PHP library for executing commands on multiple remote machines, via SSH

I can see this being a helpful tool for lots of developers, but there's also grunt.js, which is super popular and dominates the name "grunt". I don't really see why the package is not named differently. Even the gruntfile is the same name, so I doubt this is a coincidence, right?

My two cents would be to rename the package and respect the other tool's namespace, even if it is from another language.

Can you really only define a single host per name? How are you expected to deploy to an entire live cluster without being able to define an array of target hosts?

@philsturgeon

It's been nearly 3 years since the last commit on this project, and two issues I raised in December haven't had a response.

Can you please clarify the status of this project, as I may need to find an alternative if there is no longer support for this library.

Many thanks!

Keep getting this error "[1] 73643 segmentation fault shunt print_php_info production" when trying to do a remote "ls" as shown in the README. Using SSH key to connect. Is there a log file or anything I can use to debug?

hello, i have this error :

 php composer.phar update
Loading composer repositories with package information
Updating dependencies (including require-dev)
Your requirements could not be resolved to an installable set of packages.

  Problem 1
    - league/shunt v2.1.1 requires ext-ssh2 >=0.12 -> the requested PHP extension ssh2 is missing from your system.
    - league/shunt v2.1.0 requires ext-ssh2 >=0.12 -> the requested PHP extension ssh2 is missing from your system.
    - league/shunt v2.0.0 requires ext-ssh2 >=0.12 -> the requested PHP extension ssh2 is missing from your system.
    - Installation request for league/shunt ~2.0 -> satisfiable by league/shunt[v2.0.0, v2.1.0, v2.1.1].

how to fix this please ?

Added ability to type hint a Grunt task so that syntax like:

<?php

use League\Grunt\Grunt;

return [
    'hosts' => [
        'production' => 'foo.com'
    ],
    'tasks' => [
        'deploy' => function(Grunt $g) {
            $g->run('ls');
        }
    ],
];

Can be used. This gives the ability to autocomplete methods from an IDE, as well as aid in using debuggers.

Excuse my ignorance, but still did not understand how to have the script to work, put I saved the Shuntfile file and how do I call the task? Would someone send me an example?

I thank everyone's attention.

According to the description, Shunt is a

PHP library for executing commands in parallel on multiple remote machines, via SSH

As far as I can tell, nothing is run in parallel. The commands are run one at a time (desirable, I would assume), and each machine is run one at a time (undesirable, IMHO). So nothing seems to be run in parallel.

Example Shuntfile:

<?php

return [
    'hosts' => [
        'production' => 'staging.ussig.net',
        'production2' => 'dev.ussig.net',
    ],
    'auth' => [
        //...
    ],
    'tasks' => [
        'deploy' => function(\League\Shunt\Shunt $s) {
            $releasePath = '/var/www/deploytohere/releases/' . (new \DateTime())->format('YmdHis');
            $s->run('mkdir -p ' . $releasePath);
            $s->run('git clone repo_path ' . $releasePath);
            $s->run('cd ' . $releasePath . ' && php composer.phar install');
            echo "Finished:" . date('d M Y H:i:s') . "\n";
        }
    ],
];

Output:

$ vendor/bin/shunt deploy production,production2
#1. Running "deploy" task on "production"
Connecting...
staging.ussig.net < mkdir -p /var/www/deploytohere/releases/20140306201644
staging.ussig.net > [OK]
staging.ussig.net < git clone [email protected]:mrkrstphr/sales.git /var/www/deploytohere/releases/20140306201644
staging.ussig.net > Cloning into '/var/www/deploytohere/releases/20140306201644'...
                  > Checking connectivity... done
staging.ussig.net < cd /var/www/deploytohere/releases/20140306201644 && php composer.phar install
staging.ussig.net > Warning: This development build of composer is over 30 days old. It is recommended to update it by running "composer.phar self-update" to get the latest version.
                  > Loading composer repositories with package information
[...snip...]
                  > Writing lock file
                  > Generating autoload files
Finished:06 Mar 2014 20:17:16
Done.

#2. Running "deploy" task on "production2"
Connecting...
dev.ussig.net < mkdir -p /var/www/deploytohere/releases/20140306201716
dev.ussig.net > [OK]
dev.ussig.net < git clone [email protected]:mrkrstphr/sales.git /var/www/deploytohere/releases/20140306201716
dev.ussig.net > Cloning into '/var/www/deploytohere/releases/20140306201716'...
              > Checking connectivity... done
dev.ussig.net < cd /var/www/deploytohere/releases/20140306201716 && php composer.phar install
dev.ussig.net > Warning: This development build of composer is over 30 days old. It is recommended to update it by running "composer.phar self-update" to get the latest version.
              > Loading composer repositories with package information
[...snip...]
                  > Writing lock file
                  > Generating autoload files
Finished:06 Mar 2014 20:19:43
Done.

I'm having trouble installing the ssh2 extension, both on my work computer and at home. I feel this might be an issue for more users.

Perhaps Shunt could use a pure PHP SSH2 implementation like phpseclib if the extension is not available. It also supports SSH2, SFTP and SCP.

