Added options to isolate stdout or stderr only. Added documentation.

[deployable.git] / deployable

#!/usr/bin/env perl
use strict;
use warnings;
use Carp;
use version; our $VERSION = qv('0.0.1');
use Fatal qw( close );
use Pod::Usage qw( pod2usage );
use Getopt::Long qw( :config gnu_getopt );
use English qw( -no_match_vars );
use File::Basename qw( basename dirname );
use File::Spec::Functions qw( file_name_is_absolute catfile );
use POSIX qw( strftime );
use Cwd qw( cwd realpath );
use Archive::Tar;
use File::Find::Rule;
use Data::Dumper;

my %config = (
   output => '-',
   remote => catfile(dirname(realpath(__FILE__)), 'remote'),
   tarfile => [],
   heredir => [],
   rootdir => [],
   root    => [],
   deploy  => [],
);
GetOptions(
   \%config,
   qw(
      usage! help! man! version!

      bundle|all-exec|X!
      cleanup|c!
      deploy|exec|d=s@
      heredir|H=s@
      include-archive-tar|T!
      output|o=s
      root|r=s@
      rootdir|R=s@
      tarfile|tar=s
      workdir|work-directory|deploy-directory|w=s
   ),
);
pod2usage(message => "$0 $VERSION", -verbose => 99, -sections => '')
  if $config{version};
pod2usage(-verbose => 99, -sections => 'USAGE') if $config{usage};
pod2usage(-verbose => 99, -sections => 'USAGE|EXAMPLES|OPTIONS')
  if $config{help};
pod2usage(-verbose => 2) if $config{man};

pod2usage(
   message   => 'working directory must be an absolute path',
   -verbose  => 99,
   -sections => ''
  )
  if exists $config{workdir} && !file_name_is_absolute($config{workdir});

if ($config{'include-archive-tar'}) {
   $config{remote} = catfile(dirname(realpath(__FILE__)), 'remote-at');
   if (! -e $config{remote}) { # "make" it
      print {*STDERR} "### Making remote-at...\n";
      my $startdir = cwd();
      chdir dirname realpath __FILE__;
      system {'make'} qw( make remote-at );
      chdir $startdir;
   }
}

# Establish output channel
my $out_fh = \*STDOUT;
if ($config{output} ne '-') {
   open my $fh, '>', $config{output}    ## no critic
     or croak "open('$config{output}'): $OS_ERROR";
   $out_fh = $fh;
}
binmode $out_fh;

# Emit script code to be executed remotely. It is guaranteed to end
# with __END__, so that all what comes next is data
print {$out_fh} get_remote_script();

# If a tarfile was given, simply put it and exit
if (@{$config{tarfile}}) {
   open my $fh, '<', $config{tarfile}
      or croak "open('$config{tarfile}'): $OS_ERROR";
   print {$out_fh} <$fh>;
   close $fh;
   close $out_fh;
   exit 0;
}

# Where all the data will be kept
my $tar = Archive::Tar->new();

{ # Add a configuration file
   my %general_configuration;
   for my $name (qw( workdir cleanup bundle deploy )) {
      $general_configuration{$name} = $config{$name}
         if exists $config{$name};
   }
   $tar->add_data('deployable/config.pl', Dumper \%general_configuration);
}

# Process files and directories. All these will be reported in the
# extraction directory, i.e. basename() will be applied to them. For
# directories, they will be re-created
for my $file (@ARGV) {
   croak "'$file' not readable" unless -r $file;
   if (-d $file) {
      print {*STDERR} "processing directory '$file'\n";
      save_directory($tar, '.', '.', 'here');
   } ## end if (-d $file)
   else {
      print {*STDERR} "processing file '$file'\n";
      save_file($tar, $file, 'here');
   } ## end else [ if (-d $file)
} ## end for my $file (@ARGV)

# heredir-s are directories that are extracted directly into the ex dir
for my $heredir (@{$config{heredir}}) {
   croak "'$heredir' not readable" unless -r $heredir;
   print {*STDERR} "processing here-directory '$heredir'\n";
   save_directory($tar, $heredir, '.', 'here');
} ## end for my $heredir (@{$config...

