SSH Backward Connection:

ali2011 · 12-11-2011, 11:58 PM

I have a control unit to about 150 PCs globally distributed, and I am facing a trouble to send WARNING signal to the main machine. the word trouble for me means: A remote host has killed an active iperf process for some reason. It's to difficult to watch or even monitor these troubles manually, and reactivate any killed process using a new shh-connection. What I look to do is as following:

If any node in the grope trying to probe a node that killed its iperf daemon, and received an error message "write failed: Connection refused", then the one received the error message should inform the main machine to reactivate the killed ipref on that node using SHH connection. By the way, EVERY node in doing active probing IS on active SSH session with the main machine (I have no idea about to send this WARNING on this active session). The idea should contain two short scripts, 1st is an active alarm all time on each PC from the 150, and the 2nd is a re-activator at the main machine to reactivate iperf on that particular machine.

I will greatly appreciate your response.

asimba · 12-12-2011, 05:36 AM

my 2 cents - main machine/server can poll all these 150 systems after every n minutes.
or
Machines themselves can check for this process via cron - and restart if need be.
or
And these all 150 machines can run a short custom xinetd daemon on user defined port - which on server request restarts iperf

ali2011 · 12-12-2011, 10:48 AM

Can you explain with an example? I'm not that strong in programming. If the remote machine can check by itself then much efforts will be avoided but how?

asimba · 12-12-2011, 10:59 AM

I can suggest you a bit - you might want to do coding part your

cron - is akin to windows scheduler

ps aux / other switches + grep - let you know if process is running - if running then what state zombie etc.

Also you might want to check for earlier threads

http://www.linuxquestions.org/questi...-start-787605/

Following is code excerpt from same thread - seems a good one.

Code:

#!/bin/bash

SPHINXPID=$(pgrep searchd)


if [[ -z "$SPHINXPID" ]]; then
  /usr/local/sphinx/bin/searchd
fi

unset SPHINXPID