From cf7e333579561316a8f41cca9d378c1e33e5c2aa Mon Sep 17 00:00:00 2001
From: Pavel Sviderski <me@psviderski.name>
Date: Thu, 1 May 2025 11:27:27 +1000
Subject: [PATCH] feat(dns-server): use embedded DNS server for all service
 containers

---
 Makefile                          |  4 ++++
 internal/machine/docker/server.go | 22 +++++++++++++++++++---
 internal/machine/machine.go       | 12 +++++++-----
 3 files changed, 30 insertions(+), 8 deletions(-)

diff --git a/Makefile b/Makefile
index 4ba7a397..4254ae39 100644
--- a/Makefile
+++ b/Makefile
@@ -36,6 +36,10 @@ demo-reset:
 	ssh root@5.223.45.199 "AUTO_CONFIRM=true sudo -E uncloud-uninstall"
 	ssh spy@192.168.40.243 "AUTO_CONFIRM=true sudo -E uncloud-uninstall"
 
+.PHONY: ucind-cluster
+ucind-cluster:
+	go run ./cmd/ucind cluster rm && go run ./cmd/ucind cluster create -m 3
+
 .PHONY: proto
 proto:
 	protoc --go_out=. --go_opt=paths=source_relative --go-grpc_out=. --go-grpc_opt=paths=source_relative \
diff --git a/internal/machine/docker/server.go b/internal/machine/docker/server.go
index e011afe5..10d2fbba 100644
--- a/internal/machine/docker/server.go
+++ b/internal/machine/docker/server.go
@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"io"
 	"log/slog"
+	"net/netip"
 	"regexp"
 	"slices"
 	"strconv"
@@ -30,6 +31,7 @@ import (
 	"github.com/opencontainers/go-digest"
 	ocispec "github.com/opencontainers/image-spec/specs-go/v1"
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
+	"github.com/psviderski/uncloud/internal/machine/dns"
 	"github.com/psviderski/uncloud/internal/secret"
 	"github.com/psviderski/uncloud/pkg/api"
 	"google.golang.org/grpc"
@@ -45,13 +47,17 @@ type Server struct {
 	pb.UnimplementedDockerServer
 	client *client.Client
 	db     *sqlx.DB
+	// internalDNSIP is a function that returns the IP address of the internal DNS server. It may return an empty
+	// address if the address is unknown (e.g. when the machine is not initialised yet).
+	internalDNSIP func() netip.Addr
 }
 
 // NewServer creates a new Docker gRPC server with the provided Docker client.
-func NewServer(cli *client.Client, db *sqlx.DB) *Server {
+func NewServer(cli *client.Client, db *sqlx.DB, internalDNSIP func() netip.Addr) *Server {
 	return &Server{
-		client: cli,
-		db:     db,
+		client:        cli,
+		db:            db,
+		internalDNSIP: internalDNSIP,
 	}
 }
 
@@ -505,6 +511,16 @@ func (s *Server) CreateServiceContainer(
 		},
 	}
 
+	// Configure the container to use the internal DNS server if it's available.
+	dnsIP := s.internalDNSIP()
+	if dnsIP.IsValid() {
+		hostConfig.DNS = []string{dnsIP.String()}
+		// Optimize DNS resolution for service discovery by appending the search domain to names without a dot.
+		// For example, the first attempt for "my-service" will be "my-service.internal".
+		hostConfig.DNSOptions = []string{"ndots:1"}
+		hostConfig.DNSSearch = []string{dns.InternalDomain}
+	}
+
 	if spec.Container.LogDriver != nil {
 		hostConfig.LogConfig = container.LogConfig{
 			Type:   spec.Container.LogDriver.Name,
diff --git a/internal/machine/machine.go b/internal/machine/machine.go
index fb64ea78..e1293b4d 100644
--- a/internal/machine/machine.go
+++ b/internal/machine/machine.go
@@ -220,7 +220,6 @@ func NewMachine(config *Config) (*Machine, error) {
 	if err != nil {
 		return nil, fmt.Errorf("init machine database: %w", err)
 	}
-	dockerServer := machinedocker.NewServer(dockerCli, db)
 
 	// Init a local gRPC proxy server that proxies requests to the local or remote machine API servers.
 	proxyDirector := apiproxy.NewDirector(config.MachineSockPath, APIPort)
@@ -238,11 +237,16 @@ func NewMachine(config *Config) (*Machine, error) {
 		initialised:      make(chan struct{}, 1),
 		store:            corroStore,
 		cluster:          c,
-		docker:           dockerServer,
 		localProxyServer: localProxyServer,
 		proxyDirector:    proxyDirector,
 	}
-	m.localMachineServer = newGRPCServer(m, c, dockerServer)
+
+	// Machine IP will only be available after the machine is initialised as a cluster member so wrap it in a function.
+	internalDNSIP := func() netip.Addr {
+		return m.IP()
+	}
+	m.docker = machinedocker.NewServer(dockerCli, db, internalDNSIP)
+	m.localMachineServer = newGRPCServer(m, c, m.docker)
 
 	if m.Initialised() {
 		m.initialised <- struct{}{}
@@ -359,8 +363,6 @@ func (m *Machine) Run(ctx context.Context) error {
 					var err error
 
 					m.cluster.UpdateMachineID(m.state.ID)
-					// TODO: set DNS server to m.IP() on the docker server so it knows which addr to pass as --dns
-					//  for service containers.
 
 					// Ensure the corrosion config is up to date, including a new gossip address if the machine
 					// has just joined a cluster.