Getting Started

First some background. When you run rsync is parses its command-line arguments, then it either connects to a remote rsyncd daemon, or runs an rsync on the remote machine via ssh or rsh. At this point there are two rsync processes: the one you invoked and the one on the remote machine. The one on the local machine is called the client, and the one on the remote machine is the server. One side (either the client or server) will send files and the other will receive files. The sending rsync generates a file list and sends it to the receiving side. The receiving rsync will fork a child process.

File::RsyncP does not (yet) have a command-line script that mimics rsync's startup processing. Think of File::RsyncP as one level below the command-line rsync. File::RsyncP implements the client side of the connection, and File::RsyncP knows how to run the remote side (eg, via rsh or ssh) or to connect to a remote rsyncd daemon. File::RsyncP automatically adds the internal --server and --sender options (if necessary) to the options passed to the remote rsync.

To initiate any rsync session the File::RsyncP->new function should be called. It takes a hashref of parameters:

logLevel

An integer level of verbosity. Zero means be quiet, 1 will give some general information, 2 will some output per file, higher values give more output. 10 will include byte dumps of all data read/written, which will make the log output huge.

rsyncCmd

The command to run the remote peer of rsync. By default the rsyncArgs are appended to the rsyncCmd to create the complete command before it is run. This behavior is affected by rsyncCmdType.

rsyncCmd can either be a single string giving the path of the rsync command to run (eg: /bin/rsync) or a list containing the command and arguments, eg:

    rsyncCmd => [qw(
        /bin/ssh -l user host /bin/rsync
    )],
    

or:

    rsyncCmd => ["/bin/ssh", "-l", $user, $host, "/bin/rsync"],
    

Also, rsyncCmd can also be set to a code reference (ie: a perl sub). In this case the code is called without arguments or other processing. It is up to the perl code you supply to exec() the remote rsync.

This option is ignored if you are connecting to an rsyncd daemon.

rsyncCmdType

By default the complete remote rsync command is created by taking rsyncCmd and appending rsyncArgs. This beavhior can be modified by specifying certain values for rsyncCmdType:

rsyncArgs

A list of rsync arguments. The full remote rsync command that is run will be rsyncCmd appended with --server (and optionally --sender if the remote is a sender) and finally all of rsyncArgs.

protocol_version

What we advertize our protocol version to be. Default is 28.

logHandler

A subroutine reference to a function that handles all the log messages. The default is a subroutine that prints the messages to STDERR.

pidHandler

An optional subroutine reference to a function that expects two integers: the pid of the rsync process (ie: the pid on the local machine that is likely ssh) and the child pid when we are receiving files. If defined, this function is called once when the rsync process is forked, and again when the child is forked during receive.

fio

The file IO object that will handle all the file system IO. The default is File::RsyncP::FileIO->new.

This can be replaced with a new module of your choice, or you can subclass File::RsyncP::FileIO.

timeout

Timeout in seconds for IO. Default is 0, meaning no timeout. Uses alarm() and it is the caller's responsbility to catch the alarm signal.

doPartial

If set, a partial rsync is done. This is to support resuming full backups in BackupPC. When doPartial is set, the --ignore-times option can be set on a per-file basis. On each file in the file list, File::RsyncP::FileIO->ignoreAttrOnFile() is called on each file, and this returns whether or not attributes should be ignored on that file. If ignoreAttrOnFile() returns 1 then it's as though --ignore-times was set for that file.

An example of calling File::RsyncP->new is:

    my $rs = File::RsyncP->new({
                logLevel   => 1,
                rsyncCmd => ["/bin/rsh", $host,  "-l", $user, "/bin/rsync"],
                rsyncArgs  => [
                        "--numeric-ids",
                        "--perms",
                        "--owner",
                        "--group",
                        "--devices",
                        "--links",
                        "--ignore-times",
                        "--block-size=700",
                        "--relative",
                        "--recursive",
                        "-v",
                    ],
            });

A fuller example showing most of the parameters and qw() for the rsyncArgs is:

    my $rs = File::RsyncP->new({
                logLevel   => 1,
                rsyncCmd => ["/bin/rsh", $host,  "-l", $user, "/bin/rsync"],
                rsyncArgs  => [qw(
                        --numeric-ids
                        --perms
                        --owner
                        --group
                        --devices
                        --links
                        --ignore-times
                        --block-size=700
                        --relative
                        --recursive
                        -v
                    )],
                logHandler => sub {
                        my($str) = @_;    
                        print MyHandler "log: $str\n";
                    };
                fio        => File::RsyncP::FileIO->new({
                                logLevel   => 1,
                            });
            });

Talking to a remote Rsync

File::RsyncP can talk to a remote rsync using this sequence of functions:

remoteStart(remoteSend, remoteDir)

Starts the remote server by executing the command specified in the rsyncCmd parameter to File::RsyncP->new, together with the rsyncArgs.

