feat(agent): add updater-agent binary and ship it in the image

2026-10-12 14:41:05 +02:00
parent ad6ff1e066
commit 9b74286c9b
4 changed files with 199 additions and 7 deletions
@@ -8,15 +8,17 @@ COPY . .
 ARG VERSION=dev
 ARG COMMIT=unknown
 ARG BUILD_TIME=unknown
-RUN CGO_ENABLED=0 go build \
-    -ldflags="-s -w \
-      -X main.version=${VERSION} \
-      -X main.commit=${COMMIT} \
-      -X main.buildTime=${BUILD_TIME}" \
-    -o /out/package-updater ./cmd/server
+RUN for cmd in server:package-updater agent:updater-agent; do \
+      CGO_ENABLED=0 go build \
+        -ldflags="-s -w \
+          -X main.version=${VERSION} \
+          -X main.commit=${COMMIT} \
+          -X main.buildTime=${BUILD_TIME}" \
+        -o /out/${cmd#*:} ./cmd/${cmd%%:*} || exit 1; \
+    done

 FROM alpine:3.20
 RUN apk add --no-cache docker-cli docker-cli-compose ca-certificates wget
-COPY --from=build /out/package-updater /usr/local/bin/
+COPY --from=build /out/ /usr/local/bin/
 EXPOSE 8080
 ENTRYPOINT ["/usr/local/bin/package-updater"]
@@ -0,0 +1,89 @@
+// Command updater-agent reports this machine's Docker services to the
+// package-updater master.
+package main
+
+import (
+	"context"
+	"fmt"
+	"log/slog"
+	"os"
+	"os/signal"
+	"syscall"
+	"time"
+
+	"github.com/docker/docker/client"
+
+	"github.com/shcizo/package-updater/internal/agent/config"
+	"github.com/shcizo/package-updater/internal/agent/conn"
+	"github.com/shcizo/package-updater/internal/agent/inventory"
+	"github.com/shcizo/package-updater/internal/logging"
+)
+
+// Build info is injected at link time via -ldflags="-X main.version=...".
+var version = "dev"
+
+func main() {
+	if err := run(); err != nil {
+		fmt.Fprintln(os.Stderr, "fatal:", err)
+		os.Exit(1)
+	}
+}
+
+func run() error {
+	cfg, err := config.Load()
+	if err != nil {
+		return err
+	}
+	logger := logging.New(cfg.LogLevel)
+
+	dockerCli, err := client.NewClientWithOpts(client.FromEnv, client.WithAPIVersionNegotiation())
+	if err != nil {
+		return fmt.Errorf("docker client: %w", err)
+	}
+	defer dockerCli.Close()
+
+	collector := inventory.New(dockerCli, cfg.SelfID)
+	agent := conn.New(conn.Config{
+		URL: cfg.MasterURL, Token: cfg.Token, Fingerprint: cfg.MasterKey, Version: version,
+	}, collector, logger)
+
+	ctx, stop := signal.NotifyContext(context.Background(), syscall.SIGINT, syscall.SIGTERM)
+	defer stop()
+
+	go watchLoop(ctx, dockerCli, agent, logger)
+	logger.Info("agent_starting", "version", version, "master_url", cfg.MasterURL)
+	return agent.Run(ctx)
+}
+
+// watchLoop keeps a Docker event subscription alive. When the stream breaks
+// (daemon restart), it resubscribes and asks for a refresh, because events
+// may have been missed in between. It also asks for a full re-inventory every
+// 60 seconds, since a Swarm replica dying on another node emits no event here.
+func watchLoop(ctx context.Context, src inventory.EventSource, agent *conn.Agent, log *slog.Logger) {
+	go func() {
+		ticker := time.NewTicker(60 * time.Second)
+		defer ticker.Stop()
+		for {
+			select {
+			case <-ctx.Done():
+				return
+			case <-ticker.C:
+				agent.Notify()
+			}
+		}
+	}()
+
+	for {
+		err := inventory.Watch(ctx, src, 2*time.Second, agent.Notify)
+		if ctx.Err() != nil {
+			return
+		}
+		log.Warn("docker_events_lost", "error", err)
+		select {
+		case <-ctx.Done():
+			return
+		case <-time.After(5 * time.Second):
+		}
+		agent.Notify()
+	}
+}
@@ -0,0 +1,51 @@
+// Package config loads the agent's configuration from environment variables.
