xcat-core/xCAT-client/pods/man1/xdsh.1.pod
2013-08-24 06:31:21 -04:00

588 lines
22 KiB
Plaintext

=head1 B<NAME>
B<xdsh> - Concurrently runs remote commands on multiple nodes (Management Node, Service Nodes, compute nodes), or an install image.
=head1 B<SYNOPSIS>
B<xdsh> I<noderange> [B<-B> I<bypass>] [B<--devicetype> I<type_of_device>] [B<-e>] [B<-E> I<environment_file>] [B<-f> I<fanout>]
[B<-L>] [B<-l> I<userID>] [B<-m>] [B<-o>
I<node_options>] [B<-Q>] [B<-r> I<node_remote_shell>] [B<-s>] [B<-S> B<csh>|B<ksh>] [B<-t> I<timeout>]
[B<-T>] [B<-v>] [B<-X> I<env_list>] [B<-z>] [B<--sudo>] I<command_list>
B<xdsh> I<noderange> [B<-K>]
B<xdsh> I<noderange> [B<-K>] [B<-l> I<userID>] B<--devicetype> I<type_of_device>
B<xdsh> [B<-i> I<image path | nim image name>] I<command_list>
B<xdsh> I<noderange> [B<-c>]
B<xdsh> [B<-h> | B<-V> | B<-q>]
=head1 B<DESCRIPTION>
The B<xdsh> command runs commands in parallel on remote nodes and/or the Management Node. The B<xdsh> command issues a
remote shell command for each target specified, and returns the output
from all targets,
formatted so that command results from all nodes can be managed.
If the command is to be executed on the Management Node, it does not use a remote shell command, but uses the local OS copy or shell command. The Management Node must be defined in the xCAT database. The best way to do this is to use the xcatconfig -m option.
The B<xdsh> command is an xCAT Distributed Shell Utility.
B<COMMAND> B<SPECIFICATION>:
The commands to execute on the targets are specified by the
I<command_list> B<xdsh> parameter, or executing a local script using the B<-e>
flag.
The syntax for the I<command_list> B<xdsh> parameter is as follows:
I<command>[; I<command>]...
where I<command> is the command to run on the remote
target. Quotation marks are required to ensure that all commands in the
list are executed remotely, and that any special characters are interpreted
correctly on the remote target. A script file on the local host can be
executed on each of the remote targets by using the B<-e>
flag. If B<-e> is specified, I<command_list> is the
script name and arguments to the script. For example:
xdsh hostname -e I<script_filename> [I<arguments>]...
The I<script_filename> file is copied to a random filename in the B</tmp>
directory on each remote target and then executed on the targets.
The B<xdsh> command does not work with any interactive commands, including
those that read from standard input.
B<REMOTE> B<SHELL> B<COMMAND>:
The B<xdsh> command uses a configurable remote shell command to execute
remote commands on the remote targets. Support is explicitly provided
for AIX Remote Shell and OpenSSH, but any secure remote command that
conforms to the IETF (Internet Engineering Task Force) Secure Remote
Command Protocol can be used.
The remote shell is determined as follows, in order of precedence:
1. The B<-r> flag.
2. The B<DSH_NODE_RSH> environment variable.
3. The default node remote shell as defined by the target I<context>.
4. The B</usr/bin/ssh> command.
The remote shell options are determined as follows, in order of prece-
dence:
1. The B<-o> flag.
2. The B<DSH_NODE_OPTS> environment variable.
B<REMOTE> B<SHELL> B<ENVIRONMENT>:
The shell environment used on the remote target defaults to the shell
defined for the I<user_ID> on the remote target. The command
syntax that B<xdsh> uses to form the remote commands can be specified using the B<-S>
flag. If B<-S> is not specified, the syntax defaults to B<sh> syntax.
When commands are executed on the remote target, the path used is
determined by the B<DSH_PATH> environment variable defined in the shell of
the current user. If B<DSH_PATH> is not set, the path used is the remote
shell default path. For example, to set the local path for the remote
targets, use:
DSH_PATH=$PATH
The B<-E> flag exports a local environment definition file to each remote
target. Environment variables specified in this file are defined in the
remote shell environment before the I<command_list> is executed.
