KDUMP(5) User Manuals KDUMP(5)NAMEkdump - Configuration of kdumpSYNOPSIS
/etc/sysconfig/kdump
DESCRIPTION
This configuration file provides various options for the kdump loading
and saving process. It's mostly read by kdumptool(8) but also by the
kdump init script.
OPTIONS
The file format consists of shell variables (which means that the
configuration file is actually parsed by a sh(1) compatible shell) that
are described below. It's recommended to use a pair of double quotes to
assign values to that variables.
KDUMP_KERNELVER
Kernel Version string for the -kdump kernel, such as "2.6.25-5-kdump".
The init script will use a kernel named
/boot/vmlinu[zx]-$KDUMP_KERNELVER. Using "kdump" will default to the
most recently installed kdump kernel (to which the symbolic link
/boot/vmlinuz-kdump points).
If no version is specified, then the init script will try to find a
kdump kernel with the algorithm described below.
Default is "" which means that the auto-detection described above is
used.
Kernel auto-detection
Let's assume that a kernel name KERNELRELEASE consists of
BASEVERSION and a FLAVOUR on SUSE. For example consider the kernel
release ("uname -r") "2.6.25.18-0.2-default". Then the BASEVERSION
is "2.6.25.18-0.2" (but there are also BASEVERSION strings without
a dash) and the FLAVOUR is "default".
To find out if a kernel is suitable as capture kernel, there are
strong and weak requirements. The strong requirements mean that you
just cannot use the kernel as kdump kernel:
· Either the kernel must be a special kdump kernel with a
modified start address or
· the kernel is relocatable.
The weak requirements mean that a kernel is not good as kdump
kernel so it should be avoided when better kernels are installed. A
kernel does not fulfil that weak requirements when
· It is a realtime kernel (since the interrupt handling is
different and irqpoll has problems) or
· the kernel is compiled for a huge number of CPUs (above 1024 in
our case) since that means that the capure kernel needs more
memory.
After all that definition, the algorithm tries following kernels in
that order:
1. BASEVERSION-kdump (e.g. /boot/vmlinuz-2.6.27-rc1-kdump)
2. kdump (e.g. /boot/vmlinuz-kdump)
3. KERNELRELEASE (e.g. /boot/vmlinuz-2.6.27-rc1-kdump), strong
check
4. BASEVERSION-default (e.g. /boot/vmlinuz-2.6.27-rc1-default),
strong check
5. "" (e.g. /boot/vmlinuz), strong check
6. KERNELRELEASE (e.g. /boot/vmlinuz-2.6.27-rc1-kdump), weak check
7. "" (e.g. /boot/vmlinuz), weak check
On each item, it tries:
1. /boot/*vmlinuz
2. /boot/*vmlinux
Initrd
For the initrd, if the kernel name already contains "kdump", then
the initrd /boot/initrd-<kernelver> is used. Else, we use
/boot/initrd-<kernelver>-kdump (the kernel flavour is not replaced,
so for example this means /boot/initrd-2.6.25.9-0.2-default-kdump.
This is because the initrd used for kdump always has to be special,
and it makes no sense to load the normal initrd that would not save
the dump.
KDUMP_CPUS
If dumping is CPU-bound on your system, you can parallelize it on an
SMP system. Keep in mind that this increases the stress on the I/O
subsystem, so you probably don't get the maximum speed if you set this
parameter to the number of CPUs in your system.
This parameter modifies the maxcpus parameter of the kdump kernel and
the number of makedumpfile(8) processes created.
Note: This parameter does not work properly for the ELF format, because
makedumpfile(8) does not support split ELF dump files.
WARNING: SETTING THIS NUMBER TO A NUMBER HIGHER THAN 1 IS EXPERIMENTAL!
SOME MACHINES ARE KNOWN TO HANG OR REBOOT RANDOMLY DURING
INITIALISATION OF THE DUMP KERNEL IF KDUMP_CPUS > 1.
Default is 1.
KDUMP_COMMANDLINE
Command line used to boot the kdump kernel. By default, the
/etc/init.d/kdump script tries to build a command line automatically
from the command line of the running system, preserving root partition,
serial console, etc. If you manually create a command line, make sure
to include:
· the root parameter,
· any console parameters (for serial console),
· irqpoll,
· reset_devices,
· elevator=deadline to save the memory footprint.
The last three parameters are special for kdump and should always be
included, if you don't have a good reason to exclude them.
