I have a perl script which, when destilled a bit, looks like this:
my $randport = int(10000 + rand(1000)); # Random port as other scripts like this run at the same time
my $localip = '192.168.100.' . ($port - 4000); # Don't ask... backwards compatibility
system("ssh -NL $randport:$localip:23 root\@$ip -o ConnectTimeout=60 -i somekey &"); # create the tunnel in the background
sleep 10; # Give the tunnel some time to come up
# Create the telnet object
my $telnet = new Net::Telnet(
Timeout => 10,
Host => 'localhost',
Port => $randport,
Telnetmode => 0,
Errmode => \&fail,
);
# SNIPPED... a bunch of parsing data from $telnet
The thing is that the target $ip is on a link with very unpredictable bandwidth, so the tunnel might come up right away, it might take a while, it might not come up at all. So a sleep is necessary to give the tunnel some time to get up and running.
So the question is: How can i test if the tunnel is up and running? 10 seconds is a really undesirable delay if the tunnel comes up straight away. Ideally, i would like to check if it's up and continue with creating the telnet object once it is, to a maximum of, say, 30 seconds.
Edit: Ping doesn't help me mouch, as the remote end of the tunnel is generally up, but with a very high amount of packetloss
Solved: Extrapolating from the tip suggested by mikebabcock, sleep 10
has been replaced with this block which works like a charm:
my $starttime = time();
while (1)
{
# Check for success
if (system("nc -dzw10 localhost $randport > /dev/null") == 0) { last }
# Check for timeout
if (time() > $starttime + 30) { &fail() }
# 250ms delay before recheck
select (undef, undef, undef, 0.25);
}
Use netcat -- often nc
on Linux systems:
nc -dvzw10 ${HOSTNAME} 23
Works for me, with a response like:
Connection to ${HOSTNAME} 23 port [tcp/telnet] succeeded!
It also returns 0 on success, and is happy with a simple connection after which it goes away.