RELEASE 1.0.1: v2 log format, source_tag-labeled metrics, lint cleanup

Wire-format and metric overhaul. Both file and UDP ingest now share one versioned ParseLine that dispatches on the v<N>\t prefix; v1 stays unchanged, v2 adds $bytes_sent (replacing $body_bytes_sent), $request_length, $upstream_response_time, and $upstream_status. File ingest gains the same versioning, and the legacy positional file format is removed (no live deployments). Prometheus exposition is rewritten: - nginx_http_bytes_sent and nginx_http_request_duration_seconds gain a source_tag label. - nginx_http_requests_by_source_total gains status_class. - New v2-only metrics: nginx_http_request_bytes, nginx_http_upstream_duration_seconds, nginx_http_upstream_requests_total{status_class}. - Dropped nginx_http_response_body_bytes_by_source (subsumed by the dual-labeled bytes_sent metric). Adds 'make fixstyle' (gofmt -w) and clears all golangci-lint findings across the repo (errcheck, S1001, ST1005, unused). Docs in design.md FR-2/FR-8 and user-guide.md are rewritten to present v2 as the recommended log format.
2026-05-01 15:40:53 +02:00
parent d1a21a7a62
commit 6647f95be4
28 changed files with 931 additions and 724 deletions
@@ -5,7 +5,7 @@ PROTO_FILE   := $(PROTO_DIR)/logtail.proto
 GEN_FILES    := proto/logtailpb/logtail.pb.go proto/logtailpb/logtail_grpc.pb.go

 NATIVE_ARCH  := $(shell go env GOARCH)
-VERSION      := 0.9.2
+VERSION      := 1.0.1
 COMMIT_HASH  := $(shell git rev-parse --short HEAD 2>/dev/null || echo unknown)
 DATE         := $(shell date -u +%Y-%m-%dT%H:%M:%SZ)
 LDFLAGS      := -s -w \
@@ -28,7 +28,7 @@ GO_VERSION ?= 1.24.6
 # accepts. Older releases can't parse recent Go syntax.
 GOLANGCI_LINT_VERSION ?= 1.64.0

-.PHONY: help all build build-amd64 build-arm64 test lint proto pkg-deb docker docker-push clean \
+.PHONY: help all build build-amd64 build-arm64 test lint fixstyle proto pkg-deb docker docker-push clean \
        install-deps install-deps-apt install-deps-go install-deps-go-tools

 # help is the default target. Keep the list aligned with the .PHONY block above.
@@ -40,6 +40,7 @@ help:
 	@echo "  build-arm64          build all four binaries for linux/arm64 into build/arm64/"
 	@echo "  test                 run 'go test ./...'"
 	@echo "  lint                 run 'golangci-lint run ./...'"
+	@echo "  fixstyle             rewrite all .go files in place with 'gofmt -w'"
 	@echo "  proto                regenerate proto/logtailpb/*.pb.go from proto/logtail.proto"
 	@echo "  pkg-deb              build amd64 + arm64 .deb packages (requires dpkg-deb)"
 	@echo "  docker               buildx --load two tags ($(IMAGE):v$(VERSION), $(IMAGE):latest) for the native arch"
@@ -74,6 +75,11 @@ test: $(GEN_FILES)
 lint:
 	golangci-lint run ./...

+# fixstyle rewrites all tracked .go files with gofmt. Generated proto files are
+# excluded — they are regenerated from .proto via `make proto`.
+fixstyle:
+	gofmt -w $(shell find . -name '*.go' -not -path './proto/logtailpb/*')
+
 proto: $(GEN_FILES)

 # protoc's go_package option places output at the go_package path (not source-relative).
@@ -262,7 +262,7 @@ func startFakeCollector(t *testing.T, snaps []*pb.Snapshot) string {
 	}
 	srv := grpc.NewServer()
 	pb.RegisterLogtailServiceServer(srv, &fakeCollector{snaps: snaps})
-	go srv.Serve(lis)
+	go func() { _ = srv.Serve(lis) }()
 	t.Cleanup(srv.Stop)
 	return lis.Addr().String()
 }
@@ -310,7 +310,7 @@ func TestGRPCEndToEnd(t *testing.T) {
 	}
 	grpcSrv := grpc.NewServer()
 	pb.RegisterLogtailServiceServer(grpcSrv, NewServer(cache, "agg-test", NewTargetRegistry(nil)))
-	go grpcSrv.Serve(lis)
+	go func() { _ = grpcSrv.Serve(lis) }()
 	defer grpcSrv.Stop()

 	conn, err := grpc.NewClient(lis.Addr().String(),
@@ -318,7 +318,7 @@ func TestGRPCEndToEnd(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()
 	client := pb.NewLogtailServiceClient(conn)
 	qctx, qcancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer qcancel()
@@ -84,7 +84,7 @@ func dumpCollector(ctx context.Context, addr string) (fine, coarse []st.Snapshot
 	if err != nil {
 		return nil, nil, err
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()

 	client := pb.NewLogtailServiceClient(conn)
 	stream, err := client.DumpSnapshots(ctx, &pb.DumpSnapshotsRequest{})
@@ -97,15 +97,11 @@ func (c *Cache) LoadHistorical(fine, coarse []st.Snapshot) {
 	c.mu.Lock()
 	defer c.mu.Unlock()

-	for i, snap := range fine {
-		c.fineRing[i] = snap
-	}
+	copy(c.fineRing[:], fine)
 	c.fineFilled = len(fine)
 	c.fineHead = len(fine) % st.FineRingSize

-	for i, snap := range coarse {
-		c.coarseRing[i] = snap
-	}
+	copy(c.coarseRing[:], coarse)
 	c.coarseFilled = len(coarse)
 	c.coarseHead = len(coarse) % st.CoarseRingSize
 }
@@ -77,7 +77,7 @@ func (cs *CollectorSub) stream(ctx context.Context) (bool, error) {
 	if err != nil {
 		return false, err
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()

 	client := pb.NewLogtailServiceClient(conn)
 	stream, err := client.StreamSnapshots(ctx, &pb.SnapshotRequest{})
@@ -1,7 +1,6 @@
 package main

 import (
-	"bytes"
 	"context"
 	"encoding/json"
 	"net"
@@ -11,7 +10,6 @@ import (

 	pb "git.ipng.ch/ipng/nginx-logtail/proto/logtailpb"
 	"google.golang.org/grpc"
-	"google.golang.org/grpc/credentials/insecure"
 )

 // --- Unit tests ---
@@ -149,21 +147,11 @@ func startFake(t *testing.T, fs *fakeServer) string {
 	}
 	srv := grpc.NewServer()
 	pb.RegisterLogtailServiceServer(srv, fs)
-	go srv.Serve(lis)
+	go func() { _ = srv.Serve(lis) }()
 	t.Cleanup(srv.GracefulStop)
 	return lis.Addr().String()
 }

-func dialTest(t *testing.T, addr string) pb.LogtailServiceClient {
-	t.Helper()
-	conn, err := grpc.NewClient(addr, grpc.WithTransportCredentials(insecure.NewCredentials()))
-	if err != nil {
-		t.Fatal(err)
-	}
-	t.Cleanup(func() { conn.Close() })
-	return pb.NewLogtailServiceClient(conn)
-}
-
 // --- TopN tests ---

 func TestTopNSingleTarget(t *testing.T) {
@@ -225,23 +213,7 @@ func TestTopNJSON(t *testing.T) {
 	})

 	results := fanOutTopN([]string{addr}, nil, pb.GroupBy_WEBSITE, 10, pb.Window_W5M)
-	var buf bytes.Buffer
-	// Redirect stdout not needed; call JSON formatter directly.
 	r := results[0]
-	// Build expected JSON by calling printTopNJSON with a captured stdout.
-	// We test indirectly: marshal manually and compare fields.
-	type entry struct {
-		Label string `json:"label"`
-		Count int64  `json:"count"`
-	}
-	type out struct {
-		Source  string  `json:"source"`
-		Target  string  `json:"target"`
-		Entries []entry `json:"entries"`
-	}
-	_ = buf
-	_ = r
-	// Verify the response fields are correct for JSON serialization.
 	if r.resp.Source != "agg" {
 		t.Errorf("source = %q", r.resp.Source)
 	}
@@ -24,7 +24,7 @@ type streamEvent struct {
 func runStream(args []string) {
 	fs := flag.NewFlagSet("stream", flag.ExitOnError)
 	sf, targetFlag := bindShared(fs)
-	fs.Parse(args)
+	_ = fs.Parse(args) // ExitOnError: only returns nil here
 	sf.resolve(*targetFlag)

 	filter := buildFilter(sf)
@@ -85,7 +85,7 @@ func streamOnce(ctx context.Context, addr string, filter *pb.Filter, events chan
 	if err != nil {
 		return err
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()

 	stream, err := client.StreamSnapshots(ctx, &pb.SnapshotRequest{})
 	if err != nil {
@@ -18,7 +18,7 @@ func runTargets(args []string) {
 		fmt.Fprintln(os.Stderr, "usage: logtail-cli targets [--target host:port] [--json]")
 		fs.PrintDefaults()
 	}
-	fs.Parse(args)
+	_ = fs.Parse(args) // ExitOnError: only returns nil here
 	sf.resolve(*target)

 	for _, addr := range sf.targets {
@@ -30,7 +30,7 @@ func runTargets(args []string) {
 		ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 		resp, err := client.ListTargets(ctx, &pb.ListTargetsRequest{})
 		cancel()
-		conn.Close()
+		_ = conn.Close()
 		if err != nil {
 			fmt.Fprintf(os.Stderr, "targets: %s: %v\n", addr, err)
 			continue
@@ -43,7 +43,7 @@ func runTargets(args []string) {
 				Addr        string `json:"addr"`
 			}
 			for _, t := range resp.Targets {
-				json.NewEncoder(os.Stdout).Encode(row{QueryTarget: addr, Name: t.Name, Addr: t.Addr})
+				_ = json.NewEncoder(os.Stdout).Encode(row{QueryTarget: addr, Name: t.Name, Addr: t.Addr})
 			}
 		} else {
 			if len(sf.targets) > 1 {
@@ -24,7 +24,7 @@ func runTopN(args []string) {
 	n := fs.Int("n", 10, "number of entries")
 	window := fs.String("window", "5m", "time window: 1m 5m 15m 60m 6h 24h")
 	groupBy := fs.String("group-by", "website", "group by: website prefix uri status")
-	fs.Parse(args)
+	_ = fs.Parse(args) // ExitOnError: only returns nil here
 	sf.resolve(*targetFlag)

