Add support for considering connection / timeout related errors as UNKNOWN state

[atc0005]

Overview

Because the stock check_http plugin considers connection refusals and timeouts as CRITICAL, this plugin adopted the same behavior when it was initially developed.

The check_cert plugin continues to consider connection refusals and timeouts as CRITICAL state, but there are cases where it could be useful to (optionally) force those issues to be considered as UKNOWN state.

For example, having this support could allow a sysadmin to configure only expirations (approaching, exceeded) as actionable, effectively ignoring connection issues by deselecting UNKNOWN state as triggering notifications. Service checks configured in this way would presume that connection related issues are temporary, relying on a separate service check (e.g., HTTPS GET/POST/whatever) to be responsible for alerting if normal client connections are unsuccessful.

With this setup, less "noise" would be generated from service checks intended to trigger only when certificates are approaching expiration (or failing some other enforced validation requirement).

References

• https://nagios-plugins.org/doc/man/check_http.html
• Audit plugin exit states to determine if another state (e.g., UNKNOWN) would be more applicable #464

[atc0005]

Tangent:

Might be worth doing the same for DNS related errors.

Probably also worth reviewing the current handling and introducing retry tuning.