plugin/forward: configurable domain support for healthcheck (#5281)

* plugin/forward: configurable domain support for healthcheck

Signed-off-by: hansedong <admin@yinxiaoluo.com>

7 files changed, 113 insertions, 35 deletions

diff --git a/plugin/forward/README.md b/plugin/forward/README.md
index 892701f70..1ea335bf2 100644
--- a/plugin/forward/README.md
+++ b/plugin/forward/README.md
@@ -50,7 +50,7 @@ forward FROM TO... {
     tls CERT KEY CA
     tls_servername NAME
     policy random|round_robin|sequential
-    health_check DURATION [no_rec]
+    health_check DURATION [no_rec] [domain DOMAIN]
     max_concurrent MAX
 }
 ~~~
@@ -90,6 +90,8 @@ forward FROM TO... {
   * `<duration>` - use a different duration for health checking, the default duration is 0.5s.
   * `no_rec` - optional argument that sets the RecursionDesired-flag of the dns-query used in health checking to `false`.
     The flag is default `true`.
+  * `domain DOMAIN` - optional arguments that sets the domain of the dns-query used in health checking.
+    If not configured, the requested domain name is `.`. `DOMAIN` is used to configure the domain name.
 * `max_concurrent` **MAX** will limit the number of concurrent queries to **MAX**.  Any new query that would
   raise the number of concurrent queries above the **MAX** will result in a REFUSED response. This
   response does not count as a health failure. When choosing a value for **MAX**, pick a number
@@ -182,6 +184,18 @@ service with health checks.
 }
 ~~~
 
+Or configure other domain name for health check requests
+
+~~~ corefile
+. {
+    forward . tls://9.9.9.9 {
+       tls_servername dns.quad9.net
+       health_check 5s domain example.org
+    }
+    cache 30
+}
+~~~
+
 Or with multiple upstreams from the same provider
 
 ~~~ corefile
diff --git a/plugin/forward/forward.go b/plugin/forward/forward.go
index 19a469c72..e7fcaaad6 100644
--- a/plugin/forward/forward.go
+++ b/plugin/forward/forward.go
@@ -56,7 +56,7 @@ type Forward struct {
 
 // New returns a new Forward.
 func New() *Forward {
-	f := &Forward{maxfails: 2, tlsConfig: new(tls.Config), expire: defaultExpire, p: new(random), from: ".", hcInterval: hcInterval, opts: options{forceTCP: false, preferUDP: false, hcRecursionDesired: true}}
+	f := &Forward{maxfails: 2, tlsConfig: new(tls.Config), expire: defaultExpire, p: new(random), from: ".", hcInterval: hcInterval, opts: options{forceTCP: false, preferUDP: false, hcRecursionDesired: true, hcDomain: "."}}
 	return f
 }
 
@@ -234,6 +234,7 @@ type options struct {
 	forceTCP           bool
 	preferUDP          bool
 	hcRecursionDesired bool
+	hcDomain           string
 }
 
 var defaultTimeout = 5 * time.Second
diff --git a/plugin/forward/health.go b/plugin/forward/health.go
index f4cfab834..ec0b48143 100644
--- a/plugin/forward/health.go
+++ b/plugin/forward/health.go
@@ -16,6 +16,8 @@ type HealthChecker interface {
 	SetTLSConfig(*tls.Config)
 	SetRecursionDesired(bool)
 	GetRecursionDesired() bool
+	SetDomain(domain string)
+	GetDomain() string
 	SetTCPTransport()
 }
 
@@ -23,6 +25,7 @@ type HealthChecker interface {
 type dnsHc struct {
 	c                *dns.Client
 	recursionDesired bool
+	domain           string
 }
 
 var (
@@ -31,7 +34,7 @@ var (
 )
 
 // NewHealthChecker returns a new HealthChecker based on transport.