The definition file should contain entries like the following
and be executable. One environment variable per line.
export NEWENVVARIABLE="yes"
export ANOTHERENVVARIABLE="yes"
B<COMMAND> B<EXECUTION>:
The maximum number of concurrent remote shell command processes (the
fanout) can be specified with the B<-f> flag or with the B<DSH_FANOUT>
environment variable. The fanout is only restricted by the number of remote
shell commands that can be run in parallel. You can experiment with the
B<DSH_FANOUT> value on your management server to see if higher values are
appropriate.
A timeout value for remote command execution can be specified with the
B<-t> flag or with the B<DSH_TIMEOUT> environment variable. If any remote
target does not provide output to either standard output or standard
error within the timeout value, B<xdsh> displays an error message and
exits.
If streaming mode is specified with the B<-s> flag, output is returned as
it becomes available from each target, instead of waiting for the
I<command_list> to complete on all targets before returning output. This can
improve performance but causes the output to be unsorted.
The B<-z> flag displays the exit code from the last command issued on the
remote node in I<command_list>. Note that OpenSSH behaves differently; it
returns the exit status of the last remote command issued as its exit
status. If the command issued on the remote node is run in the
background, the exit status is not displayed.
The B<-m> flag monitors execution of the B<xdsh> command by printing status
messages to standard output. Each status message is preceded by B<dsh>.
The B<-T> flag provides diagnostic trace information for the execution of
the B<xdsh> command. Default settings and the actual remote shell commands
executed on the remote targets are displayed.
No error detection or recovery mechanism is provided for remote
targets. The B<xdsh> command output to standard error and standard output can
be analyzed to determine the appropriate course of action.
B<COMMAND> B<OUTPUT>:
The B<xdsh> command waits until complete output is available from each
remote shell process and then displays that output before initiating
new remote shell processes. This default behavior is overridden by the
B<-s> flag.
The B<xdsh> command output consists of standard error and standard output
from the remote commands. The B<xdsh> standard output is the standard
output from the remote shell command. The B<xdsh> standard error is the
standard error from the remote shell command. Each line is prefixed with
the host name of the node that produced the output. The host name is
followed by the B<:> character and a command output line. A filter for
displaying identical outputs grouped by node is provided separately.
See the B<xdshbak> command for more information.
A command can be run silently using the B<-Q> flag; no output from each
target's standard output or standard error is displayed.
B<SIGNALS>:
Signal 2 (INT), Signal 3 (QUIT), and Signal 15 (TERM) are propagated to
the commands executing on the remote targets.
Signal 19 (CONT), Signal 17 (STOP), and Signal 18 (TSTP) default to
B<xdsh>; the B<xdsh> command responds normally to these signals, but the
signals do not have an effect on remotely executing commands. Other
signals are caught by B<xdsh> and have their default effects on the B<xdsh>
command; all current child processes, through propagation to remotely
running commands, are terminated (SIGTERM).
=head1 B<OPTIONS>
=over 5
=item B<-c>|B<--cleanup>
This flag will have xdsh remove all files from the subdirectories of the
the directory on the servicenodes, where xdcp stages the copy to the
compute nodes as defined in the site table SNsyncfiledir and nodesyncfiledir
attribute, when the target is a service node.
It can also be used to remove the nodesyncfiledir directory on the compute
nodes, which keeps the backup copies of files for the xdcp APPEND function
support, if a compute node is the target.
=item B<-e>|B<--execute>
Indicates that I<command_list> specifies a local script
filename and arguments to be executed on the remote targets.
The script file is copied to the remote targets and then
remotely executed with the given arguments. The
B<DSH_NODE_RCP> environment variables specify the remote copy
command to use to copy the script file to node targets.
=item B<-E>|B<--environment> I<environment_file>
Specifies that the I<environment_file> contains environment
variable definitions to export to the target before
executing the I<command_list>.
=item B<--devicetype> I<type_of_device>
Specify a user-defined device type that references the location
of relevant device configuration file. The devicetype value must
correspond to a valid device configuration file.
xCAT ships some default configuration files
for Ethernet switches and and IB switches under
I</opt/xcat/share/xcat/devicetype> directory. If you want to overwrite
any of the configuration files, please copy it to I</var/opt/xcat/>
directory and cutomize it.