# rootdir-s are directories that will go under root
for my $rootdir (@{$config{rootdir}}) {
   croak "'$rootdir' not readable" unless -r $rootdir;
   print {*STDERR} "processing root-directory '$rootdir'\n";
   save_directory($tar, '.', $rootdir, 'root');
}

# root-s are directories whose contents go under root
for my $root (@{$config{root}}) {
   croak "'$root' not readable" unless -r $root;
   print {*STDERR} "processing root-directory '$root'\n";
   save_directory($tar, $root, '.', 'root');
}

# Save tar file, it will close the filehandle as well
$tar->write($out_fh);

# Set as executable
if ($config{output} ne '-') {
   chmod oct(755), $config{output}
     or carp "chmod(0755, '$config{output}'): $OS_ERROR";
}

{
   my $cwd;
   sub save_directory {
      my ($tar, $localstart, $localdir, $remotestart) = @_;

      $cwd ||= cwd();
      chdir $localstart;

      save_file($tar, $_, $remotestart) for File::Find::Rule->file()->in($localdir);

      chdir $cwd;

      return;
   } ## end sub save_directory
}

sub save_file {
   my ($tar, $filename, $remotestart) = @_;

   my ($ftar) = $tar->add_files($filename);
   $ftar->rename("$remotestart/$filename");

   return;
} ## end sub save_file

sub get_remote_script {
   open my $fh, '<', $config{remote}
     or croak "open('$config{remote}'): $OS_ERROR";
   my @lines;
   while (<$fh>) {
      last if /\A __END__ \s*\z/mxs;
      push @lines, $_;
   }
   close $fh;
   return join '', @lines, "__END__\n";
} ## end sub get_remote_script

__END__

=head1 NAME

deployable - create a deploy script for some files/scripts

=head1 VERSION

See version at beginning of script, variable $VERSION, or call

   shell$ deployable --version

=head1 USAGE

   deployable [--usage] [--help] [--man] [--version]

   deployable [--bundle|--all-exec|-X] [--cleanup|-c] 
              [--deploy|--exec|d <program>] [--heredir|-H <dirname>]
              [--include-archive-tar|-T] [--output|-o <filename>]
              [--root|-r <dirname>] [--rootdir|-R <dirname>]
              [--tarfile|--tar|-t <filename>]
              [--workdir|-w <path>] [ files or directories... ]

=head1 EXAMPLES

   # pack some files and a deploy script together.
   shell$ deployable script.sh file.txt some/directory -d script.sh

   # Use a directory's contents as elements for the target root
   shell$ ls -1 /path/to/target/root
   etc
   opt
   usr
   var
   # The above will be deployed as /etc, /opt, /usr and /var
   shell$ deployable -o dep.pl --root /path/to/target/root

   # Include directory /path/to/etc for inclusion and extraction 
   # directly as /etc
   shell$ deployable -o dep.pl --rootdir /path/to/etc

=head1 DESCRIPTION

This is a meta-script to create deploy scripts. The latter ones are
suitable to be distributed in order to deploy something.

You basically have to provide two things: files to install and programs
to be executed. Files can be put directly into the deployed script, or
can be included in gzipped tar archives.

When called, this script creates a deploy script for you. This script
includes all the specified files, and when executed it will extract
those files and execute the given programs. In this way, you can ship
both files and logic needed to correctly install those files, but this
is of course of of scope.

All files and archives will be extracted under a configured path
(see L<--workdir> below), which we'll call I<workdir> from now on. Under
the I<workdir> a temporary directory will be created, and the files
will be put in the temporary directory. You can specify if you want to
clean up this temporary directory or keep it, at your choice. (You're able
to both set a default for this cleanup when invoking deployable, or when
invoking the deploy script itself). The temporary directory will be
called I<tmpdir> in the following.

There are several ways to embed files to be shipped:

=over

=item *

pass the name of an already-prepared tar file via L</--tarfile>. This is
by far the most flexible of the options, but it requires more work on
your side. If you pass this option, all other input files/directories will
be ignored;

=item *

specify the file name directly on the command line. A file given in this
way will always be extracted into the I<tmpdir>, whatever its initial path
was;

=item *

specify the name of a directory on the command line. In this case,
C<tar> will be used to archive the directory, with the usual option to
turn absolute paths into relative ones; this means that directories will
be re-created under I<tmpdir> when extraction is performed;

=item *

give the name of a directory to be used as a "here directory", using
the C<--heredir|-H> option. This is much the same as giving the directory
name (see above), but in this case C<tar> will be told to change into the
directory first, and archive '.'. This means that the contents of the
"here-directory" will be extracted directly into I<tmpdir>.

=back

=head2 Extended Example

Suppose you have a few server which have the same configuration, apart
from some specific stuff (e.g. the hostname, the IP addresses, etc.).
You'd like to perform changes to all with the minimum work possible...
so you know you should script something.