-func NewHealthChecker(trans string, recursionDesired bool) HealthChecker {
+func NewHealthChecker(trans string, recursionDesired bool, domain string) HealthChecker {
 	switch trans {
 	case transport.DNS, transport.TLS:
 		c := new(dns.Client)
@@ -39,7 +42,7 @@ func NewHealthChecker(trans string, recursionDesired bool) HealthChecker {
 		c.ReadTimeout = hcReadTimeout
 		c.WriteTimeout = hcWriteTimeout
 
-		return &dnsHc{c: c, recursionDesired: recursionDesired}
+		return &dnsHc{c: c, recursionDesired: recursionDesired, domain: domain}
 	}
 
 	log.Warningf("No healthchecker for transport %q", trans)
@@ -58,6 +61,13 @@ func (h *dnsHc) GetRecursionDesired() bool {
 	return h.recursionDesired
 }
 
+func (h *dnsHc) SetDomain(domain string) {
+	h.domain = domain
+}
+func (h *dnsHc) GetDomain() string {
+	return h.domain
+}
+
 func (h *dnsHc) SetTCPTransport() {
 	h.c.Net = "tcp"
 }
@@ -80,7 +90,7 @@ func (h *dnsHc) Check(p *Proxy) error {
 
 func (h *dnsHc) send(addr string) error {
 	ping := new(dns.Msg)
-	ping.SetQuestion(".", dns.TypeNS)
+	ping.SetQuestion(h.domain, dns.TypeNS)
 	ping.MsgHdr.RecursionDesired = h.recursionDesired
 
 	m, _, err := h.c.Exchange(ping, addr)
diff --git a/plugin/forward/health_test.go b/plugin/forward/health_test.go
index 2d65f4353..9917b3a37 100644
--- a/plugin/forward/health_test.go
+++ b/plugin/forward/health_test.go
@@ -242,3 +242,42 @@ func TestHealthNoMaxFails(t *testing.T) {
 		t.Errorf("Expected number of health checks to be %d, got %d", 0, i1)
 	}
 }
+
+func TestHealthDomain(t *testing.T) {
+	hcReadTimeout = 10 * time.Millisecond
+	readTimeout = 10 * time.Millisecond
+	defaultTimeout = 10 * time.Millisecond
+	hcWriteTimeout = 10 * time.Millisecond
+	hcDomain := "example.org."
+	i := uint32(0)
+	q := uint32(0)
+	s := dnstest.NewServer(func(w dns.ResponseWriter, r *dns.Msg) {
+		if atomic.LoadUint32(&q) == 0 { //drop the first query to trigger health-checking
+			atomic.AddUint32(&q, 1)
+			return
+		}
+		if r.Question[0].Name == hcDomain && r.RecursionDesired == true {
+			atomic.AddUint32(&i, 1)
+		}
+		ret := new(dns.Msg)
+		ret.SetReply(r)
+		w.WriteMsg(ret)
+	})
+	defer s.Close()
+	p := NewProxy(s.Addr, transport.DNS)
+	p.health.SetDomain(hcDomain)
+	f := New()
+	f.SetProxy(p)
+	defer f.OnShutdown()
+
+	req := new(dns.Msg)
+	req.SetQuestion(".", dns.TypeNS)
+
+	f.ServeDNS(context.TODO(), &test.ResponseWriter{}, req)
+
+	time.Sleep(20 * time.Millisecond)
+	i1 := atomic.LoadUint32(&i)
+	if i1 != 1 {
+		t.Errorf("Expected number of health checks with Domain==%s to be %d, got %d", hcDomain, 1, i1)
+	}
+}
diff --git a/plugin/forward/proxy.go b/plugin/forward/proxy.go
index 80cb581f9..6a4b5693e 100644
--- a/plugin/forward/proxy.go
+++ b/plugin/forward/proxy.go
@@ -29,7 +29,7 @@ func NewProxy(addr, trans string) *Proxy {
 		probe:     up.New(),
 		transport: newTransport(addr),
 	}
-	p.health = NewHealthChecker(trans, true)
+	p.health = NewHealthChecker(trans, true, ".")
 	runtime.SetFinalizer(p, (*Proxy).finalizer)
 	return p
 }
diff --git a/plugin/forward/setup.go b/plugin/forward/setup.go
index baf0512ba..af8141898 100644
--- a/plugin/forward/setup.go
+++ b/plugin/forward/setup.go
@@ -151,6 +151,7 @@ func parseStanza(c *caddy.Controller) (*Forward, error) {
 		if f.opts.forceTCP && transports[i] != transport.TLS {
 			f.proxies[i].health.SetTCPTransport()
 		}
+		f.proxies[i].health.SetDomain(f.opts.hcDomain)
 	}
 
 	return f, nil
@@ -187,11 +188,17 @@ func parseBlock(c *caddy.Controller, f *Forward) error {
 			return fmt.Errorf("health_check can't be negative: %d", dur)
 		}
 		f.hcInterval = dur
+		f.opts.hcDomain = "."