For example, I<base/IBSwitch/Qlogic/config> is the configuration
file location if devicetype is specified as IBSwitch::Qlogic.
xCAT will first search config file using I</var/opt/xcat/> as the base.
If not found, it will search for it using
I</opt/xcat/share/xcat/devicetype/> as the base.
=item B<-f>|B<--fanout> I<fanout_value>
Specifies a fanout value for the maximum number of concur-
rently executing remote shell processes. Serial execution
can be specified by indicating a fanout value of B<1>. If B<-f>
is not specified, a default fanout value of B<64> is used.
=item B<-h>|B<--help>
Displays usage information.
=item B<-i>|B<--rootimg> I<install image>
For Linux, Specifies the path to the install image on the local node.
For AIX, specifies the name of the osimage on the local node. Run lsnim
for valid names.
xdsh will chroot (xcatchroot for AIX) to this path and run the xdsh command against the
install image. No other xdsh flags, environment variables apply with
this input. A noderange is not accepted. Only runs on the local host,
normally the Management Node. The command you run must not prompt for input, the prompt will not be returned to you, and it will appear that xdsh hangs.
=item B<-K>|B<--ssh-setup>
=item B<-K>|B<--ssh-setup> B<-l>|B<--user> I<user_ID> B<--devicetype> I<type_of_device>
Set up the SSH keys for the user running the command to the specified node list.
The userid must have the same uid, gid and password as the userid on the node
where the keys will be setup.
If the current user is root, roots public ssh keys will be put in the
authorized_keys* files under roots .ssh directory on the node(s).
If the current user is non-root, the user must be in the policy table and have credential to run the xdsh command.
The non-root users public ssh keys and root's public ssh keys will be put in
the authorized_keys* files under the non-root users .ssh directory on the node(s).
Other device types, such as IB switch, are also supported. The
device should be defined as a node and nodetype should be defined
as switch before connecting.
The xdsh -K command must be run from the Management Node.
=item B<-l>|B<--user> I<user_ID>
Specifies a remote user name to use for remote command exe-
cution.
=item B<-L>|B<--no-locale>
Specifies to not export the locale definitions of the local
host to the remote targets. Local host locale definitions
are exported by default to each remote target.
=item B<-m>|B<--monitor>
Monitors remote shell execution by displaying status
messages during execution on each target.
=item B<-o>|B<--node-options> I<node_options>
Specifies options to pass to the remote shell command for
node targets. The options must be specified within double
quotation marks ("") to distinguish them from B<xdsh> options.
=item B<-q>|B<--show-config>
Displays the current environment settings for all DSH
Utilities commands. This includes the values of all environment
variables and settings for all currently installed and
valid contexts. Each setting is prefixed with I<context>: to
identify the source context of the setting.
=item B<-Q>|B<--silent>
Specifies silent mode. No target output is written to stan-
dard output or standard error. Monitoring messages are
written to standard output.
=item B<-r>|B<--node-rsh> I<node_remote_shell>
Specifies the path of the remote shell command used
for remote command execution on node targets.
=item B<-s>|B<--stream>
Specifies that output is returned as it becomes available
from each target, instead of waiting for the I<command_list>
to be completed on a target before returning output.
=item B<-S>|B<--syntax> B<csh>|B<ksh>
Specifies the shell syntax to be used on the remote target.
If not specified, the B<ksh> syntax is used.
=item B<--sudo>|B<--sudo>
Adding the --sudo flag to the xdsh command will have xdsh run sudo before
running the command. This is particular useful when using the -e option.
This is required when you input -l with a non-root user id and want that id
to be able to run as root on the node. The non-root userid will must be
previously defined as an xCAT user, see process for defining non-root ids in
xCAT and setting up for using xdsh. The userid sudo setup will have
to be done by the admin on the node. This includes, allowing all commands that
you would like to run with xdsh by using visudo to edit the /etc/sudoers file.
You must disabl ssh tty requirements by commenting out or removing this line in the /etc/sudoes file "#Defaults requiretty".