 	win := parseWindow(*window)
@@ -66,7 +66,7 @@ func fanOutTopN(targets []string, filter *pb.Filter, groupBy pb.GroupBy, n int,
 				results[i].resp = &pb.TopNResponse{}
 				return
 			}
-			defer conn.Close()
+			defer func() { _ = conn.Close() }()
 			ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 			defer cancel()
 			resp, err := client.TopN(ctx, &pb.TopNRequest{
@@ -22,7 +22,7 @@ func runTrend(args []string) {
 	fs := flag.NewFlagSet("trend", flag.ExitOnError)
 	sf, targetFlag := bindShared(fs)
 	window := fs.String("window", "5m", "time window: 1m 5m 15m 60m 6h 24h")
-	fs.Parse(args)
+	_ = fs.Parse(args) // ExitOnError: only returns nil here
 	sf.resolve(*targetFlag)

 	win := parseWindow(*window)
@@ -63,7 +63,7 @@ func fanOutTrend(targets []string, filter *pb.Filter, window pb.Window) []trendR
 				results[i].resp = &pb.TrendResponse{}
 				return
 			}
-			defer conn.Close()
+			defer func() { _ = conn.Close() }()
 			ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 			defer cancel()
 			resp, err := client.Trend(ctx, &pb.TrendRequest{
@@ -167,9 +167,3 @@ func parseGroupBy(s string) pb.GroupBy {
 		panic("unreachable")
 	}
 }
-
-func dieUsage(fs *flag.FlagSet, msg string) {
-	fmt.Fprintln(os.Stderr, msg)
-	fs.PrintDefaults()
-	os.Exit(1)
-}
@@ -16,17 +16,17 @@ func printTable(w io.Writer, rows [][]string) {
 	}
 	tw := tabwriter.NewWriter(w, 0, 0, 2, ' ', 0)
 	for i, row := range rows {
-		fmt.Fprintln(tw, strings.Join(row, "\t"))
+		_, _ = fmt.Fprintln(tw, strings.Join(row, "\t"))
 		if i == 0 {
 			// Print a divider matching the header width.
 			dashes := make([]string, len(row))
 			for j, h := range row {
 				dashes[j] = strings.Repeat("-", len(h))
 			}
-			fmt.Fprintln(tw, strings.Join(dashes, "\t"))
+			_, _ = fmt.Fprintln(tw, strings.Join(dashes, "\t"))
 		}
 	}
-	tw.Flush()
+	_ = tw.Flush()
 }

 // fmtCount formats a count with a space as the thousands separator.
@@ -128,7 +128,7 @@ func collectPatterns(logPaths, logsFile string) []string {
 		if err != nil {
 			log.Fatalf("collector: cannot open --logs-file %s: %v", logsFile, err)
 		}
-		defer f.Close()
+		defer func() { _ = f.Close() }()
 		sc := bufio.NewScanner(f)
 		for sc.Scan() {
 			if p := strings.TrimSpace(sc.Text()); p != "" && !strings.HasPrefix(p, "#") {
@@ -8,6 +8,13 @@ import (
 )

 // LogRecord holds the dimensions extracted from a single nginx log line.
+//
+// BytesSent carries $body_bytes_sent for v1 records and $bytes_sent for v2
+// records — operators see a small step up when emitters move to v2 because v2
+// includes header overhead.
+//
+// RequestLength, UpstreamResponseTime, UpstreamStatus, HasUpstream are v2-only.
+// In v1 records HasUpstream is always false and the related fields are zero.
 type LogRecord struct {
 	Website              string
 	ClientPrefix         string
@@ -16,79 +23,42 @@ type LogRecord struct {
 	IsTor                bool
 	ASN                  int32
 	Method               string
-	BodyBytesSent int64
+	BytesSent            int64
+	RequestLength        int64
 	RequestTime          float64
+	UpstreamResponseTime float64
+	UpstreamStatus       string
+	HasUpstream          bool
 	SourceTag            string
 }

-// fileSourceTag is the SourceTag assigned to records read from on-disk log
-// files, which pre-date the tag concept. Mirrors nginx's fallback label.
-const fileSourceTag = "direct"
-
-// ParseLine parses a tab-separated logtail log line from a file:
-//
-//	$host \t $remote_addr \t $msec \t $request_method \t $request_uri \t $status \t $body_bytes_sent \t $request_time \t $is_tor \t $asn
-//
-// The is_tor (field 9) and asn (field 10) fields are optional for backward
-// compatibility with older log files that omit them; they default to false/0
-// when absent. SourceTag is always set to "direct" (file origin has no tag).
-// Returns false for lines with fewer than 8 fields.
+// ParseLine parses a versioned nginx-logtail line. Both file ingest and UDP
+// ingest funnel through here. Every line MUST start with "v<N>\t"; unknown or
+// missing versions return false so operators can ship a parser update before
+// the emitter switches.
 func ParseLine(line string, v4bits, v6bits int) (LogRecord, bool) {
-	fields := strings.SplitN(line, "\t", 10)
-	if len(fields) < 8 {
-		return LogRecord{}, false
-	}
-	prefix, ok := truncateIP(fields[1], v4bits, v6bits)
-	if !ok {
-		return LogRecord{}, false
-	}
-	isTor := len(fields) >= 9 && fields[8] == "1"
-	var asn int32
-	if len(fields) == 10 {
-		if n, err := strconv.ParseInt(fields[9], 10, 32); err == nil {
-			asn = int32(n)
-		}
-	}
-	return LogRecord{
-		Website:       fields[0],
-		ClientPrefix:  prefix,
-		URI:           stripQuery(fields[4]),
-		Status:        fields[5],
-		IsTor:         isTor,
-		ASN:           asn,
-		Method:        fields[3],
-		BodyBytesSent: parseInt(fields[6]),
-		RequestTime:   parseFloat(fields[7]),
-		SourceTag:     fileSourceTag,
-	}, true
-}
-
-// ParseUDPLine dispatches on the version prefix emitted by
-// nginx-ipng-stats-plugin's ipng_stats_logtail directive. The wire format is
-// "v<N>\t<payload>", where <payload> is version-specific. Unknown or missing
-// versions return false so operators can roll out a v2 parser before
-// upgrading emitters.
-func ParseUDPLine(line string, v4bits, v6bits int) (LogRecord, bool) {
 	i := strings.IndexByte(line, '\t')
 	if i < 0 {
 		return LogRecord{}, false
 	}
 	switch line[:i] {
 	case "v1":
-		return parseUDPLineV1(line[i+1:], v4bits, v6bits)
+		return parseV1(line[i+1:], v4bits, v6bits)
+	case "v2":
+		return parseV2(line[i+1:], v4bits, v6bits)
 	default:
 		return LogRecord{}, false
 	}
 }

-// parseUDPLineV1 parses the v1 payload (12 tab-separated fields):
+// parseV1 parses the v1 payload (12 tab-separated fields):
 //
 //	$host \t $remote_addr \t $request_method \t $request_uri \t $status \t
 //	$body_bytes_sent \t $request_time \t $is_tor \t $asn \t
 //	$ipng_source_tag \t $server_addr \t $scheme
 //
-// server_addr and scheme are parsed but discarded.
-func parseUDPLineV1(payload string, v4bits, v6bits int) (LogRecord, bool) {
+// $server_addr and $scheme are parsed but discarded.
+func parseV1(payload string, v4bits, v6bits int) (LogRecord, bool) {
 	fields := strings.Split(payload, "\t")
 	if len(fields) != 12 {
 		return LogRecord{}, false
@@ -109,12 +79,71 @@ func parseUDPLineV1(payload string, v4bits, v6bits int) (LogRecord, bool) {
 		IsTor:        fields[7] == "1",
 		ASN:          asn,
 		Method:       fields[2],
-		BodyBytesSent: parseInt(fields[5]),
+		BytesSent:    parseInt(fields[5]),
 		RequestTime:  parseFloat(fields[6]),
 		SourceTag:    fields[9],
 	}, true
 }

+// parseV2 parses the v2 payload (15 tab-separated fields):
+//
+//	$host \t $remote_addr \t $request_method \t $request_uri \t $status \t
+//	$bytes_sent \t $request_length \t $request_time \t
+//	$upstream_response_time \t $upstream_status \t
+//	$is_tor \t $asn \t $ipng_source_tag \t $server_addr \t $scheme
+//
+// $upstream_response_time and $upstream_status are "-" (or empty) when nginx
+// served the response directly — HasUpstream is left false in that case.
+// When nginx retried across multiple upstreams the fields are comma-separated;
+// the parser keeps the last entry, since that's the upstream that actually
+// served the response. $server_addr and $scheme are parsed but discarded.
+func parseV2(payload string, v4bits, v6bits int) (LogRecord, bool) {
+	fields := strings.Split(payload, "\t")
+	if len(fields) != 15 {
+		return LogRecord{}, false
+	}
+	prefix, ok := truncateIP(fields[1], v4bits, v6bits)
+	if !ok {
+		return LogRecord{}, false
+	}
+	var asn int32
+	if n, err := strconv.ParseInt(fields[11], 10, 32); err == nil {
+		asn = int32(n)
+	}
+	r := LogRecord{
+		Website:       fields[0],
+		ClientPrefix:  prefix,
+		URI:           stripQuery(fields[3]),
+		Status:        fields[4],
+		IsTor:         fields[10] == "1",
+		ASN:           asn,
+		Method:        fields[2],
+		BytesSent:     parseInt(fields[5]),
+		RequestLength: parseInt(fields[6]),
+		RequestTime:   parseFloat(fields[7]),
+		SourceTag:     fields[12],
+	}
+	if fields[8] != "-" && fields[8] != "" {
+		timeStr := lastCommaPart(fields[8])
+		statusStr := lastCommaPart(fields[9])
+		if t, err := strconv.ParseFloat(timeStr, 64); err == nil {
+			r.UpstreamResponseTime = t
+			r.UpstreamStatus = statusStr
+			r.HasUpstream = true
+		}
+	}
+	return r, true
+}
+
+// lastCommaPart returns the substring after the last ", " (nginx's separator
+// for retried upstreams). Plain values pass through unchanged.
+func lastCommaPart(s string) string {
+	if i := strings.LastIndex(s, ", "); i >= 0 {
+		return s[i+2:]
+	}
+	return s
+}
+
 func stripQuery(uri string) string {
 	if i := strings.IndexByte(uri, '?'); i >= 0 {
 		return uri[:i]
@@ -4,215 +4,7 @@ import (
 	"testing"
 )