 
 		for c.NextArg() {
 			switch hcOpts := c.Val(); hcOpts {
 			case "no_rec":
 				f.opts.hcRecursionDesired = false
+			case "domain":
+				if !c.NextArg() {
+					return c.ArgErr()
+				}
+				f.opts.hcDomain = c.Val()
 			default:
 				return fmt.Errorf("health_check: unknown option %s", hcOpts)
 			}
diff --git a/plugin/forward/setup_test.go b/plugin/forward/setup_test.go
index 3c38acf97..2a5257e98 100644
--- a/plugin/forward/setup_test.go
+++ b/plugin/forward/setup_test.go
@@ -20,25 +20,27 @@ func TestSetup(t *testing.T) {
 		expectedErr     string
 	}{
 		// positive
-		{"forward . 127.0.0.1", false, ".", nil, 2, options{hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1 {\nexcept miek.nl\n}\n", false, ".", nil, 2, options{hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1 {\nmax_fails 3\n}\n", false, ".", nil, 3, options{hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1 {\nforce_tcp\n}\n", false, ".", nil, 2, options{forceTCP: true, hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1 {\nprefer_udp\n}\n", false, ".", nil, 2, options{preferUDP: true, hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1 {\nforce_tcp\nprefer_udp\n}\n", false, ".", nil, 2, options{preferUDP: true, forceTCP: true, hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1:53", false, ".", nil, 2, options{hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1:8080", false, ".", nil, 2, options{hcRecursionDesired: true}, ""},
-		{"forward . [::1]:53", false, ".", nil, 2, options{hcRecursionDesired: true}, ""},
-		{"forward . [2003::1]:53", false, ".", nil, 2, options{hcRecursionDesired: true}, ""},
-		{"forward . 127.0.0.1 \n", false, ".", nil, 2, options{hcRecursionDesired: true}, ""},
-		{"forward 10.9.3.0/18 127.0.0.1", false, "0.9.10.in-addr.arpa.", nil, 2, options{hcRecursionDesired: true}, ""},
+		{"forward . 127.0.0.1", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s domain example.org\n}\n", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "example.org"}, ""},
+		{"forward . 127.0.0.1 {\nexcept miek.nl\n}\n", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1 {\nmax_fails 3\n}\n", false, ".", nil, 3, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1 {\nforce_tcp\n}\n", false, ".", nil, 2, options{forceTCP: true, hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1 {\nprefer_udp\n}\n", false, ".", nil, 2, options{preferUDP: true, hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1 {\nforce_tcp\nprefer_udp\n}\n", false, ".", nil, 2, options{preferUDP: true, forceTCP: true, hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1:53", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1:8080", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . [::1]:53", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . [2003::1]:53", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward . 127.0.0.1 \n", false, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
+		{"forward 10.9.3.0/18 127.0.0.1", false, "0.9.10.in-addr.arpa.", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, ""},
 		// negative
-		{"forward . a27.0.0.1", true, "", nil, 0, options{hcRecursionDesired: true}, "not an IP"},
-		{"forward . 127.0.0.1 {\nblaatl\n}\n", true, "", nil, 0, options{hcRecursionDesired: true}, "unknown property"},
+		{"forward . a27.0.0.1", true, "", nil, 0, options{hcRecursionDesired: true, hcDomain: "."}, "not an IP"},
+		{"forward . 127.0.0.1 {\nblaatl\n}\n", true, "", nil, 0, options{hcRecursionDesired: true, hcDomain: "."}, "unknown property"},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s domain\n}\n", true, "", nil, 0, options{hcRecursionDesired: true, hcDomain: "."}, "Wrong argument count or unexpected line ending after 'domain'"},
 		{`forward . ::1