See the document Granting_Users_xCAT_privileges for sudo setup requirements.
This is not supported in a hierarical cluster, that is the nodes are serviced by servicenodes.
=item B<-t>|B<--timeout> I<timeout>
Specifies the time, in seconds, to wait for output from any
currently executing remote targets. If no output is
available from any target in the specified I<timeout>, B<xdsh>
displays an error and terminates execution for the remote
targets that failed to respond. If I<timeout> is not specified,
B<xdsh> waits indefinitely to continue processing output from
all remote targets.
=item B<-T>|B<--trace>
Enables trace mode. The B<xdsh> command prints diagnostic
messages to standard output during execution to each target.
=item B<-v>|B<--verify>
Verifies each target before executing any remote commands
on the target. If a target is not responding, execution of
remote commands for the target is canceled. When specified
with the B<-i> flag, the user is prompted to retry the
verification request.
=item B<-V>|B<--version>
Displays the B<xdsh> command version information.
=item B<-X> I<env_list>
Ignore B<xdsh> environment variables. This option can take an
argument which is a comma separated list of environment
variable names that should B<NOT> be ignored. If there is no
argument to this option, or the argument is an empty
string, all B<xdsh> environment variables will be ignored.
This option is useful when running B<xdsh> from within other
scripts when you don't want the user's environment affecting
the behavior of xdsh.
=item B<-z>|B<--exit-status>
Displays the exit status for the last remotely executed
non-asynchronous command on each target. If the command
issued on the remote node is run in the background, the
exit status is not displayed.
Exit values for each remote shell execution are displayed in
messages from the B<xdsh> command, if the remote shell exit values are
non-zero. A non-zero return code from a remote shell indicates that
an error was encountered in the remote shell. This return code is
unrelated to the exit code of the remotely issued command. If a
remote shell encounters an error, execution of the remote command on
that target is bypassed.
The B<xdsh> command exit code is B<0> if the command executed without
errors and all remote shell commands finished with exit codes of B<0>.
If internal B<xdsh> errors occur or the remote shell commands do not
complete successfully, the B<xdsh> command exit value is greater than
B<0>. The exit value is increased by B<1> for each successive instance of
an unsuccessful remote command execution. If the remotely issued
command is run in the background, the exit code of the remotely
issued command is B<0>.
=back
=head1 B<Environment> B<Variables>
=over 4
=item B<DEVICETYPE>
Specify a user-defined device type. See B<--devicetype> flag.
=item B<DSH_ENVIRONMENT>
Specifies a file that contains environment variable
definitions to export to the target before executing the remote
command. This variable is overridden by the B<-E> flag.
=item B<DSH_FANOUT>
Specifies the fanout value. This variable is overridden by
the B<-f> flag.
=item B<DSH_NODE_OPTS>
Specifies the options to use for the remote shell command
with node targets only. This variable is overridden by the
B<-o> flag.
=item B<DSH_NODE_RCP>
Specifies the full path of the remote copy command to use
to copy local scripts and local environment configuration
files to node targets.
=item B<DSH_NODE_RSH>
Specifies the full path of the remote shell to use for
remote command execution on node targets. This variable is
overridden by the B<-r> flag.
=item B<DSH_PATH>
Sets the command path to use on the targets. If B<DSH_PATH> is
not set, the default path defined in the profile of the
remote I<user_ID> is used.
=item B<DSH_SYNTAX>
Specifies the shell syntax to use on remote targets; B<ksh> or
B<csh>. If not specified, the B<ksh> syntax is assumed. This
variable is overridden by the B<-S> flag.
=item B<DSH_TIMEOUT>
Specifies the time, in seconds, to wait for output from
each remote target. This variable is overridden by the B<-t>
flag.
=back
=head1 B<Compatibility with AIX dsh>
To provide backward compatibility for scripts written using dsh in
AIX and CSM, a tool has been provide B<groupfiles4dsh>,
which will build node group files from the
xCAT database that can be used by dsh. See man groupfiles4dsh.