-func TestParseLine(t *testing.T) {
-	good := "www.example.com\t1.2.3.4\t1741954800.123\tGET\t/api/v1/search?q=foo&x=1\t200\t1452\t0.043"
-
-	tests := []struct {
-		name   string
-		line   string
-		wantOK bool
-		want   LogRecord
-	}{
-		{
-			name:   "normal IPv4 line strips query string",
-			line:   good,
-			wantOK: true,
-			want: LogRecord{
-				Website:       "www.example.com",
-				ClientPrefix:  "1.2.3.0/24",
-				URI:           "/api/v1/search",
-				Status:        "200",
-				Method:        "GET",
-				BodyBytesSent: 1452,
-				RequestTime:   0.043,
-				SourceTag:     "direct",
-			},
-		},
-		{
-			name:   "URI with no query string",
-			line:   "host\t10.0.0.1\t0\tPOST\t/submit\t201\t0\t0.001",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "host",
-				ClientPrefix: "10.0.0.0/24",
-				URI:          "/submit",
-				Status:       "201",
-				Method:       "POST",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "IPv6 address truncated to /48",
-			line:   "host\t2001:db8:cafe::1\t0\tGET\t/\t200\t0\t0.001",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "host",
-				ClientPrefix: "2001:db8:cafe::/48",
-				URI:          "/",
-				Status:       "200",
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "too few fields returns false",
-			line:   "host\t1.2.3.4\t0\tGET\t/",
-			wantOK: false,
-		},
-		{
-			name:   "empty line returns false",
-			line:   "",
-			wantOK: false,
-		},
-		{
-			name:   "invalid IP returns false",
-			line:   "host\tnot-an-ip\t0\tGET\t/\t200\t0\t0.001",
-			wantOK: false,
-		},
-		{
-			name:   "status 429",
-			line:   "api.example.com\t5.6.7.8\t0\tGET\t/rate-limited\t429\t0\t0.001",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "api.example.com",
-				ClientPrefix: "5.6.7.0/24",
-				URI:          "/rate-limited",
-				Status:       "429",
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "is_tor=1 sets IsTor true",
-			line:   "tor.example.com\t1.2.3.4\t0\tGET\t/\t200\t0\t0.001\t1",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "tor.example.com",
-				ClientPrefix: "1.2.3.0/24",
-				URI:          "/",
-				Status:       "200",
-				IsTor:        true,
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "is_tor=0 sets IsTor false",
-			line:   "normal.example.com\t1.2.3.4\t0\tGET\t/\t200\t0\t0.001\t0",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "normal.example.com",
-				ClientPrefix: "1.2.3.0/24",
-				URI:          "/",
-				Status:       "200",
-				IsTor:        false,
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "missing is_tor field defaults to false (backward compat)",
-			line:   "old.example.com\t1.2.3.4\t0\tGET\t/\t200\t0\t0.001",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "old.example.com",
-				ClientPrefix: "1.2.3.0/24",
-				URI:          "/",
-				Status:       "200",
-				IsTor:        false,
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "asn field parsed",
-			line:   "asn.example.com\t1.2.3.4\t0\tGET\t/\t200\t0\t0.001\t0\t12345",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "asn.example.com",
-				ClientPrefix: "1.2.3.0/24",
-				URI:          "/",
-				Status:       "200",
-				IsTor:        false,
-				ASN:          12345,
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "asn field with is_tor=1",
-			line:   "both.example.com\t1.2.3.4\t0\tGET\t/\t200\t0\t0.001\t1\t65535",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "both.example.com",
-				ClientPrefix: "1.2.3.0/24",
-				URI:          "/",
-				Status:       "200",
-				IsTor:        true,
-				ASN:          65535,
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "missing asn field defaults to 0 (backward compat)",
-			line:   "noasn.example.com\t1.2.3.4\t0\tGET\t/\t200\t0\t0.001\t1",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "noasn.example.com",
-				ClientPrefix: "1.2.3.0/24",
-				URI:          "/",
-				Status:       "200",
-				IsTor:        true,
-				ASN:          0,
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-		{
-			name:   "invalid asn field defaults to 0",
-			line:   "badann.example.com\t1.2.3.4\t0\tGET\t/\t200\t0\t0.001\t0\tnot-a-number",
-			wantOK: true,
-			want: LogRecord{
-				Website:      "badann.example.com",
-				ClientPrefix: "1.2.3.0/24",
-				URI:          "/",
-				Status:       "200",
-				IsTor:        false,
-				ASN:          0,
-				Method:       "GET",
-				RequestTime:  0.001,
-				SourceTag:    "direct",
-			},
-		},
-	}
-
-	for _, tc := range tests {
-		t.Run(tc.name, func(t *testing.T) {
-			got, ok := ParseLine(tc.line, 24, 48)
-			if ok != tc.wantOK {
-				t.Fatalf("ParseLine ok=%v, want %v", ok, tc.wantOK)
-			}
-			if !tc.wantOK {
-				return
-			}
-			if got != tc.want {
-				t.Errorf("got %+v, want %+v", got, tc.want)
-			}
-		})
-	}
-}
-
-func TestParseUDPLine(t *testing.T) {
+func TestParseLineV1(t *testing.T) {
 	// v1 \t host \t remote_addr \t method \t uri \t status \t body_bytes \t req_time \t
 	// is_tor \t asn \t source_tag \t server_addr \t scheme
 	good := "v1\twww.example.com\t1.2.3.4\tGET\t/api/v1/search?q=foo\t200\t1452\t0.043\t0\t12345\tcdn\t10.0.0.1\thttps"
@@ -235,7 +27,7 @@ func TestParseUDPLine(t *testing.T) {
 				IsTor:        false,
 				ASN:          12345,
 				Method:       "GET",
-				BodyBytesSent: 1452,
+				BytesSent:    1452,
 				RequestTime:  0.043,
 				SourceTag:    "cdn",
 			},
@@ -252,7 +44,7 @@ func TestParseUDPLine(t *testing.T) {
 				IsTor:        true,
 				ASN:          65535,
 				Method:       "GET",
-				BodyBytesSent: 0,
+				BytesSent:    0,
 				RequestTime:  0,
 				SourceTag:    "direct",
 			},
@@ -272,28 +64,13 @@ func TestParseUDPLine(t *testing.T) {
 			line:   "v1\th\tnope\tGET\t/\t200\t0\t0\t0\t0\ttag\t10.0.0.1\thttp",
 			wantOK: false,
 		},
-		{
-			name:   "unknown version rejected (future v2)",
-			line:   "v2\twww.example.com\t1.2.3.4\tGET\t/\t200\t0\t0\t0\t0\ttag\t10.0.0.1\thttp",
-			wantOK: false,
-		},
-		{
-			name:   "missing version prefix rejected (legacy 12-field line)",
-			line:   "www.example.com\t1.2.3.4\tGET\t/\t200\t0\t0\t0\t0\ttag\t10.0.0.1\thttp",
-			wantOK: false,
-		},
-		{
-			name:   "no tab at all rejected",
-			line:   "v1",
-			wantOK: false,
-		},
 	}

 	for _, tc := range tests {
 		t.Run(tc.name, func(t *testing.T) {
-			got, ok := ParseUDPLine(tc.line, 24, 48)
+			got, ok := ParseLine(tc.line, 24, 48)
 			if ok != tc.wantOK {
-				t.Fatalf("ParseUDPLine ok=%v, want %v; got=%+v", ok, tc.wantOK, got)
+				t.Fatalf("ParseLine ok=%v, want %v; got=%+v", ok, tc.wantOK, got)
 			}
 			if !tc.wantOK {
 				return
@@ -305,6 +82,148 @@ func TestParseUDPLine(t *testing.T) {
 	}
 }

+func TestParseLineV2(t *testing.T) {
+	// v2 \t host \t remote_addr \t method \t uri \t status \t bytes_sent \t request_length \t
+	// request_time \t upstream_response_time \t upstream_status \t is_tor \t asn \t
+	// source_tag \t server_addr \t scheme
+	full := "v2\twww.example.com\t1.2.3.4\tGET\t/api/v1/search?q=foo\t200\t1500\t421\t0.043\t0.012\t200\t0\t12345\tcdn\t10.0.0.1\thttps"
+
+	tests := []struct {
+		name   string
+		line   string
+		wantOK bool
+		want   LogRecord
+	}{
+		{
+			name:   "v2 with upstream",
+			line:   full,
+			wantOK: true,
+			want: LogRecord{
+				Website:              "www.example.com",
+				ClientPrefix:         "1.2.3.0/24",
+				URI:                  "/api/v1/search",
+				Status:               "200",
+				ASN:                  12345,
+				Method:               "GET",
+				BytesSent:            1500,
+				RequestLength:        421,
+				RequestTime:          0.043,
+				UpstreamResponseTime: 0.012,
+				UpstreamStatus:       "200",
+				HasUpstream:          true,
+				SourceTag:            "cdn",
+			},
+		},
+		{
+			name:   "v2 no upstream (dash sentinels)",
+			line:   "v2\twww.example.com\t1.2.3.4\tGET\t/static.html\t200\t900\t300\t0.001\t-\t-\t0\t0\tdirect\t10.0.0.1\thttps",
+			wantOK: true,
+			want: LogRecord{
+				Website:       "www.example.com",
+				ClientPrefix:  "1.2.3.0/24",
+				URI:           "/static.html",
+				Status:        "200",
+				Method:        "GET",
+				BytesSent:     900,
+				RequestLength: 300,
+				RequestTime:   0.001,
+				SourceTag:     "direct",
+			},
+		},
+		{
+			name:   "v2 no upstream (empty fields)",
+			line:   "v2\thh\t1.2.3.4\tGET\t/\t301\t200\t100\t0\t\t\t0\t0\tdirect\t10.0.0.1\thttps",
+			wantOK: true,
+			want: LogRecord{
+				Website:       "hh",
+				ClientPrefix:  "1.2.3.0/24",
+				URI:           "/",
+				Status:        "301",
+				Method:        "GET",
+				BytesSent:     200,
+				RequestLength: 100,
+				SourceTag:     "direct",
+			},
+		},
+		{
+			name:   "v2 retried upstreams (comma-separated, last wins)",
+			line:   "v2\twww.example.com\t1.2.3.4\tGET\t/api\t502\t900\t300\t1.500\t0.500, 1.000\t504, 502\t0\t0\tcdn\t10.0.0.1\thttps",
+			wantOK: true,
+			want: LogRecord{
+				Website:              "www.example.com",
+				ClientPrefix:         "1.2.3.0/24",
+				URI:                  "/api",
+				Status:               "502",
+				Method:               "GET",
+				BytesSent:            900,
+				RequestLength:        300,
+				RequestTime:          1.500,
+				UpstreamResponseTime: 1.000,
+				UpstreamStatus:       "502",
+				HasUpstream:          true,
+				SourceTag:            "cdn",
+			},
+		},
+		{
+			name:   "v2 wrong field count (14) rejected",
+			line:   "v2\twww.example.com\t1.2.3.4\tGET\t/\t200\t0\t0\t0\t-\t-\t0\t0\tcdn\t10.0.0.1",
+			wantOK: false,
+		},
+		{
+			name:   "v2 bad IP rejected",
+			line:   "v2\thh\tnope\tGET\t/\t200\t0\t0\t0\t-\t-\t0\t0\tcdn\t10.0.0.1\thttps",
+			wantOK: false,
+		},
+		{
+			name:   "v2 bad upstream time leaves HasUpstream=false",
+			line:   "v2\thh\t1.2.3.4\tGET\t/\t200\t0\t0\t0\tnotanumber\t200\t0\t0\tcdn\t10.0.0.1\thttps",
+			wantOK: true,
+			want: LogRecord{
+				Website:      "hh",
+				ClientPrefix: "1.2.3.0/24",
+				URI:          "/",
+				Status:       "200",
+				Method:       "GET",
+				SourceTag:    "cdn",
+			},
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			got, ok := ParseLine(tc.line, 24, 48)
+			if ok != tc.wantOK {
+				t.Fatalf("ParseLine ok=%v, want %v; got=%+v", ok, tc.wantOK, got)
+			}
+			if !tc.wantOK {
+				return
+			}
+			if got != tc.want {
+				t.Errorf("got %+v, want %+v", got, tc.want)
+			}
+		})
+	}
+}
+
+func TestParseLineRejections(t *testing.T) {
+	tests := []struct {
+		name string
+		line string
+	}{
+		{"empty line", ""},
+		{"no tab at all", "v1"},
+		{"unknown version v3", "v3\twww.example.com\t1.2.3.4\tGET\t/\t200\t0\t0\t0\t0\ttag\t10.0.0.1\thttp"},
+		{"missing version prefix (legacy file format)", "www.example.com\t1.2.3.4\t1741954800.123\tGET\t/api\t200\t1452\t0.043"},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			if _, ok := ParseLine(tc.line, 24, 48); ok {
+				t.Errorf("expected rejection for %q", tc.line)
+			}
+		})
+	}
+}
+
 func TestTruncateIP(t *testing.T) {
 	tests := []struct {
 		addr string
@@ -14,27 +14,46 @@ const promNumBodyBounds = 7