-		forward com ::2`, true, "", nil, 0, options{hcRecursionDesired: true}, "plugin"},
-		{"forward . https://127.0.0.1 \n", true, ".", nil, 2, options{hcRecursionDesired: true}, "'https' is not supported as a destination protocol in forward: https://127.0.0.1"},
-		{"forward xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 127.0.0.1 \n", true, ".", nil, 2, options{hcRecursionDesired: true}, "unable to normalize 'xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx'"},
+		forward com ::2`, true, "", nil, 0, options{hcRecursionDesired: true, hcDomain: "."}, "plugin"},
+		{"forward . https://127.0.0.1 \n", true, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, "'https' is not supported as a destination protocol in forward: https://127.0.0.1"},
+		{"forward xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 127.0.0.1 \n", true, ".", nil, 2, options{hcRecursionDesired: true, hcDomain: "."}, "unable to normalize 'xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx'"},
 	}
 
 	for i, test := range tests {
@@ -221,18 +223,23 @@ func TestSetupMaxConcurrent(t *testing.T) {
 
 func TestSetupHealthCheck(t *testing.T) {
 	tests := []struct {
-		input       string
-		shouldErr   bool
-		expectedVal bool
-		expectedErr string
+		input          string
+		shouldErr      bool
+		expectedRecVal bool
+		expectedDomain string
+		expectedErr    string
 	}{
 		// positive
-		{"forward . 127.0.0.1\n", false, true, ""},
-		{"forward . 127.0.0.1 {\nhealth_check 0.5s\n}\n", false, true, ""},
-		{"forward . 127.0.0.1 {\nhealth_check 0.5s no_rec\n}\n", false, false, ""},
+		{"forward . 127.0.0.1\n", false, true, ".", ""},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s\n}\n", false, true, ".", ""},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s no_rec\n}\n", false, false, ".", ""},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s no_rec domain example.org\n}\n", false, false, "example.org", ""},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s domain example.org\n}\n", false, true, "example.org", ""},
 		// negative
-		{"forward . 127.0.0.1 {\nhealth_check no_rec\n}\n", true, true, "time: invalid duration"},
-		{"forward . 127.0.0.1 {\nhealth_check 0.5s rec\n}\n", true, true, "health_check: unknown option rec"},
+		{"forward . 127.0.0.1 {\nhealth_check no_rec\n}\n", true, true, ".", "time: invalid duration"},
+		{"forward . 127.0.0.1 {\nhealth_check domain example.org\n}\n", true, true, "example.org", "time: invalid duration"},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s rec\n}\n", true, true, ".", "health_check: unknown option rec"},
+		{"forward . 127.0.0.1 {\nhealth_check 0.5s domain\n}\n", true, true, ".", "Wrong argument count or unexpected line ending after 'domain'"},
 	}
 
 	for i, test := range tests {
@@ -247,13 +254,13 @@ func TestSetupHealthCheck(t *testing.T) {
 			if !test.shouldErr {
 				t.Errorf("Test %d: expected no error but found one for input %s, got: %v", i, test.input, err)
 			}
-
 			if !strings.Contains(err.Error(), test.expectedErr) {
 				t.Errorf("Test %d: expected error to contain: %v, found error: %v, input: %s", i, test.expectedErr, err, test.input)
 			}
 		}
-		if !test.shouldErr && (f.opts.hcRecursionDesired != test.expectedVal || f.proxies[0].health.GetRecursionDesired() != test.expectedVal) {
-			t.Errorf("Test %d: expected: %t, got: %d", i, test.expectedVal, f.maxConcurrent)
+		if !test.shouldErr && (f.opts.hcRecursionDesired != test.expectedRecVal || f.proxies[0].health.GetRecursionDesired() != test.expectedRecVal ||
+			f.opts.hcDomain != test.expectedDomain || f.proxies[0].health.GetDomain() != test.expectedDomain) {
+			t.Errorf("Test %d: expectedRec: %v, got: %v. expectedDomain: %s, got: %s. ", i, test.expectedRecVal, f.opts.hcRecursionDesired, test.expectedDomain, f.opts.hcDomain)
 		}
 	}
 }