Essentially I am trying to write a GUI cluster admin tool, whereby text typed into a console is multiplied out to 2 or more terminal windows (and where each window can be typed in individually) to minimise cluster admin headaches.

I have the basic perl outline in my head and the GUI side of it written - i just do not currently know how to control a terminal window in this way.

We have an off-the-shelf tool already but it doesnt support ssh, and it lacks a few "niceties", hence this mini-project.

So you want to run the same command simultaneously on several cluster nodes. Again, I think it would be too much of a headache to open n terminals and tell them all to pass stuff to shells. You would probably be better off using something like Net::SSH and just maintaining one session per node. Maintain an array of references to connections or something like that, and everytime you type a command into your GUI you iterate across the array and pass the command to each connection.


LAI
:eof

After a lot of hunting around on the net I have found out how to do what i want to do; just have to work out whether i not i should be doing it this way in general...

use warnings;
use strict;
use constant TIOCSTI => 29719;

my $fd;
my $c="\n";

open($fd, ">>", "/dev/pts/30") || die "Could not open: $!";

my $val=ioctl($fd, TIOCSTI, $c);

print "not ok\n" unless ($val);
[download]

which will send a RETURN into the pseudo terminal (when run as root, anyhow).

Thanks for your help.

Or even use something like SOAP or XML/RPC to do the server side work. I do this currently over SSL & apache or with the Frontier::RPC2 to collect some system/app stats with great results. I used to do similar things with ssh + keychain, but sometimes i would get some weird hangs on it.

An intellectual is someone whose mind watches itself.
- Albert Camus