Fix SIGHUP

CHANGELOG.md

@@ -7,6 +7,9 @@ Changelog for NeoFS Node
 
 ### Fixed
 - SN could panic on forwarding GET/HEAD/RANGE request (#4120)
+- gRPC SIGHUP reload no longer stops a running public API before replacement configuration is verified (#4113)
+- Storage node shuts down when SIGHUP configuration reload fails (#4113)
+- Invalid node attributes on SIGHUP no longer terminate a storage node (#4113)
 
 ### Changed
 - SNs exchange TLS certificates on inter-node connections (#4097)

cmd/neofs-node/attributes.go

@@ -8,24 +8,26 @@ import (
 	"go.uber.org/zap"
 )
 
-func parseAttributes(c *cfg) {
+func parseAttributes(c *cfg) error {
 	if c.appCfg.Node.Relay {
-		return
+		return nil
 	}
 
-	fatalOnErr(attributes.ReadNodeAttributes(&c.cfgNodeInfo.localInfo, c.appCfg.Node.Attributes))
+	if err := attributes.ReadNodeAttributes(&c.cfgNodeInfo.localInfo, c.appCfg.Node.Attributes); err != nil {
+		return err
+	}
 
 	// expand UN/LOCODE attribute if any found; keep user's attributes
 	// if any conflicts appear
 
 	locAttr := c.cfgNodeInfo.localInfo.LOCODE()
 	if locAttr == "" {
-		return
+		return nil
 	}
 
 	record, err := getRecord(locAttr)
 	if err != nil {
-		fatalOnErr(fmt.Errorf("could not get locode record from DB: %w", err))
+		return fmt.Errorf("could not get locode record from DB: %w", err)
 	}
 
 	countryCode := locAttr[:locodedb.CountryCodeLen]
@@ -84,6 +86,8 @@ func parseAttributes(c *cfg) {
 	} else {
 		setIfNotEmpty(n.SetSubdivisionName, record.SubDivName)
 	}
+
+	return nil
 }
 
 func getRecord(lc string) (locodedb.Record, error) {

cmd/neofs-node/config.go

@@ -237,6 +237,7 @@ type cfgGRPC struct {
 
 	listeners []net.Listener
 	servers   []*grpc.Server
+	config    grpcConfigSnapshot
 
 	// serviceRegistrators stores functions that register gRPC service
 	// implementations into a gRPC server.
@@ -251,7 +252,9 @@ func (g *cfgGRPC) registerService(f func(*grpc.Server)) {
 
 	g.serviceRegistrators = append(g.serviceRegistrators, f)
 	for _, srv := range g.servers {
-		f(srv)
+		if srv != nil {
+			f(srv)
+		}
 	}
 }
 
@@ -627,7 +630,6 @@ func (c *cfg) needBootstrap() bool {
 }
 
 func (c *cfg) configWatcher(ctx context.Context) {
-	var err error
 	ch := make(chan os.Signal, 1)
 	signal.Notify(ch, syscall.SIGHUP)
 
@@ -640,74 +642,79 @@ func (c *cfg) configWatcher(ctx context.Context) {
 				c.log.Warn("failed to notify systemd about reloading", zap.Error(err))
 			}
 
-			oldMetrics := writeMetricConfig(c.appCfg)
-			oldProfiler := writeProfilerConfig(c.appCfg)
-			oldGRPC := writeGRPCConfig(c.appCfg)
+			if err := c.reloadConfig(); err != nil {
+				c.internalErr <- fmt.Errorf("configuration reload: %w", err)
+				return
+			}
 
-			c.appCfg, err = config.New(config.WithConfigFile(c.appCfg.Path()))
-			if err != nil {
-				c.log.Error("configuration reading", zap.Error(err))
-				continue
+			c.log.Info("configuration has been reloaded successfully")
+
+			if err := sdnotify.Send(sdnotify.Ready); err != nil {
+				c.log.Warn("failed to notify systemd about readiness after reload", zap.Error(err))
 			}
+		case <-ctx.Done():
+			return
+		}
+	}
+}
 
-			// Prometheus and pprof
+//nolint:contextcheck // Reloading HTTP services does not receive a request context.
+func (c *cfg) reloadConfig() error {
+	oldCfg := c.appCfg
+	oldMetrics := writeMetricConfig(oldCfg)
+	oldProfiler := writeProfilerConfig(oldCfg)
 
-			// nolint:contextcheck
-			c.reloadMetricsAndPprof(oldMetrics, oldProfiler)
+	newCfg, err := config.New(config.WithConfigFile(oldCfg.Path()))
+	if err != nil {
+		return fmt.Errorf("read configuration: %w", err)
+	}
+	if err := validateConfig(newCfg); err != nil {
+		return fmt.Errorf("validate configuration: %w", err)
+	}
+	c.appCfg = newCfg
 
-			// Logger
+	// Prometheus and pprof
 
-			err = c.logLevel.UnmarshalText([]byte(c.appCfg.Logger.Level))
-			if err != nil {
-				c.log.Error("invalid logger level configuration", zap.Error(err))
-				continue
-			}
+	c.reloadMetricsAndPprof(oldMetrics, oldProfiler)
 
-			// Policer
+	// Logger
 
-			c.policer.Reload(c.policerOpts()...)
+	if err := c.logLevel.UnmarshalText([]byte(c.appCfg.Logger.Level)); err != nil {
+		return fmt.Errorf("set logger level: %w", err)
+	}
 
-			// Storage Engine
+	// Policer
 
-			var rcfg engine.ReConfiguration
-			for _, optsWithID := range c.shardOpts() {
-				rcfg.AddShard(optsWithID.configID, optsWithID.shOpts)
-			}
+	c.policer.Reload(c.policerOpts()...)
 
-			err = c.cfgObject.cfgLocalStorage.localStorage.Reload(rcfg)
-			if err != nil {
-				c.log.Error("storage engine configuration update", zap.Error(err))
-				continue
-			}
+	// Storage Engine
 
-			// Morph
+	var rcfg engine.ReConfiguration
+	for _, optsWithID := range c.shardOpts() {
+		rcfg.AddShard(optsWithID.configID, optsWithID.shOpts)
+	}
 
-			c.cli.Reload(client.WithEndpoints(c.appCfg.FSChain.Endpoints))
+	if err := c.cfgObject.cfgLocalStorage.localStorage.Reload(rcfg); err != nil {
+		return fmt.Errorf("update storage engine configuration: %w", err)
+	}
 
-			// Node
+	// Morph
 
-			err = c.reloadNodeAttributes()
-			if err != nil {
-				c.log.Error("invalid node attributes configuration", zap.Error(err))
-				continue
-			}
+	c.cli.Reload(client.WithEndpoints(c.appCfg.FSChain.Endpoints))
 
-			// gRPC
+	// Node
 
-			if err = reloadGRPC(c, oldGRPC); err != nil {
-				c.log.Error("gRPC configuration reload", zap.Error(err))
-				continue
-			}
+	if err := c.reloadNodeAttributes(); err != nil {
+		return fmt.Errorf("update node attributes: %w", err)
+	}
 
-			c.log.Info("configuration has been reloaded successfully")
+	// gRPC
 
-			if err := sdnotify.Send(sdnotify.Ready); err != nil {
-				c.log.Warn("failed to notify systemd about readiness after reload", zap.Error(err))
-			}
-		case <-ctx.Done():
-			return
-		}
+	if err := reloadGRPC(c); err != nil {
+		return fmt.Errorf("reload gRPC configuration: %w", err)
 	}
+
+	return nil
 }
 
 // writeSystemAttributes writes app version as defined at compilation