Depending on how your network setup actually is, most likely the sanest approach is to create an SSH tunnel and then connect DBI to the local end of that tunnel.

From a quick look through Net::SSH2, it doesn't seem that it readily supports setting up a tunnel. I think you can use the ->tcpip method to forward packets yourself, but that feels like far too much programming.

Personally, I would simply launch ssh -L local:remote:remoteport or however one sets up an ssh tunnel from the command line, and then direct DBI / DBD::mysql to connect to the local port. Most likely this is what HeidiSQL does as well.

After a quick search, this is almost exactly what you stated in this post in this thread nearly three years ago :)

From what I can tell, not much has changed.

I would really like there to be a nicer/more integrated solution, but I think with libssh2, you have to do the packet forwarding yourself, which means that you will need at least threads or some very clever asynchronous framework to run both the forwarding and DBI queries. And spawning an external process is much simpler than that, at least if you have the ssh executable available.

thanks i decided to install cygwin and open ssh in shell with perl. hopefully that will work. UPDATE: I apparently do not know how to use "open" to open an SSH connection. I tried:

open(SSH, "ssh user\@host.com");
[download]

as a simple example, and I have tested that syntax will work in Cygwin except it will prompt for password. I don't think I understood what you had previously.

I would either set up a passwordless keypair or use ssh-agent to store the password credentials.

Because the right way to do connect to a remote DB through a SSH gateway would be to call socketpair, attach the command ssh $remote_host -W${db_host}:${db_port} to one side and then pass the other to the DBD::mysql.

Here's some skeleton code to try. I haven't tested it much, but it seems to connect alright.

#! /usr/bin/perl

use strict;
use warnings;

use Socket;
use POSIX;

sub named_socket_spawn {
    my $code = shift;
    ref $code eq 'CODE' or return;

    my ($serv, $conn);
    socket($serv, AF_UNIX, SOCK_STREAM, PF_UNSPEC)
                                        || die "socket: $!";

    my $sname = tmpnam();
    bind($serv, sockaddr_un($sname))    || die "bind: $!";
    listen($serv, 1)                    || die "listen: $!";

    my $pid = fork()                    // die "fork: $!";

    if ($pid) {
        close $serv;
        return $sname;
    }
    accept($conn, $serv)                || warn "accept: $!";
    unlink($sname)                      || warn "unlink: $!";
    open STDIN, "<&", $conn;
    open STDOUT, ">&", $conn;
    close $serv;
    close $conn;
    exit $code->();
}

use DBD::mysql;

#my $sqlsock = named_socket_spawn(sub{exec "ssh host -W host:3306"});
my $sqlsock = named_socket_spawn(sub{ exec "nc localhost 3306" });

my $dsn = "DBI:mysql:database=test;mysql_socket=$sqlsock";

my $dbh = DBI->connect($dsn, "test", "hunter5");
[download]

Hm. According to DBD::mysql, there's the mysql_socket specifier to pass the unix socket name. Does this not work?

So basically, you'd need a small helper sub to handle the tedium of attaching a process to a socket

my $unixname = named_sock_spawn(sub{ exec "ssh" });

sub named_sock_spawn {
    socket();bind();listen();
    fork() and return $name;
    accept();dup();dup();$fun->();
}
[download]

The perlipc has an example on unix-domain TCP. But are there any modules to provide named pipe utilities?

This is only for a socket name unfortunately, not for an in-memory socket structure.