Assuming netcat exists can cause problems

[ericfranz]

See:

ood_core/lib/ood_core/batch_connect/template.rb

Lines 120 to 126 in a87f59e

	# Check if port $1 is in use
	port_used () {
	local port="${1#*:}"
	local host=$((expr "${1}" : '\\(.*\\):' || echo "localhost") | awk 'END{print $NF}')
	nc -w 2 "${host}" "${port}" < /dev/null &> /dev/null
	}
	export -f port_used

The assumption here is that netcat is available. If it is not (because the command doesn't exist or the user is not allowed to execute the command) the result will be nonzero but will be reported as the port not yet being in use.

If we are going to use netcat, we should check for its existence and handle the error case that, then use fallback solutions.

Or not use netcat at all. See https://listsprd.osu.edu/pipermail/ood-users/2018-October/000269.html for discussion and some options and Trey's comment:

I’ve used this trick in the past to check if ports are available when bootstrapping MySQL instances in jobs that shared the same node:

if (: < /dev/tcp/127.0.0.1/${MYSQL_PORT}) 2>/dev/null; then
echo "MySQL port ${MYSQL_PORT} already taken."
return 1
fi

[ericfranz]

One solution if lsof is available on the compute nodes to add to the before.sh.erb before find_port is executed:

# override
port_used () {
  local port="${1#*:}"
  local host=$((expr "${1}" : '\\(.*\\):' || echo "localhost") | awk 'END{print $NF}')
  lsof -i :"${port}" &> /dev/null
}
export -f port_used

and the diff of the file looks something like this:

 # Export the module function if it exists
 [[ $(type -t module) == "function" ]] && export -f module

+# override
+port_used () {
+  local port="${1#*:}"
+  local host=$((expr "${1}" : '\\(.*\\):' || echo "localhost") | awk 'END{print $NF}')
+  lsof -i :"${port}" &> /dev/null
+}
+export -f port_used
+
 # Find available port to run server on
 port=$(find_port ${host})

[ericfranz]

If there is a reliable way to check if the pseudo-device /dev/tcp is available, we would check if that exists first, then use it if it does; if not then try something else.

[ericfranz]

Addressed by #159