 var promBodyBounds = [promNumBodyBounds]int64{256, 1024, 4096, 16384, 65536, 262144, 1048576}

-// Request-time histogram bucket upper bounds in seconds (standard Prometheus defaults).
+// Duration histogram bucket upper bounds in seconds (Prometheus defaults).
 const promNumTimeBounds = 11

 var promTimeBounds = [promNumTimeBounds]float64{.005, .01, .025, .05, .1, .25, .5, 1, 2.5, 5, 10}

 const promCounterCap = 250_000 // safety cap on {host,method,status} counter entries

-// promCounterKey is the label set for per-request counters.
+// promCounterKey is the label set for the per-request counter.
 type promCounterKey struct {
 	Host   string
 	Method string
 	Status string
 }

-// promBodyEntry holds the body_bytes_sent histogram for one host.
+// hostSourceKey labels histograms by {host, source_tag}.
+type hostSourceKey struct {
+	Host      string
+	SourceTag string
+}
+
+// sourceClassKey labels the source-tag rollup counter.
+type sourceClassKey struct {
+	SourceTag   string
+	StatusClass string
+}
+
+// upstreamKey labels the upstream-only request counter.
+type upstreamKey struct {
+	Host        string
+	SourceTag   string
+	StatusClass string // class of $upstream_status
+}
+
+// promBodyEntry holds one body-size histogram (one label-set worth).
 type promBodyEntry struct {
 	buckets [promNumBodyBounds + 1]int64 // indices 0..N-1: le=bound[i]; index N: le=+Inf
 	sum     int64
 }

-// promTimeEntry holds the request_time histogram for one host.
+// promTimeEntry holds one duration histogram (one label-set worth).
 type promTimeEntry struct {
 	buckets [promNumTimeBounds + 1]int64
 	sum     float64
@@ -47,28 +66,29 @@ type promTimeEntry struct {
 type PromStore struct {
 	mu               sync.Mutex
 	counters         map[promCounterKey]int64
-	body     map[string]*promBodyEntry // keyed by host
-	reqTime  map[string]*promTimeEntry // keyed by host
+	bytesSent        map[hostSourceKey]*promBodyEntry
+	requestDuration  map[hostSourceKey]*promTimeEntry
+	requestBytes     map[hostSourceKey]*promBodyEntry // v2 only
+	upstreamDuration map[hostSourceKey]*promTimeEntry // v2 only
+	upstreamCounters map[upstreamKey]int64            // v2 only
+	sourceCounters   map[sourceClassKey]int64

-	// per-source_tag rollups (parallel series, not a cross-product with host)
-	sourceCounters map[string]int64          // keyed by source_tag
-	sourceBody     map[string]*promBodyEntry // keyed by source_tag
-
-	// UDP ingest counters — protected by their own atomic-friendly lock.
 	udpMu               sync.Mutex
-	udpPacketsReceived int64 // datagrams read off the socket
-	udpLoglinesSuccess int64 // successfully parsed
-	udpLoglinesConsumed int64 // successfully forwarded to the store channel
+	udpPacketsReceived  int64
+	udpLoglinesSuccess  int64
+	udpLoglinesConsumed int64
 }

 // NewPromStore returns an empty PromStore ready for use.
 func NewPromStore() *PromStore {
 	return &PromStore{
 		counters:         make(map[promCounterKey]int64, 1024),
-		body:           make(map[string]*promBodyEntry, 64),
-		reqTime:        make(map[string]*promTimeEntry, 64),
-		sourceCounters: make(map[string]int64, 32),
-		sourceBody:     make(map[string]*promBodyEntry, 32),
+		bytesSent:        make(map[hostSourceKey]*promBodyEntry, 64),
+		requestDuration:  make(map[hostSourceKey]*promTimeEntry, 64),
+		requestBytes:     make(map[hostSourceKey]*promBodyEntry, 64),
+		upstreamDuration: make(map[hostSourceKey]*promTimeEntry, 64),
+		upstreamCounters: make(map[upstreamKey]int64, 64),
+		sourceCounters:   make(map[sourceClassKey]int64, 32),
 	}
 }

@@ -76,8 +96,11 @@ func NewPromStore() *PromStore {
 // Must be called from a single goroutine.
 func (p *PromStore) Ingest(r LogRecord) {
 	p.mu.Lock()
+	defer p.mu.Unlock()

-	// --- per-{host,method,status} request counter ---
+	hsk := hostSourceKey{Host: r.Website, SourceTag: r.SourceTag}
+
+	// nginx_http_requests_total{host,method,status} — capped.
 	ck := promCounterKey{Host: r.Website, Method: r.Method, Status: r.Status}
 	if _, ok := p.counters[ck]; ok {
 		p.counters[ck]++
@@ -85,37 +108,54 @@ func (p *PromStore) Ingest(r LogRecord) {
 		p.counters[ck] = 1
 	}

-	// --- body_bytes_sent histogram (keyed by host only) ---
-	observeBody(p.body, r.Website, r.BodyBytesSent)
-
-	// --- request_time histogram (keyed by host only) ---
-	te, ok := p.reqTime[r.Website]
-	if !ok {
-		te = &promTimeEntry{}
-		p.reqTime[r.Website] = te
-	}
-	for i, bound := range promTimeBounds {
-		if r.RequestTime <= bound {
-			te.buckets[i]++
-		}
-	}
-	te.buckets[promNumTimeBounds]++ // +Inf
-	te.sum += r.RequestTime
-
-	// --- per-source_tag rollups ---
-	p.sourceCounters[r.SourceTag]++
-	observeBody(p.sourceBody, r.SourceTag, r.BodyBytesSent)
-
-	p.mu.Unlock()
+	observeBody(p.bytesSent, hsk, r.BytesSent)
+	observeTime(p.requestDuration, hsk, r.RequestTime)
+	if r.RequestLength > 0 {
+		observeBody(p.requestBytes, hsk, r.RequestLength)
 	}

-// IncUDPPacket, IncUDPSuccess, and IncUDPConsumed bump their respective
-// UDP ingest counters. They are called from the UDP listener goroutine.
+	p.sourceCounters[sourceClassKey{
+		SourceTag:   r.SourceTag,
+		StatusClass: statusClass(r.Status),
+	}]++
+
+	if r.HasUpstream {
+		observeTime(p.upstreamDuration, hsk, r.UpstreamResponseTime)
+		p.upstreamCounters[upstreamKey{
+			Host:        r.Website,
+			SourceTag:   r.SourceTag,
+			StatusClass: statusClass(r.UpstreamStatus),
+		}]++
+	}
+}
+
+// IncUDPPacket, IncUDPSuccess, IncUDPConsumed bump UDP-ingest counters from
+// the listener goroutine.
 func (p *PromStore) IncUDPPacket()   { p.udpMu.Lock(); p.udpPacketsReceived++; p.udpMu.Unlock() }
 func (p *PromStore) IncUDPSuccess()  { p.udpMu.Lock(); p.udpLoglinesSuccess++; p.udpMu.Unlock() }
 func (p *PromStore) IncUDPConsumed() { p.udpMu.Lock(); p.udpLoglinesConsumed++; p.udpMu.Unlock() }

-func observeBody(m map[string]*promBodyEntry, key string, bytes int64) {
+// statusClass folds an HTTP status code into 2xx/3xx/4xx/5xx, with anything
+// else falling to "other" (including empty input).
+func statusClass(status string) string {
+	if status == "" {
+		return "other"
+	}
+	switch status[0] {
+	case '2':
+		return "2xx"
+	case '3':
+		return "3xx"
+	case '4':
+		return "4xx"
+	case '5':
+		return "5xx"
+	default:
+		return "other"
+	}
+}
+
+func observeBody(m map[hostSourceKey]*promBodyEntry, key hostSourceKey, bytes int64) {
 	e, ok := m[key]
 	if !ok {
 		e = &promBodyEntry{}
@@ -126,53 +166,77 @@ func observeBody(m map[string]*promBodyEntry, key string, bytes int64) {
 			e.buckets[i]++
 		}
 	}
-	e.buckets[promNumBodyBounds]++ // +Inf
+	e.buckets[promNumBodyBounds]++
 	e.sum += bytes
 }