If you only want to extend the kernel command line with your own
parameter(s), use KDUMP_COMMANDLINE_APPEND.
Default is "".
KDUMP_COMMANDLINE_APPEND
This variable describes all command line parameters that are passed to
the kdump kernel additionally to the default parameters.
For network based dumping, you may have to add a net_delay parameter to
specify the extra delay in seconds that is needed for the network
device to become fully operational, e.g. net_delay=20 will wait 20
seconds before continuing after the network device is configured. This
parameter is used by the network setup code in the resulting initrd.
See also KDUMP_COMMANDLINE.
Default is "".
KDUMP_FADUMP
If this option is set to "yes", kdump will use the FADUMP mechanism
(firmware-assisted dump). With this option, the secondary kernel is
never loaded from the primary kernel. Instead, the kdump feature is
added to the standard initrd. If the primary kernel crashes, the
firmware reboots the machine in a special environment. The initrd
detects this environment on boot, saves the dump, and continues with
normal startup.
Note: FADUMP is only available on powerpc.
Default is "no".
KEXEC_OPTIONS
Additional options for kexec(8).
Default is "" except on IA64 where it is "--noio".
MAKEDUMPFILE_OPTIONS
Additional options for makedumpfile(8). makedumpfile will be used to
save the dump if KDUMP_DUMPLEVEL is non-zero or KDUMP_DUMPFORMAT is not
ELF. Normally, you don't have to specify any options here, but you may
be asked in Bugzilla to add the -D option for debugging.
Default is "".
KDUMP_IMMEDIATE_REBOOT
Immediately reboot after saving the core in the kdump kernel? Use "yes"
or "no".
Default is "yes".
KDUMP_TRANSFER
A script or command executed to process and transfer the dump image. It
can read the dump image either via /proc/vmcore or /dev/oldmem. Most
other variables will have no effect if you use a custom command here.
It's important that all requirements including that script itself are
included in KDUMP_REQUIRED_PROGRAMS.
It's strongly recommended that the default mechanism is used, it should
provide everything that is needed, together with KDUMP_PRESCRIPT and
KDUMP_POSTSCRIPT. We can't support problems with using that variable.
KDUMP_PRESCRIPT and KDUMP_POSTSCRIPT are not executed if KDUMP_TRANSFER
is set.
Default: "".
KDUMP_SAVEDIR
The directory where dumps are saved to. If the directory does not
exist, it will be created. The directory is a URL and must conform to
the syntax described in the section "URL FORMAT" below. If the
directory does not exist, it will be created. (That is true for all
targets, even network targets. Of course the user must have the
permission to create directories.)
Below that directory, for each dump a directory with a time stamp in
the following format will be created: "YYYY-MM-TT-HH:MM". That
directory contains the dump and various other files, depending which
dump target will be used and which dump settings are made.
If KDUMP_COPY_KERNEL is set, that directory will also contain the
kernel.
You can specify multiple targets separated by spaces. The dump will
then be split and saved to all target directories in parallel. This is
useful if the targets are on different storage devices, because their
combined I/O bandwidth can be used. Note how this option interacts with
KDUMP_CPUS. If you specify more CPUs than target locations, then they
will be assigned to the available processes in a round-robin fashion
(i.e. more than one process will be writing to the same directory). If
you specify more locations than CPUs, then only the first KDUMP_CPUS
locations will be used. This feature is supported only for local files
using the kdump-compressed format.
Default: "file:///var/log/dump".
KDUMP_KEEP_OLD_DUMPS
Number of old dumps to keep. That variable is only honored on local
directories (i.e., if KDUMP_SAVEDIR starts with file) because we think
it's bad from a security point of view if other hosts delete stuff
(that may be from another hosts) on a dump server. The deletion process
takes place before the dumps are saved. So if you specify 3 here, then
after the dump has been saved, 4 dumps are on disk.
Set that variable to "0" to disable the deletion of dumps entirely, and
set that variable to "-1" to delete all dumps, i.e. then only the just
saved dump is on disk.
Default: "5"
KDUMP_FREE_DISK_SIZE
Don't save the dump (or delete it afterwards) if after saving the dump
less than KDUMP_FREE_DISK_SIZE megabytes are free. Because when dump
compression or dump filtering (see KDUMP_DUMPLEVEL and
KDUMP_DUMPFORMAT) is used, we don't know in advance how large the dump
will be, we have to delete the dump after saving it.