If the client is receiving files from the server then remoteSend should be non-zero and remoteDir is the source directory on the remote machine. If the client is sending files to the remote server then remoteSend should be zero and remoteDir is the destination directory on the remote machine. Returns undef on success and non-zero on error.

go(localDir)

Run the client rsync. localDir is the source directory on the local machine if the client is sending files, or it is the destination directory on the local machine if the client is receiving files. Returns undef on success.

serverClose()

Call this after go() to finish up. Returns undef on success.

statsFinal()

This can be optionally called to pickup the transfer stats. It returns a hashref containing elements totalRead, totalWritten, totalSize, plus whatever the FileIO module might add.

abort()

Call this function to abort the transfer.

An example of sending files to a remote rsync is:

    #
    # Send files to remote destDirectory from local srcDirectory
    # by running rsyncCmd with rsyncArgs.
    #
    $rs->remoteStart(0, destDirectory);
    $rs->go(srcDirectory);
    $rs->serverClose;

An example of receiving files from a remote rsync is:

    #
    # Receive files from remote srcDirectory to local destDirectory
    # by running rsyncCmd with rsyncArgs.
    #
    $rs->remoteStart(1, srcDirectory);
    $rs->go(destDirectory);
    $rs->serverClose;

Talking to a remote Rsync daemon

File::RsyncP can connect to a remote Rsync daemon using this sequence of functions:

serverConnect(host, port)

Connect to the Rsync daemon on the given string host and integer port. The port argument is optional and it defaults to 873. On error serverConnect returns a string error message. On success it returns undef.

serverService(module, authUser, authPasswd, authRequired)

Specify which module to use (a "module" is the symbolic name that appears inside "[...]" /etc/rsyncd.conf), the user's credentials (authUser and authPasswd) and whether authorization is mandatory (authRequired). If set to a non-zero value, authRequired ensures that the remote Rsync daemon requires authentication. If necessary, this is to ensure that you don't connect to an insecure Rsync daemon. The auth arguments are optional if the selected rsyncd module doesn't require authentication.

See the rsyncd.conf manual page for more information. For example, if a host called navajo had a /etc/rsyncd.conf contains these lines:

   [test]
           path = /data/test
           comment = test module
           auth users = craig, celia
           secrets file = /etc/rsyncd.secrets
    

and /etc/rsyncd.secrets contained:

    craig:xxx
    

then you could connect to this rsyncd using:

    $rs->serverConnect("navajo", 873);
    $rs->serverService("test", "craig", "xxx", 0);
    

The value of the authRequired argument doesn't matter in this case.

On error serverService returns a string error message. On success it returns undef.

serverStart(remoteSend, remoteDir)

Starts the remote server. If the client is receiving files from the server then remoteSend should be non-zero. If the client is sending files to the remote server then remoteSend should be zero. The remoteDir typically starts with the module name, followed by any directory below the module. Or remoteDir can be just "." to refer to the top-level module directory. Returns undef on success.

go(localDir)

Run the client rsync. localDir is the source directory on the local machine if the client is sending files, or it is the destination directory on the local machine if the client is receiving files. Returns undef on success.

serverClose()

Call this after go() to finish up. Returns undef on success.

abort()

Call this function to abort the transfer.

An example of sending files to a remote rsyncd daemon is:

    #
    # Send files to a remote module from a local srcDirectory by
    # connecting to an rsyncd server.  ($module is the name from
    # /etc/rsyncd.conf.)
    #
    my $port = 873;
    $rs->serverConnect($host, $port);
    $rs->serverService($module, $authUser, $authPasswd);
    $rs->serverStart(0, ".");
    $rs->go(srcDirectory);
    $rs->serverClose;

An example of receiving files from a remote rsyncd daemon is:

    #
    # Receive files from a remote module to local destDirectory by
    # connecting to an rsyncd server.  ($module is the name from
    # /etc/rsyncd.conf.)
    #
    my $port = 873;
    $rs->serverConnect($host, $port);
    $rs->serverService($module, $authUser, $authPasswd);
    $rs->serverStart(1, ".");
    $rs->go(destDirectory);
    $rs->serverClose;