+func observeTime(m map[hostSourceKey]*promTimeEntry, key hostSourceKey, seconds float64) {
+	e, ok := m[key]
+	if !ok {
+		e = &promTimeEntry{}
+		m[key] = e
+	}
+	for i, bound := range promTimeBounds {
+		if seconds <= bound {
+			e.buckets[i]++
+		}
+	}
+	e.buckets[promNumTimeBounds]++
+	e.sum += seconds
+}
+
 // ServeHTTP renders all metrics in the Prometheus text exposition format (0.0.4).
 func (p *PromStore) ServeHTTP(w http.ResponseWriter, _ *http.Request) {
-	// Snapshot everything under the lock, then render without holding it.
-	p.mu.Lock()
-
 	type counterSnap struct {
 		k promCounterKey
 		v int64
 	}
+	type bodySnap struct {
+		k hostSourceKey
+		e promBodyEntry
+	}
+	type timeSnap struct {
+		k hostSourceKey
+		e promTimeEntry
+	}
+	type upstreamCounterSnap struct {
+		k upstreamKey
+		v int64
+	}
+	type sourceCounterSnap struct {
+		k sourceClassKey
+		v int64
+	}
+
+	p.mu.Lock()
+
 	counters := make([]counterSnap, 0, len(p.counters))
 	for k, v := range p.counters {
 		counters = append(counters, counterSnap{k, v})
 	}
-
-	type bodySnap struct {
-		label string
-		e     promBodyEntry
+	bytesSnaps := make([]bodySnap, 0, len(p.bytesSent))
+	for k, e := range p.bytesSent {
+		bytesSnaps = append(bytesSnaps, bodySnap{k, *e})
 	}
-	bodySnaps := make([]bodySnap, 0, len(p.body))
-	for h, e := range p.body {
-		bodySnaps = append(bodySnaps, bodySnap{h, *e})
+	requestBytesSnaps := make([]bodySnap, 0, len(p.requestBytes))
+	for k, e := range p.requestBytes {
+		requestBytesSnaps = append(requestBytesSnaps, bodySnap{k, *e})
 	}
-
-	type timeSnap struct {
-		host string
-		e    promTimeEntry
+	requestDurationSnaps := make([]timeSnap, 0, len(p.requestDuration))
+	for k, e := range p.requestDuration {
+		requestDurationSnaps = append(requestDurationSnaps, timeSnap{k, *e})
 	}
-	timeSnaps := make([]timeSnap, 0, len(p.reqTime))
-	for h, e := range p.reqTime {
-		timeSnaps = append(timeSnaps, timeSnap{h, *e})
+	upstreamDurationSnaps := make([]timeSnap, 0, len(p.upstreamDuration))
+	for k, e := range p.upstreamDuration {
+		upstreamDurationSnaps = append(upstreamDurationSnaps, timeSnap{k, *e})
 	}
-
-	type sourceCounterSnap struct {
-		tag string
-		v   int64
+	upstreamCounters := make([]upstreamCounterSnap, 0, len(p.upstreamCounters))
+	for k, v := range p.upstreamCounters {
+		upstreamCounters = append(upstreamCounters, upstreamCounterSnap{k, v})
 	}
 	sourceCounters := make([]sourceCounterSnap, 0, len(p.sourceCounters))
-	for t, v := range p.sourceCounters {
-		sourceCounters = append(sourceCounters, sourceCounterSnap{t, v})
-	}
-	sourceBodySnaps := make([]bodySnap, 0, len(p.sourceBody))
-	for t, e := range p.sourceBody {
-		sourceBodySnaps = append(sourceBodySnaps, bodySnap{t, *e})
+	for k, v := range p.sourceCounters {
+		sourceCounters = append(sourceCounters, sourceCounterSnap{k, v})
 	}

 	p.mu.Unlock()
@@ -183,7 +247,6 @@ func (p *PromStore) ServeHTTP(w http.ResponseWriter, _ *http.Request) {
 	udpConsumed := p.udpLoglinesConsumed
 	p.udpMu.Unlock()

-	// Sort for stable, human-readable output.
 	sort.Slice(counters, func(i, j int) bool {
 		a, b := counters[i].k, counters[j].k
 		if a.Host != b.Host {
@@ -194,85 +257,139 @@ func (p *PromStore) ServeHTTP(w http.ResponseWriter, _ *http.Request) {
 		}
 		return a.Status < b.Status
 	})
-	sort.Slice(bodySnaps, func(i, j int) bool { return bodySnaps[i].label < bodySnaps[j].label })
-	sort.Slice(timeSnaps, func(i, j int) bool { return timeSnaps[i].host < timeSnaps[j].host })
-	sort.Slice(sourceCounters, func(i, j int) bool { return sourceCounters[i].tag < sourceCounters[j].tag })
-	sort.Slice(sourceBodySnaps, func(i, j int) bool { return sourceBodySnaps[i].label < sourceBodySnaps[j].label })
+	sortBody := func(s []bodySnap) {
+		sort.Slice(s, func(i, j int) bool {
+			a, b := s[i].k, s[j].k
+			if a.Host != b.Host {
+				return a.Host < b.Host
+			}
+			return a.SourceTag < b.SourceTag
+		})
+	}
+	sortTime := func(s []timeSnap) {
+		sort.Slice(s, func(i, j int) bool {
+			a, b := s[i].k, s[j].k
+			if a.Host != b.Host {
+				return a.Host < b.Host
+			}
+			return a.SourceTag < b.SourceTag
+		})
+	}
+	sortBody(bytesSnaps)
+	sortBody(requestBytesSnaps)
+	sortTime(requestDurationSnaps)
+	sortTime(upstreamDurationSnaps)
+	sort.Slice(upstreamCounters, func(i, j int) bool {
+		a, b := upstreamCounters[i].k, upstreamCounters[j].k
+		if a.Host != b.Host {
+			return a.Host < b.Host
+		}
+		if a.SourceTag != b.SourceTag {
+			return a.SourceTag < b.SourceTag
+		}
+		return a.StatusClass < b.StatusClass
+	})
+	sort.Slice(sourceCounters, func(i, j int) bool {
+		a, b := sourceCounters[i].k, sourceCounters[j].k
+		if a.SourceTag != b.SourceTag {
+			return a.SourceTag < b.SourceTag
+		}
+		return a.StatusClass < b.StatusClass
+	})

 	w.Header().Set("Content-Type", "text/plain; version=0.0.4; charset=utf-8")
 	bw := bufio.NewWriterSize(w, 256*1024)

-	// nginx_http_requests_total
-	fmt.Fprintln(bw, "# HELP nginx_http_requests_total Total number of HTTP requests processed.")
-	fmt.Fprintln(bw, "# TYPE nginx_http_requests_total counter")
+	// pf, pln are short helpers so the metric block reads cleanly. Errors on a
+	// bufio writer wrapping http.ResponseWriter mean the client disconnected;
+	// there's nothing useful to do mid-write — the next call will simply no-op.
+	pf := func(format string, a ...any) { _, _ = fmt.Fprintf(bw, format, a...) }
+	pln := func(s string) { _, _ = fmt.Fprintln(bw, s) }
+
+	pln("# HELP nginx_http_requests_total Total number of HTTP requests processed.")
+	pln("# TYPE nginx_http_requests_total counter")
 	for _, c := range counters {
-		fmt.Fprintf(bw, "nginx_http_requests_total{host=%q,method=%q,status=%q} %d\n",
+		pf("nginx_http_requests_total{host=%q,method=%q,status=%q} %d\n",
 			c.k.Host, c.k.Method, c.k.Status, c.v)
 	}

-	// nginx_http_response_body_bytes (histogram, labeled by host)
-	fmt.Fprintln(bw, "# HELP nginx_http_response_body_bytes HTTP response body size distribution in bytes.")
-	fmt.Fprintln(bw, "# TYPE nginx_http_response_body_bytes histogram")
-	for _, s := range bodySnaps {
-		writeBodyHistogram(bw, "nginx_http_response_body_bytes", "host", s.label, s.e)
+	pln("# HELP nginx_http_bytes_sent HTTP response size distribution in bytes (body for v1 records, full wire bytes for v2).")
+	pln("# TYPE nginx_http_bytes_sent histogram")
+	for _, s := range bytesSnaps {
+		writeBodyHistogramHS(bw, "nginx_http_bytes_sent", s.k, s.e)
 	}

-	// nginx_http_request_duration_seconds (histogram, labeled by host)
-	fmt.Fprintln(bw, "# HELP nginx_http_request_duration_seconds HTTP request processing time in seconds.")
-	fmt.Fprintln(bw, "# TYPE nginx_http_request_duration_seconds histogram")
-	for _, s := range timeSnaps {
-		for i, bound := range promTimeBounds {
-			fmt.Fprintf(bw, "nginx_http_request_duration_seconds_bucket{host=%q,le=%q} %d\n",
-				s.host, formatFloat(bound), s.e.buckets[i])
-		}
-		fmt.Fprintf(bw, "nginx_http_request_duration_seconds_bucket{host=%q,le=\"+Inf\"} %d\n",
-			s.host, s.e.buckets[promNumTimeBounds])
-		fmt.Fprintf(bw, "nginx_http_request_duration_seconds_count{host=%q} %d\n",
-			s.host, s.e.buckets[promNumTimeBounds])
-		fmt.Fprintf(bw, "nginx_http_request_duration_seconds_sum{host=%q} %g\n",
-			s.host, s.e.sum)
+	pln("# HELP nginx_http_request_bytes HTTP request size distribution in bytes (v2 emitters only).")
+	pln("# TYPE nginx_http_request_bytes histogram")
+	for _, s := range requestBytesSnaps {
+		writeBodyHistogramHS(bw, "nginx_http_request_bytes", s.k, s.e)
 	}

-	// nginx_http_requests_by_source_total (counter, labeled by source_tag)
-	fmt.Fprintln(bw, "# HELP nginx_http_requests_by_source_total HTTP requests rolled up by nginx source tag.")
-	fmt.Fprintln(bw, "# TYPE nginx_http_requests_by_source_total counter")
+	pln("# HELP nginx_http_request_duration_seconds HTTP request processing time in seconds.")
+	pln("# TYPE nginx_http_request_duration_seconds histogram")
+	for _, s := range requestDurationSnaps {
+		writeTimeHistogramHS(bw, "nginx_http_request_duration_seconds", s.k, s.e)
+	}
+
+	pln("# HELP nginx_http_upstream_duration_seconds Upstream response time in seconds (v2 emitters only).")
+	pln("# TYPE nginx_http_upstream_duration_seconds histogram")
+	for _, s := range upstreamDurationSnaps {
+		writeTimeHistogramHS(bw, "nginx_http_upstream_duration_seconds", s.k, s.e)
+	}
+
+	pln("# HELP nginx_http_upstream_requests_total Requests served via an upstream, by upstream-status class (v2 emitters only).")
+	pln("# TYPE nginx_http_upstream_requests_total counter")
+	for _, c := range upstreamCounters {
+		pf("nginx_http_upstream_requests_total{host=%q,source_tag=%q,status_class=%q} %d\n",
+			c.k.Host, c.k.SourceTag, c.k.StatusClass, c.v)
+	}
+
+	pln("# HELP nginx_http_requests_by_source_total HTTP requests rolled up by source_tag and status class.")
+	pln("# TYPE nginx_http_requests_by_source_total counter")
 	for _, c := range sourceCounters {
-		fmt.Fprintf(bw, "nginx_http_requests_by_source_total{source_tag=%q} %d\n", c.tag, c.v)
+		pf("nginx_http_requests_by_source_total{source_tag=%q,status_class=%q} %d\n",
+			c.k.SourceTag, c.k.StatusClass, c.v)
 	}