+package config
+
+import (
+	"errors"
+	"os"
+	"regexp"
+	"strings"
+)
+
+var masterKeyPattern = regexp.MustCompile(`^sha256:[0-9a-f]{64}$`)
+
+// Config holds the agent's runtime configuration.
+type Config struct {
+	MasterURL string // MASTER_URL, wss://host:port/agent
+	MasterKey string // MASTER_KEY, fingerprint shown in the master UI
+	Token     string // AGENT_TOKEN, shown once when the agent was added
+	SelfID    string // SELF_ID, defaults to the hostname (= container ID prefix)
+	LogLevel  string // LOG_LEVEL, default info
+}
+
+// Load reads and validates the agent configuration. It fails fast so a
+// misconfigured agent never starts dialling.
+func Load() (*Config, error) {
+	cfg := &Config{
+		MasterURL: os.Getenv("MASTER_URL"),
+		MasterKey: os.Getenv("MASTER_KEY"),
+		Token:     os.Getenv("AGENT_TOKEN"),
+		SelfID:    os.Getenv("SELF_ID"),
+		LogLevel:  os.Getenv("LOG_LEVEL"),
+	}
+	switch {
+	case cfg.MasterURL == "":
+		return nil, errors.New("MASTER_URL is required")
+	case !strings.HasPrefix(cfg.MasterURL, "wss://"):
+		return nil, errors.New("MASTER_URL must start with wss://")
+	case cfg.MasterKey == "":
+		return nil, errors.New("MASTER_KEY is required")
+	case !masterKeyPattern.MatchString(cfg.MasterKey):
+		return nil, errors.New("MASTER_KEY must look like sha256:<64 hex>")
+	case cfg.Token == "":
+		return nil, errors.New("AGENT_TOKEN is required")
+	}
+	if cfg.SelfID == "" {
+		cfg.SelfID, _ = os.Hostname()
+	}
+	if cfg.LogLevel == "" {
+		cfg.LogLevel = "info"
+	}
+	return cfg, nil
+}
@@ -0,0 +1,50 @@
+package config
+
+import (
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+const validKey = "sha256:0000000000000000000000000000000000000000000000000000000000000000"
+
+func setValid(t *testing.T) {
+	t.Setenv("MASTER_URL", "wss://10.0.0.5:8090/agent")
+	t.Setenv("MASTER_KEY", validKey)
+	t.Setenv("AGENT_TOKEN", "pua_abc")
+	t.Setenv("SELF_ID", "")
+	t.Setenv("LOG_LEVEL", "")
+}
+
+func TestLoad_Valid(t *testing.T) {
+	setValid(t)
+	cfg, err := Load()
+	require.NoError(t, err)
+	host, _ := os.Hostname()
+	assert.Equal(t, &Config{
+		MasterURL: "wss://10.0.0.5:8090/agent", MasterKey: validKey, Token: "pua_abc",
+		SelfID: host, LogLevel: "info",
+	}, cfg)
+}
+
+func TestLoad_Invalid(t *testing.T) {
+	cases := []struct {
+		name, key, value, wantErr string
+	}{
+		{"missing url", "MASTER_URL", "", "MASTER_URL is required"},
+		{"plain ws", "MASTER_URL", "ws://10.0.0.5:8090/agent", "MASTER_URL must start with wss://"},
+		{"missing key", "MASTER_KEY", "", "MASTER_KEY is required"},
+		{"malformed key", "MASTER_KEY", "abc", "MASTER_KEY must look like sha256:<64 hex>"},
+		{"missing token", "AGENT_TOKEN", "", "AGENT_TOKEN is required"},
+	}
+	for _, c := range cases {
+		t.Run(c.name, func(t *testing.T) {
+			setValid(t)
+			t.Setenv(c.key, c.value)
+			_, err := Load()
+			assert.ErrorContains(t, err, c.wantErr)
+		})
+	}
+}