That option applies only to local file systems, i.e. KDUMP_SAVEDIR must
start with file.
Default: "64"
KDUMP_VERBOSE
Bit mask of several options that set the verbosity of the dump process.
To apply several options, you have to add their value. The options are:
1
kdump command line is written to system log when executing
/etc/init.d/kdump.
2
Progress is written to stdout while dumping.
4
kdump command line is written so standard output when executing
/etc/init.d/kdump.
8
Debugging for kdump transfer mechanism. That is executed in initrd
and generates lots of output. However, it's very useful if
something does not work.
16
Debugging when loading the kdump kernel. That creates lots of
output to stdout, so only use if you really want to debug that. It
is useful if kdump reports that it cannot find a suitable kernel
but you think kdump is wrong.
So, for example, if you want to have the command line on stdout when
loading the kernel and also in the system log, just use KDUMP_VERBOSE =
5 (i.e. 1 + 4).
Default: "3".
KDUMP_DUMPLEVEL
Determines the dump level. If KDUMP_DUMPLEVEL is non-zero, then
makedumpfile(8) is used to strip pages that may not be necessary for
analysing. 0 means no stripping, and 31 is the maximum dump level, i.e.
0 produces the largest dump files and 31 the smallest.
The following table from makedumpfile(8) shows what each dump level
means:
┌───────────┬───────────┬────────────┬─────────┬───────────┬───────────┐
│dump level │ zero page │ cache page │ cache │ user data │ free page │
│ │ │ │ private │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 0 │ │ │ │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 1 │ X │ │ │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 2 │ │ X │ │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 3 │ X │ X │ │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 4 │ │ X │ X │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 5 │ X │ X │ X │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 6 │ │ X │ X │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 7 │ X │ X │ X │ │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 8 │ │ │ │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 9 │ X │ │ │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 10 │ │ X │ │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 11 │ X │ X │ │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 12 │ │ X │ X │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 13 │ X │ X │ X │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 14 │ │ X │ X │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 15 │ X │ X │ X │ X │ │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 16 │ │ │ │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 17 │ X │ │ │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 18 │ │ X │ │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 19 │ X │ X │ │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 20 │ │ X │ X │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 21 │ X │ X │ X │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 22 │ │ X │ X │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 23 │ X │ X │ X │ │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 24 │ │ │ │ X │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 25 │ X │ │ │ X │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 26 │ │ X │ │ X │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 27 │ X │ X │ │ X │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 28 │ │ X │ X │ X │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 29 │ X │ X │ X │ X │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 30 │ │ X │ X │ X │ X │
├───────────┼───────────┼────────────┼─────────┼───────────┼───────────┤
│ 31 │ X │ X │ X │ X │ X │
└───────────┴───────────┴────────────┴─────────┴───────────┴───────────┘
Default: "31"
KDUMP_DUMPFORMAT
This variable specifies the dump format.
none
Do not generate any dump file. This is useful if you only want the
kernel log buffer, because it minimizes disk space requirements and
also greatly reduces the time involved in waiting for the dump to
complete.
ELF
ELF has the advantage that it's a standard format and GDB can be
used to analyse the dumps. The disadvantage is that the dump files
are larger.
compressed
compressed is the kdump compressed format that produces small
dumps, see makedumpfile(8). However, only crash(8) can analyse the
dumps and makedumpfile must be installed (but you need it anyway if
you set KDUMP_DUMPLEVEL to non-zero before).
lzo
lzo is similar to compressed, but it uses the LZO compression
algorithm instead of gzip. This usually results in slightly bigger
files, but it is much faster. You will need crash(8) version 6.0.7
or later to open the dump file.
Default: "compressed"
KDUMP_CONTINUE_ON_ERROR
If something goes wrong while saving the dump or deleting old dumps in
initrd, then kdumptool(8) normally tries to execute the next action.
However, it can be useful to open a shell at that point in time and fix
or debug the problem. This happens when KDUMP_CONTINUE_ON_ERROR is set
to "false".
Default: "true"
KDUMP_REQUIRED_PROGRAMS
This is a space-separated list of programs (full path) that are
required by KDUMP_TRANSFER, KDUMP_PRESCRIPT or KDUMP_POSTSCRIPT.
