/ 
SSH-Batch-0.020
River stage zero No dependents
/ SSH::Batch

NAME

SSH::Batch - Cluster operations based on parallel SSH, set and interval arithmetic

VERSION

This document describes SSH::Batch 0.020 released on May 3, 2009.

SYNOPSIS

The following scripts are provided:

fornodes

Expand patterns to machine host list.

$ cat > ~/.fornodesrc
ps=blah.ps.com bloo.ps.com boo[2-25,32,41-70].ps.com
as=ws[1101-1105].as.com
# use set operations to define new sets:
foo={ps} + {ps} * {as} - {ps} / {as}
 bar = foo.com bar.org \
    bah.cn \
    baz.com
^D

$ fornodes 'api[02-10].foo.bar.com' 'boo*.ps.com'
$ fornodes 'tq[ab-ac].[1101-1105].foo.com'
$ fornodes '{ps} + {as} - ws1104.as.com'  # set union and subtraction
$ fornodes '{ps} * {as}'  # set intersect
atnodes

Run command on clusters. (atnodes calls fornodes internally.)

# run a command on the specified servers:
$ atnodes $'ps -fe|grep httpd' 'ws[1101-1105].as.com'

# multiple-arg command requires "--":
$ atnodes ls /opt/ -- '{ps} + {as}' 'localhost'

# or use single arg command:
$ atnodes 'ls /opt/' '{ps} + {as}' 'localhost' # ditto

# specify a different user name and SSH server port:
$ atnodes hostname '{ps}' -u agentz -p 12345

# use -w to prompt for password if w/o SSH key (no echo back)
$ atnodes hostname '{ps}' -u agentz -w

# or prompt for password if sudo required...
$ atnodes 'sudo apachectl restart' '{ps}' -w

# or specify a timeout:
$ atnodes 'ping foo.com' '{ps}' -t 3
tonodes

Upload local files/directories to remote clusters

$ tonodes /tmp/*.inst -- '{as}:/tmp/'
$ tonodes foo.txt 'ws1105*' :/tmp/bar.txt

# use rsync instead of scp:
$ tonodes foo.txt 'ws1105*' :/tmp/bar.txt -rsync

$ tonodes -r /opt /bin/* -- 'ws[1101-1102].foo.com' 'bar.com' :/foo/bar/
key2nodes

Push the SSH public key (or generate one if not any) to the remote clusters.

$ key2nodes 'ws[1101-1105].as.com'

DESCRIPTION

System administration (sysadmin) is also part of my $work. Playing with a (big) bunch of machines without a handy tool is painful. So I refactored some of our old scripts and hence this module.

This is a high-level abstraction over the powerful Net::OpenSSH module. A bunch of handy scripts are provided to simplify big cluster operations: fornodes, atnodes, tonodes, and key2nodes.

SSH::Batch allows you to name your clusters using variables and interval/set syntax in your ~/.fornodesrc config file. For instance:

$ cat ~/.fornodesrc
A=foo[01-03].com bar.org
B=bar.org baz[a-b,d,e-g].cn foo02.com
C={A} * {B}
D={A} - {B}

where cluster C is the intersection set of cluster A and B while D is the sef of machines that are in A but not in B.

And then you can query machine host list by using SSH::Batch's fornodes script:

$ fornodes '{C}'
bar.org foo02.com

$ fornodes '{D}'
foo01.com foo03.com

$ fornodes blah.com '{C} + {D}'
bar.org blah.com foo01.com foo02.com foo03.com

It's always best practice to put spaces around set operators like +, -, *, and /, so as to allow these characters (notably the dash -) in your host names, as in:

$ fornodes 'foo-bar-[a-d].com - foo-bar-c.com'
foo-bar-a.com foo-bar-b.com foo-bar-d.com

for the ranges like [a-z], there's also an alternative syntax:

[a..z]

To exclude some discrete values from certain range, you need set subtration:

foo[1-100].com - foo[32,56].com

or equivalently

foo[1-31,33-55,57-100].com

fornodes could be very handy in shell programming. For example, to test the 80 port HTTP service of a cluster A, simply put

$ for node in `fornodes '{A}'`; \
    do curl "http://$node:80/blah'; \
  done

Also, other scripts in this module, like atnodes, tonodes, and key2nodes also call fornodes internally so that you can use the cluster spec syntax in those scripts' command line as well.

atnodes meets the common requirement of running a command on a remote cluster. For example:

# at the concurrency level of 6:
atnodes 'ls -lh' '{A} + {B}' my.more.com -c 6

Or upload a local file to the remote cluster:

tonodes ~/my.tar.gz '{A} / {B}' :/tmp/

or multiple files as well as some directories:

tonodes -r ~/mydir ~/mydir2/*.so -- foo.com bar.cn :~/

It's also possible to use wildcards in the cluster spec expression, as in

atnodes 'ls ~' 'api??.*.com'

where atnodes will match the pattern api??.*.com against the "universal set" consisting of those hosts appeared in ~/fornodesrc and those host names apeared before this pattern on the command line (if any). Note that only ? (match any character) and * (match 0 or more characters) are supported here.

There's also a key2nodes script to push SSH public keys to remote machines ;)

TIPS

There's some extra tips found in our own's everyday use:

Running sudo commands

Often, we want to run commands requiring root access, such as when installing software packages on remote machines. So you'll have to tell atnodes to prompt for your password:

$ atnodes 'sudo yum install blah' '{my_cluster}' -w

Then you'll be prompted by the Password: prompt after which you enter your remote password (with echo back turned off).

