Input net_response does not return measurements when target port is not open.

[ceseuron]

Bug report

Similar to another bug #278, the net_response listener does not return a value for response time when the target port being tested is not open or becomes closed due to a service failure or network problem.

Note that I went the net_response route because of the bug mentioned in #2784 and have essentially wound up in the same boat with a different input.

Relevant telegraf.conf:

[[inputs.net_response]]
protocol = "tcp"
address = ":443"
timeout = "1s"
read_timeout = "1s"
[[inputs.net_response]]
protocol = "tcp"
address = ":80"
timeout = "1s"
read_timeout = "1s"

System info:

Telegraf 1.2.1
InfluxDB 1.2.2

Metrics Target OS: CentOS 7

Steps to reproduce:

1. Have a server running an application or service on a known port. In my case, it's NGINX running over TCP port 80 and 443. Make sure the service is running and the ports are up.
2. Setup Telegraf net_response to test the service ports.
3. Verify that telegraf is returning the response_time to InfluxDB.
4. Stop the service using the target port or make that port otherwise unreachable.
5. Note that Telegraf is no longer returning a metric for net_response and that only the last good measurement is available.

Expected behavior:

Telegraf should return metrics to InfluxDB indicating a problem.

Actual behavior:

Telegraf returns no data.