=head1 B<Security>
The B<xdsh> command has no security configuration requirements. All
remote command security requirements - configuration,
authentication, and authorization - are imposed by the underlying remote
command configured for B<xdsh>. The command assumes that authentication
and authorization is configured between the local host and the
remote targets. Interactive password prompting is not supported; an
error is displayed and execution is bypassed for a remote target if
password prompting occurs, or if either authorization or
authentication to the remote target fails. Security configurations as they
pertain to the remote environment and remote shell command are
userdefined.
=head1 B<Exit Status>
The dsh command exit code is 0 if the command executed without errors and all remote shell commands finished with exit codes of 0. If internal dsh errors occur or the remote shell commands do not complete successfully, the dsh command exit value is greater than 0. The exit value is increased by 1 for each successive instance of an unsuccessful remote command execution. If the remotely issued command is run in the background, the exit code of the remotely issued command is 0.
=head1 B<Examples>
=over 3
=item *
To set up the SSH keys for root on node1, run as root:
B<xdsh> I<node1 -K>
=item *
To run the B<ps -ef > command on node targets B<node1> and B<node2>, enter:
B<xdsh> I<node1,node2 "ps -ef">
=item *
To run the B<ps> command on node targets B<node1> and run the remote command with the -v and -t flag, enter:
B<xdsh> I<node1,node2 -o"-v -t" ps>
=item *
To execute the commands contained in B<myfile> in the B<XCAT>
context on several node targets, with a fanout of B<1>, enter:
B<xdsh> I<node1,node2 -f 1 -e myfile>
=item *
To run the ps command on node1 and ignore all the dsh
environment variable except the DSH_NODE_OPTS, enter:
B<xdsh> I<node1 -X `DSH_NODE_OPTS' ps>
=item *
To run on Linux, the xdsh command "rpm -qa | grep xCAT"
on the service node fedora9 diskless image, enter:
B<xdsh> I<-i /install/netboot/fedora9/x86_64/service/rootimg "rpm -qa | grep xCAT">
=item *
To run on AIX, the xdsh command "lslpp -l | grep bos"
on the NIM 611dskls spot, enter:
B<xdsh> I<-i 611dskls "/usr/bin/lslpp -l | grep bos">
=item *
To cleanup the servicenode directory that stages the copy of files to the
nodes, enter:
B<xdsh> I<servicenoderange -c >
=item *
To define the QLogic IB switch as a node and to set up the SSH keys for IB switch
B<qswitch> with device configuration file
B</var/opt/xcat/IBSwitch/Qlogic/config> and user name B<username>, Enter
B<chdef> I<-t node -o qswitch groups=all nodetype=switch>
B<xdsh> I<qswitch -K -l username --devicetype IBSwitch::Qlogic>
=item *
To define the Management Node in the database so you can use xdsh, Enter
B<xcatconfig -m>
=item *
To define the Mellanox switch as a node and run a command to show the ssh keys.
B<mswitch> with and user name B<username>, Enter
B<chdef> I<-t node -o mswitch groups=all nodetype=switch>
B<xdsh> I<mswitch -l admin --devicetype IBSwitch::Mellanox 'enable;configure terminal;show ssh server host-keys'>
=item *
To define a BNT Ethernet switch as a node and run a command to create a new vlan with vlan id 3 on the switch.
B<chdef> I<myswitch groups=all>
B<tabch> I<switch=myswitch switches.sshusername=admin switches.sshpassword=passw0rd switches.protocol=[ssh|telnet]>
where I<admin> and I<passw0rd> are the SSH user name and password for the switch. If it is for Telnet, add I<tn:> in front of the user name: I<tn:admin>.
<xdsh> I<myswitch --devicetype EthSwitch::BNT 'enable;configure terminal;vlan 3;end;show vlan'>
=item *
To run xdsh with the non-root userid "user1" that has been setup as an xCAT userid and with sudo on node1 and node2 to run as root, do the following, see xCAT doc on Granting_Users_xCAT_privileges:
B<xdsh> I<node1,node2 --sudo -l user1 "cat /etc/passwd">
=back
=head1 B<Files>
=head1 B<SEE ALSO>
L<xdshbak(1)|xdshbak.1>, L<noderange(3)|noderange.3>, L<groupfiles4dsh(1)|groupfiles4dsh.1>