. This will be picked by
# kexec as the next kernel to be softbooted. The next kernel
# can normally be the same kernel on which the dump occurs.
#
# Only relevant for dumping via kexec (present only in lkcd for 2.6 kernels)
KEXEC_IMAGE=/boot/vmlinuz
## Type: string
## Default: "root console=tty0"
#
# KEXEC_CMDLINE command line arguments to the kernel(specified by
# KEXEC_IMAGE) which will be booted after the dump This can
# normally be the output of /proc/cmdline.
#
# Note the value given here is only an example.
KEXEC_CMDLINE="root console=tty0"
## Type: string
## Default: 192.168.100.1
#
# TARGET_HOST is the hostname/ip address of the target machine to which
# network dump needs to be sent. This target machine needs to be running
# netdump server application. The ip address should be in standard
# dotted format.
# TARGET_HOST _must_ be configured before netdump can be used.
# No default value is given here; the setting needs to be adapted
# to the local network configuration.
TARGET_HOST="localhost"
## Type: integer
## Default: 6688
#
# TARGET_PORT is the port on which the netdump server in the TARGET_HOST
# machine is listening for network dumps. Default is 6688.
#
TARGET_PORT=6688
## Type: integer
## Default: 6688
#
# SOURCE_PORT is the port used on dumping machine to send dump data over
# the network. Default is 6688.
#
SOURCE_PORT=6688
## Type: string
## Default: ff:ff:ff:ff:ff:ff
#
# ETH_ADDRESS is the Ethernet (MAC) address of the network device on the
# TARGET_HOST used to receive dumps. This should be specified in standard
# hex-digits-and-colons notation.
# By default, broadcast ethernet address is used which can increase
# the load on other systems in the same ethernet segment.
# To reduce the load use the ethernet address of the netdump server
#
ETH_ADDRESS=ff:ff:ff:ff:ff:ff
## Type: integer
## Default: 4
#
# DUMP_MAX_CONCURRENT is the maximum number of concurrent dumps the server
# will allow.
#
DUMP_MAX_CONCURRENT=4
## Type: yesno
## Default: no
#
# NETDUMP_VERBOSE enables verbose output of the netdump server to be
# sent to syslog. Useful for debugging.
#
NETDUMP_VERBOSE=no
Thanks for everything.
