Don't report 0ms on timeout in dns_query

[danielnelson]

Adds result tag and result_code field for results: success, timeout, error. If the result is not success, the query_time_ms field will be omitted.

closes #2548

Required for all PRs:

• Signed CLA.
• Associated README.md updated.
• Has appropriate unit tests.

[discoduck2x]

@danielnelson exited to test this - is there a way to get hold of nightly builds for telegraf?

[danielnelson]

Yes, these packages are updated by the CircleCI build:

https://github.com/influxdata/telegraf#nightly-builds

[randallt]

Why wouldn't you report the timeout value for query_time_ms? I'm trying to use this right now to help debug an issue and if a request times out, I would expect to see a high response time recorded.

[danielnelson]

It has been some time, but I believe it stems back to the changes on http_response where we had an existing field response_time and we didn't want to set it to the timeout as there was no response. The changes made here were based on the earlier change although you could argue that with query_time_ms the field name is valid even if the query times out.

One potential issue with mixing success and timeout together is that when you aggregate the timeout values can skew the results. This is probably not a big problem because you could use a tag to exclude the timeout times.

[randallt]

The timeouts are exactly the thing I'm looking for. At first I didn't even notice them until I saw the log entries. Seems a backward way of doing it.

[danielnelson]

If the query_time_ms is used, one might still miss them as it could appear that they are successfully completing after timeout seconds. I think gaps in the data are probably more noticeable but it depends on how you are viewing the data, and for alerts you really need to watch the result tag.