-	// nginx_http_response_body_bytes_by_source (histogram, labeled by source_tag)
-	fmt.Fprintln(bw, "# HELP nginx_http_response_body_bytes_by_source HTTP response body size distribution by nginx source tag.")
-	fmt.Fprintln(bw, "# TYPE nginx_http_response_body_bytes_by_source histogram")
-	for _, s := range sourceBodySnaps {
-		writeBodyHistogram(bw, "nginx_http_response_body_bytes_by_source", "source_tag", s.label, s.e)
+	pln("# HELP logtail_udp_packets_received_total Datagrams read from the UDP socket.")
+	pln("# TYPE logtail_udp_packets_received_total counter")
+	pf("logtail_udp_packets_received_total %d\n", udpPackets)
+	pln("# HELP logtail_udp_loglines_success_total UDP loglines that parsed successfully.")
+	pln("# TYPE logtail_udp_loglines_success_total counter")
+	pf("logtail_udp_loglines_success_total %d\n", udpSuccess)
+	pln("# HELP logtail_udp_loglines_consumed_total UDP loglines forwarded to the store (not dropped).")
+	pln("# TYPE logtail_udp_loglines_consumed_total counter")
+	pf("logtail_udp_loglines_consumed_total %d\n", udpConsumed)
+
+	_ = bw.Flush()
 }

-	// UDP ingest counters — lets operators distinguish parse failures
-	// (received - success) from channel-full drops (success - consumed).
-	fmt.Fprintln(bw, "# HELP logtail_udp_packets_received_total Datagrams read from the UDP socket.")
-	fmt.Fprintln(bw, "# TYPE logtail_udp_packets_received_total counter")
-	fmt.Fprintf(bw, "logtail_udp_packets_received_total %d\n", udpPackets)
-	fmt.Fprintln(bw, "# HELP logtail_udp_loglines_success_total UDP loglines that parsed successfully.")
-	fmt.Fprintln(bw, "# TYPE logtail_udp_loglines_success_total counter")
-	fmt.Fprintf(bw, "logtail_udp_loglines_success_total %d\n", udpSuccess)
-	fmt.Fprintln(bw, "# HELP logtail_udp_loglines_consumed_total UDP loglines forwarded to the store (not dropped).")
-	fmt.Fprintln(bw, "# TYPE logtail_udp_loglines_consumed_total counter")
-	fmt.Fprintf(bw, "logtail_udp_loglines_consumed_total %d\n", udpConsumed)
-
-	bw.Flush()
-}
-
-func writeBodyHistogram(bw *bufio.Writer, metric, labelName, labelValue string, e promBodyEntry) {
+func writeBodyHistogramHS(bw *bufio.Writer, metric string, k hostSourceKey, e promBodyEntry) {
+	pf := func(format string, a ...any) { _, _ = fmt.Fprintf(bw, format, a...) }
 	for i, bound := range promBodyBounds {
-		fmt.Fprintf(bw, "%s_bucket{%s=%q,le=%q} %d\n",
-			metric, labelName, labelValue, fmt.Sprintf("%d", bound), e.buckets[i])
+		pf("%s_bucket{host=%q,source_tag=%q,le=\"%d\"} %d\n",
+			metric, k.Host, k.SourceTag, bound, e.buckets[i])
 	}
-	fmt.Fprintf(bw, "%s_bucket{%s=%q,le=\"+Inf\"} %d\n",
-		metric, labelName, labelValue, e.buckets[promNumBodyBounds])
-	fmt.Fprintf(bw, "%s_count{%s=%q} %d\n",
-		metric, labelName, labelValue, e.buckets[promNumBodyBounds])
-	fmt.Fprintf(bw, "%s_sum{%s=%q} %d\n",
-		metric, labelName, labelValue, e.sum)
+	pf("%s_bucket{host=%q,source_tag=%q,le=\"+Inf\"} %d\n",
+		metric, k.Host, k.SourceTag, e.buckets[promNumBodyBounds])
+	pf("%s_count{host=%q,source_tag=%q} %d\n",
+		metric, k.Host, k.SourceTag, e.buckets[promNumBodyBounds])
+	pf("%s_sum{host=%q,source_tag=%q} %d\n",
+		metric, k.Host, k.SourceTag, e.sum)
+}
+
+func writeTimeHistogramHS(bw *bufio.Writer, metric string, k hostSourceKey, e promTimeEntry) {
+	pf := func(format string, a ...any) { _, _ = fmt.Fprintf(bw, format, a...) }
+	for i, bound := range promTimeBounds {
+		pf("%s_bucket{host=%q,source_tag=%q,le=%q} %d\n",
+			metric, k.Host, k.SourceTag, formatFloat(bound), e.buckets[i])
+	}
+	pf("%s_bucket{host=%q,source_tag=%q,le=\"+Inf\"} %d\n",
+		metric, k.Host, k.SourceTag, e.buckets[promNumTimeBounds])
+	pf("%s_count{host=%q,source_tag=%q} %d\n",
+		metric, k.Host, k.SourceTag, e.buckets[promNumTimeBounds])
+	pf("%s_sum{host=%q,source_tag=%q} %g\n",
+		metric, k.Host, k.SourceTag, e.sum)
 }

 // formatFloat renders a float64 bucket bound without trailing zeros but always
@@ -280,7 +397,7 @@ func writeBodyHistogram(bw *bufio.Writer, metric, labelName, labelValue string,
 func formatFloat(f float64) string {
 	s := fmt.Sprintf("%g", f)
 	if !strings.Contains(s, ".") && !strings.Contains(s, "e") {
-		s += ".0" // ensure it looks like a float, not an integer
+		s += ".0"
 	}
 	return s
 }
@@ -6,22 +6,22 @@ import (
 	"testing"
 )

-func TestPromStoreIngestBodyBuckets(t *testing.T) {
+func TestPromStoreIngestBytesBuckets(t *testing.T) {
 	ps := NewPromStore()
 	// 512 bytes: > 256, ≤ 1024 → bucket[0] stays 0, buckets[1..N] get 1
-	ps.Ingest(LogRecord{Website: "example.com", Method: "GET", Status: "200", BodyBytesSent: 512})
+	ps.Ingest(LogRecord{Website: "example.com", Method: "GET", Status: "200", BytesSent: 512, SourceTag: "direct"})

 	ps.mu.Lock()
-	be := ps.body["example.com"]
+	be := ps.bytesSent[hostSourceKey{"example.com", "direct"}]
 	ps.mu.Unlock()

 	if be == nil {
-		t.Fatal("expected body entry, got nil")
+		t.Fatal("expected bytes entry, got nil")
 	}
-	if be.buckets[0] != 0 { // le=256: 512 > 256
+	if be.buckets[0] != 0 {
 		t.Errorf("le=256 bucket = %d, want 0", be.buckets[0])
 	}
-	if be.buckets[1] != 1 { // le=1024: 512 ≤ 1024
+	if be.buckets[1] != 1 {
 		t.Errorf("le=1024 bucket = %d, want 1", be.buckets[1])
 	}
 	for i := 2; i <= promNumBodyBounds; i++ {
@@ -37,24 +37,21 @@ func TestPromStoreIngestBodyBuckets(t *testing.T) {
 func TestPromStoreIngestTimeBuckets(t *testing.T) {
 	ps := NewPromStore()
 	// 0.075s: > 0.05, ≤ 0.1
-	ps.Ingest(LogRecord{Website: "example.com", Method: "GET", Status: "200", RequestTime: 0.075})
+	ps.Ingest(LogRecord{Website: "example.com", Method: "GET", Status: "200", RequestTime: 0.075, SourceTag: "direct"})

 	ps.mu.Lock()
-	te := ps.reqTime["example.com"]
+	te := ps.requestDuration[hostSourceKey{"example.com", "direct"}]
 	ps.mu.Unlock()

 	if te == nil {
-		t.Fatal("expected time entry, got nil")
+		t.Fatal("expected request_duration entry, got nil")
 	}
-	// le=0.05 (index 3): 0.075 > 0.05 → 0
 	if te.buckets[3] != 0 {
 		t.Errorf("le=0.05 bucket = %d, want 0", te.buckets[3])
 	}
-	// le=0.1 (index 4): 0.075 ≤ 0.1 → 1
 	if te.buckets[4] != 1 {
 		t.Errorf("le=0.1 bucket = %d, want 1", te.buckets[4])
 	}
-	// +Inf (last): always 1
 	if te.buckets[promNumTimeBounds] != 1 {
 		t.Errorf("+Inf bucket = %d, want 1", te.buckets[promNumTimeBounds])
 	}
@@ -83,7 +80,7 @@ func TestPromStoreServeHTTP(t *testing.T) {
 	ps := NewPromStore()
 	ps.Ingest(LogRecord{
 		Website: "example.com", Method: "GET", Status: "200",
-		BodyBytesSent: 100, RequestTime: 0.042,
+		BytesSent: 100, RequestTime: 0.042, SourceTag: "direct",
 	})

 	req := httptest.NewRequest("GET", "/metrics", nil)
@@ -95,13 +92,15 @@ func TestPromStoreServeHTTP(t *testing.T) {
 	checks := []string{
 		"# TYPE nginx_http_requests_total counter",
 		`nginx_http_requests_total{host="example.com",method="GET",status="200"} 1`,
-		"# TYPE nginx_http_response_body_bytes histogram",
-		`nginx_http_response_body_bytes_bucket{host="example.com",le="256"} 1`, // 100 ≤ 256
-		`nginx_http_response_body_bytes_count{host="example.com"} 1`,
-		`nginx_http_response_body_bytes_sum{host="example.com"} 100`,
+		"# TYPE nginx_http_bytes_sent histogram",
+		`nginx_http_bytes_sent_bucket{host="example.com",source_tag="direct",le="256"} 1`,
+		`nginx_http_bytes_sent_count{host="example.com",source_tag="direct"} 1`,
+		`nginx_http_bytes_sent_sum{host="example.com",source_tag="direct"} 100`,
 		"# TYPE nginx_http_request_duration_seconds histogram",
-		`nginx_http_request_duration_seconds_bucket{host="example.com",le="0.05"} 1`, // 0.042 ≤ 0.05
-		`nginx_http_request_duration_seconds_count{host="example.com"} 1`,
+		`nginx_http_request_duration_seconds_bucket{host="example.com",source_tag="direct",le="0.05"} 1`,
+		`nginx_http_request_duration_seconds_count{host="example.com",source_tag="direct"} 1`,
+		"# TYPE nginx_http_requests_by_source_total counter",
+		`nginx_http_requests_by_source_total{source_tag="direct",status_class="2xx"} 1`,
 	}
 	for _, want := range checks {
 		if !strings.Contains(body, want) {
@@ -110,12 +109,12 @@ func TestPromStoreServeHTTP(t *testing.T) {
 	}
 }