For example, suppose you want to update a few files in /etc, setting these
files equal for all hosts. You would typically do the following:

   # In your computer
   shell$ mkdir -p /tmp/newfiles/etc
   shell$ cd /tmp/newfiles/etc
   # Craft the new files
   shell$ cd ..
   shell$ tar cvzf newetc.tar.gz etc

   # Now, for each server:
   shell$ scp newetc.tar.gz $server:/tmp
   shell$ ssh $server tar xvzf /tmp/newetc.tar.gz -C /


So far, so good. But what if you need to kick in a little more logic?
For example, if you update some configuration files, you'll most likey
want to restart some services. So you could do the following:

   shell$ mkdir -p /tmp/newfiles/tmp
   shell$ cd /tmp/newfiles/tmp
   # craft a shell script to be executed remotely and set the exec bit
   # Suppose it's called deploy.sh
   shell$ cd ..
   shell$ tar cvzf newetc.tar.gz etc tmp

   # Now, for each server:
   shell$ scp newetc.tar.gz $server:/tmp
   shell$ ssh $server tar xvzf /tmp/newetc.tar.gz -C /
   shell$ ssh $server /tmp/deploy.sh

And what if you want to install files depending on the particular machine?
Or you have a bundle of stuff to deploy and a bunch of scripts to execute?
You can use deployable. In this case, you can do the following:

   shell$ mkdir -p /tmp/newfiles/etc
   shell$ cd /tmp/newfiles/etc
   # Craft the new files
   shell$ cd ..
   # craft a shell script to be executed remotely and set the exec bit
   # Suppose it's called deploy.sh
   shell$ deployable -o deploy.pl etc deploy.sh --exec deploy.sh

   # Now, for each server
   shell$ scp deploy.pl $server:/tmp
   shell$ ssh $server /tmp/deploy.pl

And you're done. This can be particularly useful if you have another
layer of deployment, e.g. if you have to run a script to decide which
of a group of archives should be deployed. For example, you could craft
a different new "etc" for each server (which is particularly true if
network configurations are in the package), and produce a simple script
to choose which file to use based on the MAC address of the machine. In
this case you could have:

=over

=item newetc.*.tar.gz

a bunch of tar files with the configurations for each different server

=item newetc.list

a list file with the association between the MAC addresses and the
real tar file to deploy from the bunch in the previous bullet

=item deploy-the-right-stuff.sh

a script to get the real MAC address of the machine, select the right
tar file and do the deployment.

=back

So, you can do the following:

   shell$ deployable -o deploy.pl newetc.*.tar.gz newetc.list \
      deploy-the-right-stuff.sh --exec deploy-the-right-stuff.sh

   # Now, for each server:
   shell$ scp deploy.pl $server:/tmp
   shell$ ssh $server /tmp/deploy.pl

So, once you have the deploy script on the target machine all you need
to do is to execute it. This can come handy when you cannot access the
machines from the network, but you have to go there physically: you
can prepare all in advance, and just call the deploy script.


=head1 OPTIONS

Meta-options:

=over

=item B<--help>

print a somewhat more verbose help, showing usage, this description of
the options and some examples from the synopsis.

=item B<--man>

print out the full documentation for the script.

=item B<--usage>

print a concise usage line and exit.

=item B<--version>

print the version of the script.

=back

Real-world options:

=over

=item B<< --bundle | --all-exec | -X >>

Set bundle flag in the produced script. If the bundle flag is set, the
I<deploy script> will treat all executables in the main deployment
directory as scripts to be executed.

By default the flag is not set.

=item B<< --cleanup | -c >>

Set cleanup flag in the produced script. If the cleanup flag is set, the
I<deploy script> will clean up after having performed all operations.

You can set this flag to C<0> by using C<--no-cleanup>.

=item B<< --deploy | --exec | -d <filename> >>

Set the name of a program to execute after extraction. You can provide
multiple program names, they will be executed in the same order.

=item B<< --heredir | -H <path> >>

Set the name of a "here directory" (see L<DESCRIPTION>). You can use this
option multiple times to provide multiple directories.

=item B<< --include-archive-tar | -T >>

Embed L<Archive::Tar> (with its dependencies L<Archive::Tar::Constant> and
L<Archive::Tar::File>) inside the final script. Use this when you know (or
aren't sure) that L<Archive::Tar> will not be available in the target
machine.

