DNS services are not resolving properly

The change on Sept 15th changed how services resolved in the absence
of search paths, which resulted in very long times to resolve DNS in
some cases.

Author: smarterclayton

pkg/dns/server.go

@@ -3,6 +3,8 @@ package dns
 import (
 	client "k8s.io/kubernetes/pkg/client/unversioned"
 
+	"github.com/golang/glog"
+
 	"github.com/coreos/go-etcd/etcd"
 	"github.com/prometheus/client_golang/prometheus"
 	backendetcd "github.com/skynetservices/skydns/backends/etcd"
@@ -12,8 +14,9 @@ import (
 // NewServerDefaults returns the default SkyDNS server configuration for a DNS server.
 func NewServerDefaults() (*server.Config, error) {
 	config := &server.Config{
-		Domain: "cluster.local.",
-		Local:  "openshift.default.svc.cluster.local.",
+		Domain:  "cluster.local.",
+		Local:   "openshift.default.svc.cluster.local.",
+		Verbose: bool(glog.V(4)),
 	}
 	return config, server.SetDefaults(config)
 }

pkg/dns/serviceresolver.go

@@ -4,6 +4,8 @@ import (
 	"fmt"
 	"strings"
 
+	"github.com/golang/glog"
+
 	kapi "k8s.io/kubernetes/pkg/api"
 	"k8s.io/kubernetes/pkg/api/errors"
 	kclient "k8s.io/kubernetes/pkg/client/unversioned"
@@ -60,20 +62,20 @@ func NewServiceResolver(config *server.Config, accessor ServiceAccessor, endpoin
 //   * endpoint_id is "portal" when portalIP is set
 // * endpoints always returns each individual endpoint as A records
 //
-func (b *ServiceResolver) Records(name string, exact bool) ([]msg.Service, error) {
-	if !strings.HasSuffix(name, b.base) {
+func (b *ServiceResolver) Records(dnsName string, exact bool) ([]msg.Service, error) {
+	if !strings.HasSuffix(dnsName, b.base) {
 		return nil, nil
 	}
-	prefix := strings.Trim(strings.TrimSuffix(name, b.base), ".")
+	prefix := strings.Trim(strings.TrimSuffix(dnsName, b.base), ".")
 	segments := strings.Split(prefix, ".")
 	for i, j := 0, len(segments)-1; i < j; i, j = i+1, j-1 {
 		segments[i], segments[j] = segments[j], segments[i]
 	}
 	if len(segments) == 0 {
 		return nil, nil
 	}
-
-	switch segments[0] {
+	glog.V(4).Infof("Answering query %s:%t", dnsName, exact)
+	switch base := segments[0]; base {
 	case "svc", "endpoints":
 		if len(segments) < 3 {
 			return nil, nil
@@ -94,7 +96,8 @@ func (b *ServiceResolver) Records(name string, exact bool) ([]msg.Service, error
 			return nil, nil
 		}
 
-		retrieveEndpoints := segments[0] == "endpoints" || (len(segments) > 3 && segments[3] == "_endpoints")
+		subdomain := buildDNSName(b.base, base, namespace, name)
+		retrieveEndpoints := base == "endpoints" || (len(segments) > 3 && segments[3] == "_endpoints")
 
 		// if has a portal IP and looking at svc
 		if svc.Spec.ClusterIP != kapi.ClusterIPNone && !retrieveEndpoints {
@@ -117,8 +120,8 @@ func (b *ServiceResolver) Records(name string, exact bool) ([]msg.Service, error
 				if len(portName) == 0 {
 					portName = fmt.Sprintf("unknown-port-%d", port)
 				}
-				srvName := fmt.Sprintf("%s.portal.%s", portName, name)
-				keyName := fmt.Sprintf("_%s._%s.%s", portName, p.Protocol, name)
+				srvName := buildDNSName(subdomain, "portal", portName)
+				keyName := buildDNSName(subdomain, "_"+strings.ToLower(string(p.Protocol)), "_"+portName)
 				services = append(services,
 					msg.Service{
 						Host: svc.Spec.ClusterIP,
@@ -144,6 +147,7 @@ func (b *ServiceResolver) Records(name string, exact bool) ([]msg.Service, error
 					},
 				)
 			}
+			glog.V(4).Infof("Answered %s:%t with %#v", dnsName, exact, services)
 			return services, nil
 		}
 
@@ -181,11 +185,11 @@ func (b *ServiceResolver) Records(name string, exact bool) ([]msg.Service, error
 					if len(p.Protocol) == 0 {
 						p.Protocol = kapi.ProtocolTCP
 					}
-					portName := p.Name
+					portName := strings.ToLower(p.Name)
 					if len(portName) == 0 {
 						portName = fmt.Sprintf("unknown-port-%d", port)
 					}
-					srvName := fmt.Sprintf("%s.%s.%s", portName, shortName, name)
+					srvName := buildDNSName(subdomain, shortName, portName)
 					services = append(services, msg.Service{
 						Host: a.IP,
 						Port: port,
@@ -197,7 +201,7 @@ func (b *ServiceResolver) Records(name string, exact bool) ([]msg.Service, error
 						Text: "",
 						Key:  msg.Path(srvName),
 					})
-					keyName := fmt.Sprintf("_%s._%s.%s", portName, p.Protocol, name)
+					keyName := buildDNSName(subdomain, "_"+strings.ToLower(string(p.Protocol)), "_"+portName)
 					services = append(services, msg.Service{
 						Host: srvName,
 						Port: port,
@@ -225,6 +229,7 @@ func (b *ServiceResolver) Records(name string, exact bool) ([]msg.Service, error
 				}
 			}
 		}
+		glog.V(4).Infof("Answered %s:%t with %#v", dnsName, exact, services)
 		return services, nil
 	}
 	return nil, nil
@@ -246,8 +251,9 @@ func (b *ServiceResolver) ReverseRecord(name string) (*msg.Service, error) {
 	if len(svc.Spec.Ports) > 0 {
 		port = svc.Spec.Ports[0].Port
 	}
+	hostName := buildDNSName(b.base, "svc", svc.Namespace, svc.Name)
 	return &msg.Service{
-		Host: fmt.Sprintf("%s.%s.svc.%s", svc.Name, svc.Namespace, b.base),
+		Host: hostName,
 		Port: port,
 
 		Priority: 10,
@@ -278,3 +284,16 @@ func extractIP(reverseName string) (string, bool) {
 	}
 	return strings.Join(segments, "."), true
 }
+
+// buildDNSName reverses the labels order and joins them with dots.
+func buildDNSName(labels ...string) string {
+	var res string
+	for _, label := range labels {
+		if res == "" {
+			res = label
+		} else {
+			res = fmt.Sprintf("%s.%s", label, res)
+		}
+	}
+	return res
+}