Because the remote sshd might be smart enough to "remember" the sudo password for a (small) amount of time, immediate subsequent "sudo" might omit the -w option, as in

$ atnodes 'sudo mv ~/foo /usr/local/bin/' {my_cluster}

But remember, you can use sudo without passwords just for a small amount of time ;)

If you see the following error message while doing sudo with atnodes

sudo: sorry, you must have a tty to run sudo

then you should probably comment out the "Defaults requiretty" line in your server's /etc/sudoers file (or just do this for your own account).

Passing custom options to the underlying ssh

By default, atnodes relies on Net::OpenSSH to locate the OpenSSH client executable "ssh". But you can define the SSH_BATCH_SSH_CMD environment to specify the command explicitly. You can use the -ssh option to override it further.

Note that to specify your own "ssh" is also a way to pass more options to the underlying OpenSSH client executable when using atnodes:

$ cat > ~/bin/myssh
#!/bin/sh
# to enable X11 forwarding:
exec ssh -X "$@"

$ chmod +x ~/bin/myssh

$ export SSH_BATCH_SSH_CMD=~/bin/myssh
$ atnodes 'ls -lh' '{my_cluster_name}'

It's important to use "exec" in your own ssh wrapper script, or you may see atnodes hangs.

Use wildcard for cluster expressions to save typing

Wildcards in cluster spec could save a lot of typing. Say, if you have api10.foo.bar.baz.bah.com.cn appeared in your ~/.fornodesrc file:

$ cat ~/.fornodesrc
MyCluster=api[01-22].foo.bar.baz.bah.com.cn

then in case you want to refer to the api10.foo.bar.baz.bah.com.cn node alone on the command line, you can just say api10*, or api10.*.com.cn, or something more specific.

But use wildcards with care. You may have nodes that you don't want in your resulting host list. So it's best practice to use -l option when you use wildcards with atnodes or tonodes, as in

$ atnodes 'rm -rf /opt/blah' 'api10*' -l

So that atnodes will just echos out the exact host list that it would operate on but without doing anything. (It's effectively a "dry-run".) After checking, you can safely remove the -l option and go on.

Specify a different ssh port or user name.

You may have already learned that you can use the -u and -p options to specify a non-default user account or SSH port. But it's also possible and often more convenient to put it as part of your cluster spec expression, either in ~/.fornodesrc or on the command line, as in

$ cat > ~/.fornodesrc
# cluster A uses the default user name:
A=foo[01-25].com
# cluster B uses the non-default user name "jim" and a port 12345
B=jim@foo[26-28].com:12345

$ atnodes 'ls -lh' '{B} + bob@bar[29-31].org:5678'
Use -L to help grepping the outputs by hostname

When managing hundreds or even thousands of machines, it's often more convenient to grep over the outputs of atnodes or tonodes by host names. The -L option makes atnodes and tonodes to prefixing every output lines of the remote commands (if any) by the host name. As in

$ atnodes 'top -b|head -n5' '{my_big_cluster}' -L > out.txt 2>&1
$ grep 'some.specific.host.com' out.txt
Specify a timeout to prevent hanging

It's often wise to specify a timeout for SSH operations. For example, if there's 3 sec of network traffic silence, the following command will quit with an error message printed:

$ atnodes -t 3 'sleep 4' {my_cluster}
Limit the bandwith used by tonodes to be firewall-friendly

You can use the -b option to tell tonodes to use limited bandwidth if your intranet's Firewall is paranoid about your bandwidth use:

$ tonodes my_big_file {my_cluster}:/tmp/ -b 8000

where 8000 is in the unit of Kbits/sec, so it will not transfer faster than 1 MByte/sec.

PREREQUISITES

This module uses Net::OpenSSH behind the scene, so it requires the OpenSSH client executable (usually spelled "ssh") with multiplexing support (at least OpenSSH 4.1). To check your ssh version, use the command:

$ ssh -v

On my machine, it echos

OpenSSH_4.7p1 Debian-8ubuntu1.2, OpenSSL 0.9.8g 19 Oct 2007
usage: ssh [-1246AaCfgKkMNnqsTtVvXxY] [-b bind_address] [-c cipher_spec]
           [-D [bind_address:]port] [-e escape_char] [-F configfile]
           [-i identity_file] [-L [bind_address:]port:host:hostport]
           [-l login_name] [-m mac_spec] [-O ctl_cmd] [-o option] [-p port] [-R [bind_address:]port:host:hostport] [-S ctl_path]
           [-w local_tun[:remote_tun]] [user@]hostname [command]

There's no spesial requirement on the server side ssh service. Even a non-OpenSSH server-side deamon should work as well.

INSTALLATION

perl Makefile.PL
make
make test
sudo make install

Win32 users should replace "make" with "nmake".

SOURCE CONTROL

You can always get the latest SSH::Batch source from its public Git repository:

http://github.com/agentzh/sshbatch/tree/master

If you have a branch for me to pull, please let me know ;)

TODO

  • Cache the parsing and evaluation results of the config file ~/.fornodesrc to somewhere like the fiel ~/.fornodesrc.cached.

  • Abstract the duplicate code found in the scripts to a shared .pm file.

  • Add the fromnodes script to help downloading files from the remote clusters to local file system (maybe grouped by host name).

  • Add the betweennodes script to transfer files between clusters through localhost.

SEE ALSO

fornodes, atnodes, tonodes, key2nodes, SSH::Batch::ForNodes, Net::OpenSSH.

COPYRIGHT AND LICENSE

This module as well as its programs are licensed under the BSD License.

Copyright (c) 2009, Yahoo! China EEEE Works, Alibaba Inc. All rights reserved.

Copyright (C) 2009, Agent Zhang (agentzh). All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

  • Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.

  • Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

  • Neither the name of the Yahoo! China EEEE Works, Alibaba Inc. nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.