Dynamic libraries that are required by that programs don't have to be
included as they are resolved automatically. However, it's necessary to
include other runtime requirements because it's not possible to resolve
them automatically.
Default: ""
KDUMP_PRESCRIPT
Program that is executed before taking the dump. You have to include
that program in KDUMP_REQUIRED_PROGRAMS.
Default: ""
KDUMP_POSTSCRIPT
Program that is executed after taking the dump and before the system is
rebooted. You have to include that program in KDUMP_POSTSCRIPT.
Default: ""
KDUMP_COPY_KERNEL
Not only copy the dump to KDUMP_SAVEDIR but also the installed kernel.
Valid values are "yes" and "no". If you want to have the debugging
information also copied, which is required to open the program, you
also have to install the kernel-<flavour>-debuginfo package.
Modules are not copied, only the kernel image and the debugging file.
Default: "yes"
KDUMPTOOL_FLAGS
This is a space-separated list of flags to tweak the run-time behaviour
of kdumptool(8). Currently, only one flag is supported:
NOSPARSE
Disable the creation of sparse-files. This flag is for debugging
purposes, e.g. if the file system or network protocol has problems
with sparse files. Because SFTP and FTP are not mounted, that
option has no meaning when saving the dump to SFTP and FTP.
Default: ""
KDUMP_NETCONFIG
Network configuration for kdump. Because the dump process runs in
initrd, the network configuration is different from the normal network
configuration. Use auto to auto-detect the network configuration, this
is also the default.
Use a netdevice:mode string to force a specific network device to be
used. A netdevice is for example "eth0". The mode can be either "dhcp"
or "static". If you use "static", you have to set the IP address with
ip=ipspec. ipspec is
<client>:<server>:<gateway>:<netmask>:<hostname>:<device>:<proto>. See
mkinitrd(8) for details.
You can set KDUMP_NETCONFIG to "" if you want no network in initrd,
i.e. you use disk dumping.
Default: "auto"
KDUMP_SMTP_SERVER
If you want to get an email after the dump has been copied, you can
specify an SMTP server here. It's valid to specify a hostname, an IP
address. Optionally, you can also specify a port with the usual :port
syntax.
Example: smtp.myprovider.de:25
If you provide a username (KDUMP_SMTP_USER) and a password
(KDUMP_SMTP_PASSWORD), then SMTP AUTH will be used. Otherwise, plain
SMTP with no authentication will be used.
Default: ""
KDUMP_SMTP_USER
Username used for SMTP authentication. Also specify KDUMP_SMTP_PASSWORD
in that case. If no username and/or no password will be provided, plain
SMTP will be used.
Default: ""
KDUMP_SMTP_PASSWORD
Password used for SMTP authentication. Also specify KDUMP_SMTP_USER in
that case. If no username and/or no password will be provided, plain
SMTP will be used.
Default: ""
KDUMP_NOTIFICATION_TO
Email address where notification mails should be sent to. Don't specify
a real name here, only an email address. You have to configure a SMTP
server with KDUMP_SMTP_SERVER to use that feature.
Example: john@myprovider.de
Default: ""
KDUMP_NOTIFICATION_CC
Space-separated list of email addresses where notification mails should
be sent to via Cc. Don't specify real names here, only email addresses.
You have to configure a SMTP server with KDUMP_SMTP_SERVER to use that
feature.
Example: "john@myprovider.de doe@myprovider.de"
Default: ""
KDUMP_SSH_HOST_KEY
Target host key encoded with base64, i.e. as found in
~/.ssh/known_hosts. This is used for the SFTP transfer protocol to
verify the authenticity of the remote host. If this field is left
empty, the setup scripts will try to get it automatically using
ssh-keygen -F $target_host.
If you want to disable host key checking, set this variable to "*".
Example:
"AAAAB3NzaC1yc2EAAAABIwAAAIEAxIJIeuVX0F0u53wMgH22fdl62ifLXRwaPAXPsCrvyerMEb5DXHnIv1sYdzlJ1g/2Iw+ppjGDh1En/P12Dpl2uLBZh0MOrM71spTjvTwV2MKZtXrYjZUIPIZdd9vrI+/LEI0nw7LSdnKKgkAIXMhaRrq9N6uoZWe41ilU4NpnL/k="
Default: ""
URL FORMAT
In general, the URL format is
protocol://specification
where protocol is one of:
· file,
· ftp,
· sftp (aliases are ssh or scp),
· nfs
· cifs (alias is smb).