I've looked at the source code and it seems it is very implementation-dependent by using constants to dynamically call ssh2-specific functions, for example. Making Shunt SSH2 implementation-independent seems cumbersome.

Any thoughts on whether this is a pursuable notion?

I may have found a bug in the processing of the auth configuration of the Shuntfile recipe.

With this config:

<?php

return [
    //hosts

    'auth' => [
          'username' => '<username>',
          'privkeyfile' => '/home/<username>/.ssh/id_rsa',
          'pubkeyfile'  => '/home/<username>/.ssh/id_rsa.pub',
     ],
   
     // tasks
];

I got the following error when running my shunt task:

SSH authorization failed. REASON : ssh2_auth_pubkey_file(): Authentication failed for <username> using public key: Invalid public key, too short

To get it to work, I swapped the private / public key arguments like this:

<?php

return [
    //hosts

    'auth' => [
          'username' => '<username>',
          'privkeyfile' => '/home/<username>/.ssh/id_rsa.pub',
          'pubkeyfile'  => '/home/<username>/.ssh/id_rsa',
     ],

     // tasks
];

It appears that Shunt is not interpreting the arguments correctly, as I don't believe the above configuration is correct.

I'm using Shunt v2.1.1

Is it possible to pass custom arguments to the shunt script?

For example, consider a use-case where shunt is included as part of a CI pipeline, and the code deploy tag is determined dynamically. In essence doing something like the following:

vendor/bin/shunt some_task production --tag=v1.0.0

and later access the tag argument within a task like this:

<?php

return [
    //hosts
    //auth
    'tasks' => [
         'some_task' => function($s) {
             $tag = $s->option('tag');
          }
     ]
];

Hi,

first i want to say, thanks for that awesome piece of code :+1:

We've the requirement, that we want to execute a remote command by invoking the run() method and need to have access to the result.

<?php

return array(

    'hosts' => array(
        'jenkins' => '127.0.0.1',
    ),

    'auth' => array(
        'username' => 'root',
        'password' => NULL,
        'pubkeyfile' => '/home/user/.ssh/id_rsa.pub',
        'privkeyfile' => '/home/user/.ssh/id_rsa',
        'passphrase' => NULL,
    ),

    'tasks' => array(
        'download_test' => function($s) {

            $local = '';
            if (is_file('/tmp/test_encoded.php')) {
                $local = md5_file('/tmp/test_encoded.php');
            }

            $remote = '';
            $resultHandler = function ($resultDio) use (&$remote) {
                $result = explode('  ', $resultDio);
                $remote = array_shift($result);
            };

            $s->run('md5sum test_encoded.php', false, $resultHandler);

            if ($local === $remote) {
                error_log(sprintf('Hashes ARE equal %s/%s', $local, $remote));
            } else {
                error_log(sprintf('Hashes are NOT equal %s/%s', $local, $remote));
                $s->scp()->get('test_encoded.php', '/tmp/test_encoded.php');
            }
        }
    ),
);

The pull request added a third parameter that allows you to specify a closure that can be used to access the command's result. Actually i think there is no possibility to do this, right?

After using your config as starting point to get the ssh2 extension available in travis, I've just realized that:

• it's not required to track an empty file
• you don't need to enable the already installed extension
• the check step is redundant

So, here you have a PR just to say thanks!

This makes it possible to use agent authentication for hosts by specifying an agent key in the authentication section:

'auth' => array(
    'username' => 'shunt',
    'agent' => true,
)

A username must be provided and the key must be registered with the SSH agent.

First, thanks for the extraordinary package :)

Secondly, inside my tasks, I would like to access the raw output from the SSH command that was run so I can make decisions. After a couple hours of trying to creatively inject my custom CapturedOutput outputter into the doRun() function, I realized that this outputter is not used for command output, since the Shunt object has its own outputter that it got during instantiation. It seems to me that getting the raw output would be useful for others as well, so why not do one of these:

1. Save the raw output in the Shunt object, then provide access so the task can run() something, then grab the output.
2. Add a third parameter to Shunt::run() called OutputInterface $command_output so we can pass in a custom Outputter to grab the raw output of the command.
3. Add the raw output to the ArbitraryCommand itself, along with the return value and maybe some timing information so you can do ::getDuration() ::getStartTime() ::getEndTime(), etc, from the command.

I think no. 1 would be the most convenient, what you do you think?

Here's my very basic CapturedOutput class that I referenced:

    <?php

    namespace ScientiaMobile\UserManager\Output;

    use Symfony\Component\Console\Formatter\OutputFormatterInterface;
    use Symfony\Component\Console\Output\Output;

    class CapturedOutput extends Output {

            private $output;

            public function __construct($verbosity = self::VERBOSITY_NORMAL) {
                    $this->flushOutput();
                    parent::__construct($verbosity, false, null);
            }

            public function getOutput() {
                    return $this->output;
            }

            public function getOutputLines() {
                    return explode("\n", $this->output);
            }

            public function flushOutput() {
                    $this->output = '';
            }

            protected function doWrite($message, $newline) {
                    $this->output .= $message;
                    if ($newline) $this->output .= "\n";
            }
    }