-func TestPromStoreSourceTagRollup(t *testing.T) {
+func TestPromStoreSourceTagAndStatusClass(t *testing.T) {
 	ps := NewPromStore()
-	// same host, two tags; each tag should appear with its own series.
-	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "200", BodyBytesSent: 100, SourceTag: "direct"})
-	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "200", BodyBytesSent: 300, SourceTag: "cdn"})
-	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "200", BodyBytesSent: 100, SourceTag: "cdn"})
+	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "200", BytesSent: 100, SourceTag: "direct"})
+	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "200", BytesSent: 300, SourceTag: "cdn"})
+	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "404", BytesSent: 100, SourceTag: "cdn"})
+	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "500", BytesSent: 50, SourceTag: "cdn"})

 	req := httptest.NewRequest("GET", "/metrics", nil)
 	rec := httptest.NewRecorder()
@@ -124,13 +123,17 @@ func TestPromStoreSourceTagRollup(t *testing.T) {

 	checks := []string{
 		"# TYPE nginx_http_requests_by_source_total counter",
-		`nginx_http_requests_by_source_total{source_tag="direct"} 1`,
-		`nginx_http_requests_by_source_total{source_tag="cdn"} 2`,
-		"# TYPE nginx_http_response_body_bytes_by_source histogram",
-		`nginx_http_response_body_bytes_by_source_sum{source_tag="direct"} 100`,
-		`nginx_http_response_body_bytes_by_source_sum{source_tag="cdn"} 400`,
-		// host-series totals are unchanged (one row, counting 3 requests).
-		`nginx_http_requests_total{host="h",method="GET",status="200"} 3`,
+		`nginx_http_requests_by_source_total{source_tag="direct",status_class="2xx"} 1`,
+		`nginx_http_requests_by_source_total{source_tag="cdn",status_class="2xx"} 1`,
+		`nginx_http_requests_by_source_total{source_tag="cdn",status_class="4xx"} 1`,
+		`nginx_http_requests_by_source_total{source_tag="cdn",status_class="5xx"} 1`,
+		// dual-labeled histogram subsumes the old by-source bytes metric
+		`nginx_http_bytes_sent_sum{host="h",source_tag="direct"} 100`,
+		`nginx_http_bytes_sent_sum{host="h",source_tag="cdn"} 450`,
+		// host counter unchanged
+		`nginx_http_requests_total{host="h",method="GET",status="200"} 2`,
+		`nginx_http_requests_total{host="h",method="GET",status="404"} 1`,
+		`nginx_http_requests_total{host="h",method="GET",status="500"} 1`,
 	}
 	for _, want := range checks {
 		if !strings.Contains(body, want) {
@@ -139,6 +142,79 @@ func TestPromStoreSourceTagRollup(t *testing.T) {
 	}
 }

+func TestPromStoreUpstreamMetrics(t *testing.T) {
+	ps := NewPromStore()
+	// One upstream-served 200, one upstream-served 502, one no-upstream 200.
+	ps.Ingest(LogRecord{
+		Website: "h", Method: "GET", Status: "200", BytesSent: 100, SourceTag: "cdn",
+		RequestLength:        500,
+		HasUpstream:          true,
+		UpstreamResponseTime: 0.020,
+		UpstreamStatus:       "200",
+	})
+	ps.Ingest(LogRecord{
+		Website: "h", Method: "GET", Status: "502", BytesSent: 100, SourceTag: "cdn",
+		RequestLength:        500,
+		HasUpstream:          true,
+		UpstreamResponseTime: 0.150,
+		UpstreamStatus:       "502",
+	})
+	ps.Ingest(LogRecord{
+		Website: "h", Method: "GET", Status: "200", BytesSent: 100, SourceTag: "direct",
+		RequestLength: 200,
+	})
+
+	req := httptest.NewRequest("GET", "/metrics", nil)
+	rec := httptest.NewRecorder()
+	ps.ServeHTTP(rec, req)
+	body := rec.Body.String()
+
+	checks := []string{
+		// upstream counter: only the two upstream-served requests
+		`nginx_http_upstream_requests_total{host="h",source_tag="cdn",status_class="2xx"} 1`,
+		`nginx_http_upstream_requests_total{host="h",source_tag="cdn",status_class="5xx"} 1`,
+		// upstream duration histogram only for upstream-served requests
+		`nginx_http_upstream_duration_seconds_count{host="h",source_tag="cdn"} 2`,
+		// request_bytes only observed when RequestLength > 0 — all three here
+		`nginx_http_request_bytes_count{host="h",source_tag="cdn"} 2`,
+		`nginx_http_request_bytes_count{host="h",source_tag="direct"} 1`,
+	}
+	for _, want := range checks {
+		if !strings.Contains(body, want) {
+			t.Errorf("missing %q in output:\n%s", want, body)
+		}
+	}
+
+	// no-upstream request must not bump upstream counters
+	if strings.Contains(body, `source_tag="direct",status_class=`) &&
+		strings.Contains(body, "nginx_http_upstream_requests_total") {
+		// Better check: ensure no direct-tagged upstream entry
+		for _, bad := range []string{
+			`nginx_http_upstream_requests_total{host="h",source_tag="direct"`,
+			`nginx_http_upstream_duration_seconds_bucket{host="h",source_tag="direct"`,
+		} {
+			if strings.Contains(body, bad) {
+				t.Errorf("unexpected %q in output:\n%s", bad, body)
+			}
+		}
+	}
+}
+
+func TestPromStoreRequestBytesSkippedWhenZero(t *testing.T) {
+	ps := NewPromStore()
+	// v1 record — RequestLength=0, so request_bytes histogram should be empty.
+	ps.Ingest(LogRecord{Website: "h", Method: "GET", Status: "200", BytesSent: 100, SourceTag: "cdn"})
+
+	req := httptest.NewRequest("GET", "/metrics", nil)
+	rec := httptest.NewRecorder()
+	ps.ServeHTTP(rec, req)
+	body := rec.Body.String()
+
+	if strings.Contains(body, "nginx_http_request_bytes_bucket{") {
+		t.Errorf("expected no request_bytes series for v1 record:\n%s", body)
+	}
+}
+
 func TestPromStoreUDPCounters(t *testing.T) {
 	ps := NewPromStore()
 	ps.IncUDPPacket()
@@ -167,7 +243,6 @@ func TestPromStoreUDPCounters(t *testing.T) {

 func TestPromStoreCounterCap(t *testing.T) {
 	ps := NewPromStore()
-	// Fill to cap with distinct {host,method,status} combos
 	for i := 0; i < promCounterCap+10; i++ {
 		host := strings.Repeat("x", i%10+1) + ".com"
 		status := "200"
@@ -183,3 +258,18 @@ func TestPromStoreCounterCap(t *testing.T) {
 		t.Errorf("counter map size %d exceeds cap %d", n, promCounterCap)
 	}
 }
+
+func TestStatusClass(t *testing.T) {
+	cases := map[string]string{
+		"200": "2xx", "201": "2xx", "299": "2xx",
+		"301": "3xx",
+		"404": "4xx", "418": "4xx",
+		"500": "5xx", "504": "5xx",
+		"100": "other", "": "other", "abc": "other",
+	}
+	for in, want := range cases {
+		if got := statusClass(in); got != want {
+			t.Errorf("statusClass(%q) = %q, want %q", in, got, want)
+		}
+	}
+}
@@ -32,9 +32,18 @@ func BenchmarkIngest(b *testing.B) {
 	}
 }

-// BenchmarkParseLine measures parser throughput.
+// BenchmarkParseLine measures parser throughput on a v1 line.
 func BenchmarkParseLine(b *testing.B) {
-	line := "www.example.com\t1.2.3.4\t1741954800.123\tGET\t/api/v1/search?q=foo\t200\t1452\t0.043"
+	line := "v1\twww.example.com\t1.2.3.4\tGET\t/api/v1/search?q=foo\t200\t1452\t0.043\t0\t12345\tcdn\t10.0.0.1\thttps"
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		ParseLine(line, 24, 48)
+	}
+}
+
+// BenchmarkParseLineV2 measures parser throughput on a v2 line with upstream.
+func BenchmarkParseLineV2(b *testing.B) {
+	line := "v2\twww.example.com\t1.2.3.4\tGET\t/api/v1/search?q=foo\t200\t1500\t421\t0.043\t0.012\t200\t0\t12345\tcdn\t10.0.0.1\thttps"
 	b.ResetTimer()
 	for i := 0; i < b.N; i++ {
 		ParseLine(line, 24, 48)
@@ -118,7 +127,7 @@ func TestGRPCEndToEnd(t *testing.T) {
 	}
 	grpcSrv := grpc.NewServer()
 	pb.RegisterLogtailServiceServer(grpcSrv, NewServer(store, "e2e-test"))
-	go grpcSrv.Serve(lis)
+	go func() { _ = grpcSrv.Serve(lis) }()
 	defer grpcSrv.Stop()

 	// Dial it
@@ -127,7 +136,7 @@ func TestGRPCEndToEnd(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()
 	client := pb.NewLogtailServiceClient(conn)

 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
@@ -52,7 +52,7 @@ func (mt *MultiTailer) Run(ctx context.Context) {
 	if err != nil {
 		log.Fatalf("tailer: failed to create watcher: %v", err)
 	}
-	defer watcher.Close()
+	defer func() { _ = watcher.Close() }()

 	files := make(map[string]*fileState)
 	retrying := make(map[string]struct{}) // paths currently in a retryOpen goroutine
@@ -85,7 +85,7 @@ func (mt *MultiTailer) Run(ctx context.Context) {
 		select {
 		case <-ctx.Done():
 			for _, fs := range files {
-				fs.f.Close()
+				_ = fs.f.Close()
 			}
 			return