The specification varies for each protocol.
Local files (file)
That type describes a local path.
Format: file://]path
As shown, the file prefix is optional, i.e. a simple path is still
valid. The path must start with a leading slash (/), i.e. only absolute
paths are allowed.
Examples:
· /var/log/dump
· file:///var/log/dump
File Transfer Protocol (ftp)
This URL type is used to specify paths on a remote FTP server.
Format: ftp:// [ user [:password]@ ] hostname [:port] ] path
The remote user is optional, if no user is specified, anonymous FTP is
assumed. Even if a username is specified, the password is optional, if
no password is specified, an empty password is used. When the username
is "anonymous", then "$USER@$HOSTNAME" is used as password because some
FTP servers deny anonymous FTP with empty passwords.
hostname can be either a (DNS-)resolvable hostname, with or without a
domain name, or an IP address in totted decimal format. port can be
used to specify the remote port, if no port is specified in the URL,
the default FTP port is used. Finally, path must conform to the same
rules as for local files (see above).
Examples:
· ftp://neptunium/var/log/dump
· ftp://bwalle@neptunium.suse.de/var/log/dump
· ftp://bwalle:dontsay@strauss.suse.de:123/var/log/dump
· ftp://192.168.0.70/var/log/dump
Secure File Transfer Protocol (sftp)
This URL type is used to specify paths on a remote server that is
reachable via SFTP. It's important that an SFTP server must be running
(which is the default on SUSE), the kdumptool does not use FISH (FIles
transferred over SHell protocol).
Format: sftp:// [ user [:password]@ ] hostname [:port] ] path
It's valid to use the alias "ssh" or "scp" as alias for "sftp" (for
backward compatibility), but the "sftp" prefix is recommended.
For the elements, see the description of FTP above. Please note that
the use of an password here is not recommended. The recommended way is
to add the key where the file should be copied to to the
"authorized_keys" file of the user where the dump is saved. For
example, if you want to save the dump to the "kdump" user of the
machine "collector", then generate a private/public key pair on the
machine that saves the dump with ssh-keygen -t dsa. Append the file
id_dsa.pub (in root/.ssh/) to ~dump/.ssh/authorized_keys on
"collector".
root@earth:~# ssh-keygen -t dsa
...
root@earth:~# cat /root/.ssh/id_dsa.pub \
| ssh kdump@collector "cat - >> /root/.ssh/authorized_keys"
Don't use any password to encrypt the key. kdumptool is designed to be
used non-interactively.
Examples:
· sftp://kdump@collector:22/var/log/dump
· sftp://neptunium/var/log/dump
· scp://neptunium/var/log/dump
Network File System (nfs)
NFS is used to specify remote hosts that export their file system via
NFS. NFS doesn't use username and password. Also, the URL format cannot
be used to specify options. As default, the nolock option is used.
Format: +nfs://host [:port:] _path
host is the hostname, port is optional (should not be used, it is only
implemented for completeness with the other URL formats). path
specifies the path that is exported via NFS in /etc/exports (or a
subdirectory of a path that is exported).
Examples:
· nfs://neptunium/var/log/dump
· nfs://10.10.2.120/var/log/dump
Common Internet File System (cifs)
CIFS is used to specify remote hosts that export a file system via
CIFS, also known as SMB (although they are technically different, CIFS
is the successor of SMB). This could be a Microsoft Windows server, or
a Samba Server running on Unix or MacOS.
Format: +cifs:// [user [:password] @] host path
Although it's technically not 100 % correct, smb is used as an alias
for cifs, so the program will always return cifs even if you specify
smb.
The parts of the URL are described in the FTP section above.
Note: You have to use the slash as path separator, not the backslash as
on Microsoft Windows.
Examples:
· cifs://neptunium/var/log/dump
· cifs://bwalle:dontsay@neptunium:/var/log/dump
· smb://bwalle@192.168.0.70:/var/log
BUGS
Please report bugs and enhancement requests at
https://bugzilla.novell.com.
COPYING
Copyright (c) 2008 Bernhard Walle <bwalle@suse.de>. Free use of this
software is granted under the terms of the GNU General Public License
(GPL), version 2 or later.
SEE ALSOkexec(8), kdump(5), makedumpfile(8), mkdumprd(8)
http://en.opensuse.org/Kdump
kdump 0.8.4 06/14/2013 KDUMP(5)