=item B<< --output | -o <filename> >>

Set the output file name. By default the I<deploy script> will be given
out on the standard output; if you provide a filename (different from
C<->, of course!) the script will be saved there and the permissions will
be set to 0755.

=item B<< --root | -r <dirname> >>

Include C<dirname> contents for deployment under root directory. The
actual production procedure is: hop into C<dirname> and grab a tarball
of C<.>. During deployment, hop into C</> and extract the tarball.

This is useful if you're already building up the absolute deployment
layout under a given directory: just treat that directory as if it were
the root of the target system.

=item B<< --rootdir | -R <dirname >>

Include C<dirname> as a directory that will be extracted under root
directory. The actual production procedure is: grab a tarball of
C<dirname>. During deployment, hop into C</> and extract the tarball.

This is useful if you have a directory (or a group of directories) that
you want to deploy directly under the root.

=item B<< --tarfile | --tar | -t <filename> >>

Use the given file as the only contents for the deployed script. The
filename should point to a valid tar file, which will be the only
one stored into the resulting script (i.e. all other files/directories
will be ignored). This requires more work on your side, but gives you
full flexibility.

Note that the generated script will interpret the generated tar file
as follows:

=over

=item B<< deployable/config.pl >> (file)

will be C<eval>ed, expecting to receive a reference to an anonymous
hash with the configurations;

=item B<< here >> (directory)

will be normally deployed into the working directory in the target
system;

=item B<< root >> (directory)

will be normally deployed under C</> in the target system.

=back

=item B<< --workdir | --deploy-directory | -w <path> >>

Set the working directory for the deploy.

=back

=head1 THE DEPLOY SCRIPT

The net result of calling this script is to produce another script,
that we call the I<deploy script>. This script is made of two parts: the
code, which is fixed, and the configurations/files, which is what is
actually produced. The latter part is put after the C<__END__> marker,
as usual.

Stuff in the configuration part is always hexified in order to prevent
strange tricks or errors. Comments will help you devise what's inside the
configurations themselves.

The I<deploy script> has options itself, even if they are quite minimal.
In particular, it supports the same options C<--workdir|-w> and
C<--cleanup> described above, allowing the final user to override the
configured values. By default, the I<workdir> is set to C</tmp/our-deploy>
and the script will clean up after itself.

The following options are supported in the I<deploy script>:

=over

=item B<--usage | --man | --help>

print a minimal help and exit

=item B<--version>

print script version and exit

=item B<--bundle | --all-exec | -X>

treat all executables in the main deployment directory as scripts
to be executed

=item B<--cleanup | --no-cleanup>

perform / don't perform temporary directory cleanup after work done

=item B<< --deploy | --no-deploy >>

deploy scripts are executed by default (same as specifying '--deploy')
but you can prevent it.

=item B<--dryrun | --dry-run>

print final options and exit

=item B<< --filelist | --list | -l >>

print a list of files that are shipped in the deploy script

=item B<< --inspect <dirname> >>

just extract all the stuff into <dirname> for inspection. Implies
C<--no-deploy>, C<--no-tempdir>, ignores C<--bundle> (as a consequence of
C<--no-deploy>), disables C<--cleanup> and sets the working directory
to C<dirname>

=item B<--show | --show-options | -s>

print configured options and exit

=item B<< --tar | -t >>

print out the tar file that contains all the shipped files, useful
to redirect to file or pipe to the B<tar> program

=item B<< --tempdir | --no-tempdir >>

by default a temporary directory is created (same as specifying
C<--tempdir>), but you can execute directly in the workdir (see below)
without creating it.

=item B<--workdir | --work-directory | --deploy-directory | -w>

working base directory (a temporary subdirectory will be created
there anyway)

=back

Note the difference between C<--show> and C<--dryrun>: the former will
give you the options that are "embedded" in the I<deploy script> without
taking into account other options given on the command line, while the
latter will give you the final options that would be used if the script
were called without C<--dryrun>.

=head2 Deploy Script Example Usage

In the following, we'll assume that the I<deploy script> is called
C<deploy.pl>.

To execute the script with the already configured options, you just have
to call it:

   shell$ ./deploy.pl

If you just want to see which configurations are in the I<deploy script>:

   shell$ ./deploy.pl --show

To see which files are included, you have two options. One is asking the
script:

   shell$ ./deploy.pl --filelist

the other is piping to tar:

   shell$ ./deploy.pl --tar | tar tvf -

Extract contents of the script in a temp directory and simply inspect
what's inside:

   # extract stuff into subdirectory 'inspect' for... inspection
   shell$ ./deploy.pl --no-tempdir --no-deploy --workdir inspect

=head2 Deploy Script Requirements

You'll need a working Perl with version at least 5.6.2.

