feat: switch agent to a periodic dump rather than long running process

internal/config/preflight.go

@@ -3,10 +3,13 @@ package config
 import (
 	"fmt"
 	"net/http"
+	"os"
+	"path/filepath"
 	"strings"
 )
 
-// Checks to make sure all required environment variables are set
+// Checks to make sure all required environment variables are set.
+// TSAuthKey is only required when there is no existing tsnet state.
 func validateRequired(config *Config) error {
 	if config.Hostname == "" {
 		return fmt.Errorf("%s is required", HostnameEnv)
@@ -16,17 +19,23 @@ func validateRequired(config *Config) error {
 		return fmt.Errorf("%s is required", TSControlURLEnv)
 	}
 
-	if config.TSAuthKey == "" {
-		return fmt.Errorf("%s is required", TSAuthKeyEnv)
-	}
-
 	if config.WorkDir == "" {
 		return fmt.Errorf("%s is required", WorkDirEnv)
 	}
 
+	if config.TSAuthKey == "" && !hasExistingState(config.WorkDir) {
+		return fmt.Errorf("%s is required for first run (no existing state in %s)", TSAuthKeyEnv, config.WorkDir)
+	}
+
 	return nil
 }
 
+// hasExistingState checks if tsnet has previously stored identity state.
+func hasExistingState(workDir string) bool {
+	_, err := os.Stat(filepath.Join(workDir, "tailscaled.state"))
+	return err == nil
+}
+
 // Pings the Tailscale control server to make sure it's up and running
 func validateTSReady(config *Config) error {
 	testURL := config.TSControlURL

internal/tsnet/peers.go

@@ -69,17 +69,17 @@ func (s *TSAgent) GetStatusForPeer(id string) (*tailcfg.HostinfoView, error) {
 	return &whois.Node.Hostinfo, nil
 }
 
-// Dispatches ALL the HostInfo entries in our Tailnet to the master
-func (s *TSAgent) DispatchHostInfo(ctx context.Context) error {
+// FetchAllHostInfo fetches hostinfo for all peers and returns them as a map
+// keyed by node public key (e.g., "nodekey:abc123...").
+func (s *TSAgent) FetchAllHostInfo(ctx context.Context) (map[string]json.RawMessage, error) {
 	log := util.GetLogger()
 
 	stat, err := s.Lc.Status(ctx)
 	if err != nil {
 		log.Debug("Failed to get status: %s", err)
-		return fmt.Errorf("failed to get status: %w", err)
+		return nil, fmt.Errorf("failed to get status: %w", err)
 	}
 
-	// Do lookups for all peers with a hint of parallelism for speed!
 	const maxParallel = 8
 	sema := make(chan struct{}, maxParallel)
 	var wg sync.WaitGroup
@@ -96,6 +96,8 @@ func (s *TSAgent) DispatchHostInfo(ctx context.Context) error {
 		nodeMap[nodeKey] = peer
 	}
 
+	result := make(map[string]json.RawMessage)
+
 	for nodeKey, peer := range nodeMap {
 		idBytes, err := nodeKey.MarshalText()
 		if err != nil {
@@ -138,11 +140,11 @@ func (s *TSAgent) DispatchHostInfo(ctx context.Context) error {
 			}
 
 			mu.Lock()
-			fmt.Println("HOSTINFO " + nodeID + " " + string(data))
+			result[nodeID] = json.RawMessage(data)
 			mu.Unlock()
 		}()
 	}
 
 	wg.Wait()
-	return nil
+	return result, nil
 }

internal/util/logger.go

@@ -59,7 +59,7 @@ func (l *Logger) SetDebug(enabled bool) {
 func (l *Logger) log(level LogLevel, format string, v ...any) {
 	msg := fmt.Sprintf(format, v...)
 
-	fmt.Printf("LOG %s %s\n", level, msg)
+	fmt.Fprintf(os.Stderr, "LOG %s %s\n", level, msg)
 	if level == LevelFatal {
 		os.Exit(1)
 	}