HTTP/HTTPS check accepts non-2xx responses, leaks response body, and has no timeout

[dolph]

Summary

The HTTP check in https.go has three independent defects that, combined, make it unreliable as a connectivity validator:

1. Status code is never checked. The README states "An HTTP 2xx response is expected", but the code returns success on any non-error response — including 4xx and 5xx.
2. Response body is never closed. This leaks file descriptors and connections; in monitor mode it accumulates over time.
3. No timeout. http.Get uses the default client, which has no timeout. A slow or hung server will block the check indefinitely. In monitor mode that destination will simply stop being checked.

There is also a subtler design issue: the per-IP Dial loop validates connectivity to each address returned by DNS, but http.Get(dest.URL) re-resolves DNS internally via the OS resolver, so the actual HTTP request may end up at a different IP than what Lookup/Dial verified.

Code

https.go:

func HTTPS(dest *Destination) bool {
	dest.Increment("connectivity.http", []string{})
	_, err := http.Get(dest.URL)               // no timeout; response discarded; body never closed
	if err != nil {
		dest.Increment("connectivity.http.error", []string{})
		LogDestinationError(dest, "Failed HTTP GET", err)
		return false
	}
	dest.Increment("connectivity.http.success", []string{})
	return true
}

Impact

• A destination returning HTTP 500 is reported as "Connected", so the tool returns 0 from connectivity check and gates do not catch the failure.
• Long-running monitor accumulates leaked connections and may hang on slow upstreams forever.

Suggested fix

func HTTPS(dest *Destination) bool {
	dest.Increment("connectivity.http", []string{})
	client := &http.Client{Timeout: 15 * time.Second}
	resp, err := client.Get(dest.URL)
	if err != nil {
		dest.Increment("connectivity.http.error", []string{})
		LogDestinationError(dest, "Failed HTTP GET", err)
		return false
	}
	defer resp.Body.Close()
	io.Copy(io.Discard, resp.Body)              // drain so the connection can be reused
	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
		dest.Increment("connectivity.http.error", []string{})
		LogDestinationError(dest, fmt.Sprintf("Non-2xx status: %d", resp.StatusCode), nil)
		return false
	}
	dest.Increment("connectivity.http.success", []string{})
	return true
}

Consider also making the timeout configurable and emitting an http.timeout metric tag to distinguish hangs from connect failures.