If you specify L</--include-archive-tar>, the module L<Archive::Tar> will
be included as well. This should ease your life and avoid you to have
B<tar> on the target machine. On the other hand, if you already know
that B<tar> will be available, you can avoid including C<Archive::Tar>
and have the generated script use it (it could be rather slower anyway).

=head1 DIAGNOSTICS

Each error message should be enough explicit to be understood without the
need for furter explainations. Which is another way to say that I'm way
too lazy to list all possible ways that this script has to fail.


=head1 CONFIGURATION AND ENVIRONMENT

deployable requires no configuration files or environment variables.

Please note that deployable B<needs> to find its master B<remote> file
to produce the final script. This must be put in the same directory where
deployable is put. You should be able to B<symlink> deployable where you
think it's better, anyway - it will go search for the original file
and look for B<remote> inside the same directory. This does not apply to
hard links, of course.


=head1 DEPENDENCIES

All core modules, apart the following:

=over

=item B<< Archive::Tar >>

=item B<< File::Find::Rule >>

=back

=head1 BUGS AND LIMITATIONS

No bugs have been reported.

Please report any bugs or feature requests to the AUTHOR below.

Be sure to read L<CONFIGURATION AND ENVIRONMENT> for a slight limitation
about the availability of the B<remote> script.

=head1 AUTHOR

Flavio Poletti C<flavio [AT] polettix.it>


=head1 LICENSE AND COPYRIGHT

Copyright (c) 2008, Flavio Poletti C<flavio [AT] polettix.it>. All rights reserved.

This script is free software; you can redistribute it and/or
modify it under the same terms as Perl itself. See L<perlartistic>
and L<perlgpl>.

Questo script è software libero: potete ridistribuirlo e/o
modificarlo negli stessi termini di Perl stesso. Vedete anche
L<perlartistic> e L<perlgpl>.


=head1 DISCLAIMER OF WARRANTY

BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER
EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE
ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH
YOU. SHOULD THE SOFTWARE PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
NECESSARY SERVICING, REPAIR, OR CORRECTION.

IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE
LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL,
OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE
THE SOFTWARE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING
RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A
FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF
SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES.

=head1 NEGAZIONE DELLA GARANZIA

Poiché questo software viene dato con una licenza gratuita, non
c'è alcuna garanzia associata ad esso, ai fini e per quanto permesso
dalle leggi applicabili. A meno di quanto possa essere specificato
altrove, il proprietario e detentore del copyright fornisce questo
software "così com'è" senza garanzia di alcun tipo, sia essa espressa
o implicita, includendo fra l'altro (senza però limitarsi a questo)
eventuali garanzie implicite di commerciabilità e adeguatezza per
uno scopo particolare. L'intero rischio riguardo alla qualità ed
alle prestazioni di questo software rimane a voi. Se il software
dovesse dimostrarsi difettoso, vi assumete tutte le responsabilità
ed i costi per tutti i necessari servizi, riparazioni o correzioni.

In nessun caso, a meno che ciò non sia richiesto dalle leggi vigenti
o sia regolato da un accordo scritto, alcuno dei detentori del diritto
di copyright, o qualunque altra parte che possa modificare, o redistribuire
questo software così come consentito dalla licenza di cui sopra, potrà
essere considerato responsabile nei vostri confronti per danni, ivi
inclusi danni generali, speciali, incidentali o conseguenziali, derivanti
dall'utilizzo o dall'incapacità di utilizzo di questo software. Ciò
include, a puro titolo di esempio e senza limitarsi ad essi, la perdita
di dati, l'alterazione involontaria o indesiderata di dati, le perdite
sostenute da voi o da terze parti o un fallimento del software ad
operare con un qualsivoglia altro software. Tale negazione di garanzia
rimane in essere anche se i dententori del copyright, o qualsiasi altra
parte, è stata avvisata della possibilità di tali danneggiamenti.

Se decidete di utilizzare questo software, lo fate a vostro rischio
e pericolo. Se pensate che i termini di questa negazione di garanzia
non si confacciano alle vostre esigenze, o al vostro modo di
considerare un software, o ancora al modo in cui avete sempre trattato
software di terze parti, non usatelo. Se lo usate, accettate espressamente
questa negazione di garanzia e la piena responsabilità per qualsiasi
tipo di danno, di qualsiasi natura, possa derivarne.

=cut