Ruby interface to remctl

OVERVIEW

The Ruby interface to remctl provides Ruby bindings to the libremctl client library plus a high-level interface that translates the libremctl API into something closer to idiomatic Ruby.

This module provides two interfaces: a simple interface in Ruby that performs a single call to a remctl server and returns the result as an object; and a full interface in Ruby which provides more control over the connection, returns individual output tokens, and allows multiple commands to be sent via the same connection.

REQUIREMENTS

The module has been tested with Ruby 1.8 and 1.9.1 and may not work with older (or newer) versions of Ruby.

SIMPLIFIED INTERFACE

Remctl.remctl(host, *args)

Runs a command on the remote system and returns an object containing the results.

Arguments:

In order to change the port or principal used, assign to Remctl.default_port and Remctl.default_principal. These default to 0 and nil, meaning to respect the library defaults: port 4373 then fall back to attempting the legacy 4444 port, and a principal of host/<host>, with the realm determined by the domain-realm mapping.

args must be an array of things which respond to #to_s.

Returns an object of type Remctl::Result with the following attributes:

Exceptions:

The message attribute (or string value) of the Remctl::Error exception contains the string returned by the remctl library. This may be an internal error (such as a network error) or an error returned by the remote server (such as an unknown command error).

Here is an example using the simplified interface:

    require 'remctl'
    command = %w{testing 1 2 3}
    result = Remctl.remctl("foo.example.com", *command)
    if result.stdout != "":
        puts "stdout: #{result.stdout}"
    if result.stderr != "":
        puts "stderr: #{result.stderr}"
    puts "exit status: #{result.status}"

FULL INTERFACE

The full remctl interface requires the user to do more bookkeeping, but provides more flexibility and visibility into what is happening at a protocol level. It allows issuing multiple commands on the same persistent connection (provided that the remote server supports protocol version two; if it doesn't, the library will transparently fall back to opening a connection for each command).

To use the full interface, first create a connection object with Remctl.new and then call the #command method to issue a command. Read output tokens with #output until a status token has been received. Then the command is complete and another command can be issued.

Remctl.new(host, port, principal)

The constructor. Create a new Remctl object. The host argument is required. port and principal default to Remctl.default_port and Remctl.default_principal respectively.

Arguments:

This method immediately connects to the server. It raises ArgumentError in the event that the port is out of range and Remctl::Error in case of a protocol error.

If port is not given, the library default is used (try 4373 first and fall back to attempting the legacy 4444 port). If principal is not given, the library default (host/<host> with the realm determined by the domain-realm mapping) is used.

To override the credential cache used by GSS-API, assign the credential cache (as a string) to Remctl.ccache before calling Remctl.new. If the remctl client library was built against a Kerberos library and the GSS-API library supported gss_krb5_import_cred, this will only affect subsequently-created Remctl objects. Otherwise, this will change the default credential cache for all uses of GSS-API in the same process, including other remctl objects, rather than just this remctl object due to a limitation in the GSS-API. Some GSS-API implementations do not support setting the credential cache, in which case Remctl.new will throw a Remctl::Error exception if this variable is set.

Normally, the source IP used to connect to the server is assigned by the kernel. To use a specific source IP, assign that IP address (as a string) to Remctl.source_ip before calling Remctl.new.

To set a default timeout for any newly-created Remctl objects, assign the timeout (in seconds) to Remctl.timeout before calling Remctl.new. See below under r.set_timeout for more information about how the timeout is interpreted.

Exceptions:

The message attribute (or string value) of the Remctl::Error exception contains the string returned by the remctl library remctl_error() function.

All further methods below must be called on a Remctl object as returned or yielded by the constructor.

r.command(*args)

Send a command to the remote host. There is no return value; an exception is thrown on any error.

The Remctl object must already be connected. The command may, under the remctl protocol, contain any character, but be aware that most remctl servers will reject commands or arguments containing ASCII 0 (NUL) unless that argument is configured on the server to be passed via standard input.

Arguments:

Exceptions:

The message attribute (or string value) of the Remctl::Error exception contains the string returned by the remctl library.

r.output()

Reads an output token from the server and returns it as an array. A command will result in either one error token or zero or more output tokens followed by a status token. The output is complete as soon as any token other than an output token has been received, but the module will keep returning done tokens to the caller for as long as #output is called without another call to #command.

The members of the returned array are, in order:

The array always has 5 elements. The other members of the array may be nil depending on the type of token. Output tokens will have output and stream information, error tokens will have output and error information, and status tokens will have status information. Done tokens will return nil for all other elements.

For error tokens, error holds the numeric error code (see the remctl protocol specification), which is the recommended value for programs to check when looking for specific errors. output will contain an English text translation of the error code and the exact text may change.

Exceptions:

Errors from the server are returned as an error token, not as an exception.

r.noop()

Send a NOOP message to the server and read the reply. This is primarily used to keep a connection to a remctl server alive, such as through a firewall with a session timeout, while waiting to issue further commands. Returns true on success, false on failure.

The NOOP message requires protocol version 3 support in the server, so the caller should be prepared for this function to fail, indicating that the connection could not be kept alive and possibly that it was closed by the server. In this case, the client will need to explicitly reopen the connection with open().

Exceptions:

Arguments:

If the timeout is 0, this clears any timeout that was previously set.

The timeout is a timeout on network activity from the server, not on a complete operation. So, for example, a timeout of ten seconds just requires that the server send some data every ten seconds. If the server sends only tiny amounts of data at a time, the complete operation could take much longer than ten seconds without triggering the timeout.

Exceptions:

Exceptions:

HISTORY

The original implementation was written by Anthony Martinez <twopir@nmt.edu> in 2010. As of remctl 2.16 it is part of the stock remctl distribution.

LICENSE

Copyright 2010, 2011, 2012 The Board of Trustees of the Leland Stanford Junior University

Copying and distribution of this file, with or without modification, are permitted in any medium without royalty provided the copyright notice and this notice are preserved. This file is offered as-is, without any warranty.

Converted to XHTML by faq2html version 1.33