Need a bash script which checks connectivity between servers

[shinyb]

The script should read from a file the list of hosts and ports to check.

Example:

server-1.dev.internal:443

server-2.prod.internal:25

For each line, the script should check TCP network connectivity between the server it's running on and that host:port combination and, in case it fails, find out the reason.

Possible results for a check:

- Success.

- DNS record does not exist. (cannot resolve hostname)

- Service not running (remote host is not listening on that port)

- Connection rejected (remote host is listening on that port but network connection is rejected)

- Timeout (remote host is listening on that port but network connection is timing out)

Script's output will consist on a list of checks, results and the time it took for each one of them.

Example:

server-1.dev.internal:443 | Success | 3.4s

server-2.prod.internal:25 | Service not running | 0.2s

Note: Script is expecting a text file and is executed as: ./script_name <path-to-file>

[cooper]

`sed s/^/time nc\ -vz\ / <path-to-file> | sed s/:/\ /`

Edit: Above but as a single command:

`sed s/^\\\(\[^:\]*\\\):/time\ nc\ -vz\ \\1\ / <path-to-file>`

Edited November 20, 2014 by Cooper

[Sitwon]

Trying to emulate your example relatively closely...

#!/usr/bin/env bash

TIMEOUT=30

stime(){
	local START="$( date +%s )"
	"$@"
	local END="$( date +%s )"
	TIME=$(( END - START ))
}

get_status(){
	STATUS="$( nc -w "${TIMEOUT}" -z ${DOMAIN} ${PORT} 2>&1 && echo Success )"
	if [ "${STATUS}" != "Success" ]; then
		STATUS="${STATUS##*: }"
	fi
}

report_status(){
	stime get_status
	echo "${STATUS} | ${TIME}s"
}

while read LINE ; do
	DOMAIN="${LINE%%:*}"  # alternatively: "$( cut -d : -f 1 <<<"${LINE}" )"
	PORT="${LINE##*:}"  # alternatively: "$( cut -d : -f 2 <<<"${LINE}" )"
	echo "${DOMAIN}:${PORT} | $( report_status )"
done < "${1}"

The output doesn't exactly match your examples, but it's pretty close. There are many potential improvements that could be made, such as testing servers in parallel rather than sequential, improving the resolution of the timer, or performing additional diagnostics on failed connections. If I get bored today I may continue poking at this and update the thread.

[cooper]

I bow to your superior script-fu. Though wouldn't it be easier in get_status to run the nc command and then test $? to see if it was a success?

[Sitwon]

Easier? I don't know. It would also work. I guess I could have written it like this.

get_status(){
	STATUS="$( nc -w "${TIMEOUT}" -z ${DOMAIN} ${PORT} 2>&1 )"
	if [ "$?" -eq 0 ]; then
		STATUS="Success"
	else
		STATUS="${STATUS##*: }"
	fi
}

Or even like this.

get_status(){
	STATUS="$( nc -w "${TIMEOUT}" -z ${DOMAIN} ${PORT} 2>&1 )" \
		&& STATUS="Success" \
		|| STATUS="${STATUS##*: }"
}

Just up to your style/preference I suppose.

[shinyb]

Thanks for your responses guys, too bad nc "-z" option is not available in my centos 7; it fails with nc: invalid option -- 'z'

I guess I'll have to stick with some other tool.

[cooper]

See here for the -z problem.

https://bugs.launchpad.net/percona-xtradb-cluster/+bug/1349384

Edited November 20, 2014 by Cooper

[cooper]

Easier? I don't know. It would also work. [snip] Just up to your style/preference I suppose.

I was thinking more like this

get_status(){
    nc -w "${TIMEOUT}" -z ${DOMAIN} ${PORT}
    if [ "$?" -eq 0 ]; then
        STATUS="Success"
    else
        STATUS="${STATUS##*: }"
    fi
}

[Sitwon]

I was thinking more like this

get_status(){
    nc -w "${TIMEOUT}" -z ${DOMAIN} ${PORT}
    if [ "$?" -eq 0 ]; then
        STATUS="Success"
    else
        STATUS="${STATUS##*: }"
    fi
}

That would not work. You need to capture the output of nc into the $STATUS variable, otherwise your else clause will just result in $STATUS=="".

[cooper]

Oh, that's what that does...

I once again humbly bow to your superior script-fu.