@@ -117,7 +117,7 @@ func (mt *MultiTailer) Run(ctx context.Context) {
 			if event.Has(fsnotify.Rename) || event.Has(fsnotify.Remove) {
 				// Drain remaining bytes in the old fd before it disappears.
 				mt.readLines(fs.reader)
-				fs.f.Close()
+				_ = fs.f.Close()
 				delete(files, event.Name)
 				_ = watcher.Remove(event.Name)
 				startRetry(event.Name)
@@ -167,7 +167,7 @@ func (mt *MultiTailer) rescan(
 	for path, fs := range files {
 		if _, matched := current[path]; !matched {
 			mt.readLines(fs.reader)
-			fs.f.Close()
+			_ = fs.f.Close()
 			_ = watcher.Remove(path)
 			delete(files, path)
 			log.Printf("tailer: retired %s (no longer matched by any pattern)", path)
@@ -182,11 +182,11 @@ func openAndSeekEOF(path string, watcher *fsnotify.Watcher) (*fileState, error)
 		return nil, err
 	}
 	if _, err := f.Seek(0, io.SeekEnd); err != nil {
-		f.Close()
+		_ = f.Close()
 		return nil, err
 	}
 	if err := watcher.Add(path); err != nil {
-		f.Close()
+		_ = f.Close()
 		return nil, err
 	}
 	return &fileState{f: f, reader: bufio.NewReader(f)}, nil
@@ -11,7 +11,8 @@ import (

 func writeLine(t *testing.T, f *os.File, website string) {
 	t.Helper()
-	_, err := fmt.Fprintf(f, "%s\t1.2.3.4\t0\tGET\t/path\t200\t0\t0.001\n", website)
+	// v1 layout: 12 payload fields after the v1 prefix.
+	_, err := fmt.Fprintf(f, "v1\t%s\t1.2.3.4\tGET\t/path\t200\t0\t0.001\t0\t0\tdirect\t10.0.0.1\thttps\n", website)
 	if err != nil {
 		t.Fatalf("writeLine: %v", err)
 	}
@@ -25,7 +26,7 @@ func TestMultiTailerReadsLines(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	defer f.Close()
+	defer func() { _ = f.Close() }()

 	ch := make(chan LogRecord, 100)
 	mt := NewMultiTailer([]string{path}, time.Hour, 24, 48, ch)
@@ -62,7 +63,7 @@ func TestMultiTailerMultipleFiles(t *testing.T) {
 		if err != nil {
 			t.Fatal(err)
 		}
-		defer f.Close()
+		defer func() { _ = f.Close() }()
 		files[i] = f
 	}

@@ -108,7 +109,7 @@ func TestMultiTailerLogRotation(t *testing.T) {

 	// Simulate logrotate: rename the old file, create a new one
 	rotated := filepath.Join(dir, "access.log.1")
-	f.Close()
+	_ = f.Close()
 	if err := os.Rename(path, rotated); err != nil {
 		t.Fatal(err)
 	}
@@ -121,7 +122,7 @@ func TestMultiTailerLogRotation(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	defer newF.Close()
+	defer func() { _ = newF.Close() }()

 	// Allow retry goroutine to pick it up
 	time.Sleep(300 * time.Millisecond)
@@ -137,7 +138,7 @@ func TestExpandGlobs(t *testing.T) {
 	dir := t.TempDir()
 	for _, name := range []string{"a.log", "b.log", "other.txt"} {
 		f, _ := os.Create(filepath.Join(dir, name))
-		f.Close()
+		_ = f.Close()
 	}

 	pattern := filepath.Join(dir, "*.log")
@@ -151,7 +152,7 @@ func TestExpandGlobsDeduplication(t *testing.T) {
 	dir := t.TempDir()
 	p := filepath.Join(dir, "access.log")
 	f, _ := os.Create(p)
-	f.Close()
+	_ = f.Close()

 	// Same file listed twice via explicit path and glob
 	paths := expandGlobs([]string{p, filepath.Join(dir, "*.log")})
@@ -16,7 +16,7 @@ const udpReadBufBytes = 4 << 20
 const udpPacketBuf = 64 << 10

 // UDPListener receives nginx_ipng_stats_logtail datagrams on a local socket,
-// parses each packet as one log line, and forwards LogRecords to ch.
+// parses each line through the versioned ParseLine, and forwards LogRecords to ch.
 type UDPListener struct {
 	addr   string
 	v4bits int
@@ -50,7 +50,7 @@ func (u *UDPListener) Run(ctx context.Context) {
 	if err != nil {
 		log.Fatalf("udp: listen %s: %v", u.addr, err)
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()
 	if err := conn.SetReadBuffer(udpReadBufBytes); err != nil {
 		log.Printf("udp: SetReadBuffer(%d): %v", udpReadBufBytes, err)
 	}
@@ -58,7 +58,7 @@ func (u *UDPListener) Run(ctx context.Context) {

 	go func() {
 		<-ctx.Done()
-		conn.Close()
+		_ = conn.Close()
 	}()

 	buf := make([]byte, udpPacketBuf)
@@ -84,7 +84,7 @@ func (u *UDPListener) Run(ctx context.Context) {
 			if line == "" {
 				continue
 			}
-			rec, ok := ParseUDPLine(line, u.v4bits, u.v6bits)
+			rec, ok := ParseLine(line, u.v4bits, u.v6bits)
 			if !ok {
 				continue
 			}
@@ -17,7 +17,7 @@ func TestUDPListenerRoundTrip(t *testing.T) {
 		t.Fatalf("listen probe: %v", err)
 	}
 	addr := pc.LocalAddr().String()
-	pc.Close() // release; listener will re-bind
+	_ = pc.Close() // release; listener will re-bind

 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -31,15 +31,15 @@ func TestUDPListenerRoundTrip(t *testing.T) {
 	if err != nil {
 		t.Fatalf("dial: %v", err)
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()

 	// The listener is started asynchronously; retry for up to 1s.
 	good := "v1\twww.example.com\t1.2.3.4\tGET\t/\t200\t42\t0.010\t0\t12345\tdirect\t10.0.0.1\thttps"
 	bad := "not enough\tfields"
 	deadline := time.Now().Add(time.Second)
 	for time.Now().Before(deadline) {
-		conn.Write([]byte(good))
-		conn.Write([]byte(bad))
+		_, _ = conn.Write([]byte(good))
+		_, _ = conn.Write([]byte(bad))
 		select {
 		case rec := <-ch:
 			if rec.Website != "www.example.com" || rec.SourceTag != "direct" {
@@ -80,7 +80,7 @@ func TestUDPListenerBatchedDatagram(t *testing.T) {
 		t.Fatalf("listen probe: %v", err)
 	}
 	addr := pc.LocalAddr().(*net.UDPAddr)
-	pc.Close()
+	_ = pc.Close()

 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -100,7 +100,7 @@ func TestUDPListenerBatchedDatagram(t *testing.T) {
 	if err != nil {
 		t.Fatalf("src listen: %v", err)
 	}
-	defer src.Close()
+	defer func() { _ = src.Close() }()

 	// Drive the listener with retries until all three records land.
 	got := make(map[string]bool)
@@ -148,7 +148,7 @@ func TestUDPListenerMultipleSources(t *testing.T) {
 		t.Fatalf("listen probe: %v", err)
 	}
 	addr := pc.LocalAddr().(*net.UDPAddr)
-	pc.Close()
+	_ = pc.Close()

 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -166,7 +166,7 @@ func TestUDPListenerMultipleSources(t *testing.T) {
 		if err != nil {
 			t.Fatalf("%s listen: %v", tag, err)
 		}
-		defer src.Close()
+		defer func() { _ = src.Close() }()
 		deadline := time.Now().Add(time.Second)
 		for time.Now().Before(deadline) {
 			if _, err := src.WriteTo(good, addr); err != nil {
@@ -51,7 +51,7 @@ func applyTerm(term string, fs *filterState) error {
 	// Find the first operator character: ~, !, >, <, =
 	opIdx := strings.IndexAny(term, "~!><=")
 	if opIdx <= 0 {
-		return fmt.Errorf("invalid term %q: expected field=value, field>=value, field~=regex, etc.", term)
+		return fmt.Errorf("invalid term %q: expected field=value, field>=value, field~=regex, etc", term)
 	}

 	field := strings.ToLower(strings.TrimSpace(term[:opIdx]))
@@ -52,7 +52,7 @@ func startFake(t *testing.T, fs *fakeServer) string {
 	}
 	srv := grpc.NewServer()
 	pb.RegisterLogtailServiceServer(srv, fs)
-	go srv.Serve(lis)
+	go func() { _ = srv.Serve(lis) }()
 	t.Cleanup(srv.GracefulStop)
 	return lis.Addr().String()
 }
@@ -541,7 +541,7 @@ func TestDialFake(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()

 	_ = client
 	// If we got here without error, the fake server is reachable.
@@ -542,7 +542,7 @@ func (h *Handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 			fmt.Sprintf("cannot connect to %s: %v", params.Target, err)))
 		return
 	}
-	defer conn.Close()
+	defer func() { _ = conn.Close() }()

 	ctx, cancel := context.WithTimeout(r.Context(), 5*time.Second)
 	defer cancel()
@@ -589,7 +589,7 @@ func (h *Handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 		}
 		resp, err := ltClient.ListTargets(ctx, &pb.ListTargetsRequest{})
 		if ltConn != nil {
-			ltConn.Close()
+			_ = ltConn.Close()
 		}
 		if err != nil {
 			ltCh <- nil
@@ -683,5 +683,5 @@ func writeRawJSON(w http.ResponseWriter, params QueryParams, resp *pb.TopNRespon
 		o.Entries[i] = entry{Label: e.Label, Count: e.Count}
 	}
 	w.Header().Set("Content-Type", "application/json")
-	json.NewEncoder(w).Encode(o)
+	_ = json.NewEncoder(w).Encode(o)
 }
@@ -59,7 +59,7 @@ func main() {

 	<-ctx.Done()
 	log.Printf("frontend: shutting down")
-	srv.Shutdown(context.Background())
+	_ = srv.Shutdown(context.Background())
 }

 func envOr(key, def string) string {
@@ -64,8 +64,9 @@ natively, so operators can run it either from on-disk log files, from the UDP fe

 ### Non-Goals

- The system does **not** parse arbitrary nginx `log_format` strings. Two fixed tab-separated formats are supported: a file format and
-  a UDP format (see FR-2). Operators who need general parsing should use Vector, Fluent Bit, or Promtail.
+- The system does **not** parse arbitrary nginx `log_format` strings. A single versioned tab-separated format is
+  supported on both file and UDP ingest (see FR-2). Operators who need general parsing should use Vector, Fluent Bit, or
+  Promtail.
 - The system does **not** store raw log lines. Counts are aggregated at ingest; the original log lines are not kept in memory or on
  disk. The project does not replace an access log.
 - The system does **not** persist counters across restarts. Ring buffers are in-memory only. On aggregator restart, historical state
@@ -98,50 +99,92 @@ Each requirement carries a unique identifier (`FR-X.Y` or `NFR-X.Y`) so that lat
  working set.
 - **FR-1.5** `http_response` MUST be the HTTP status code as recorded by nginx.
 - **FR-1.6** `is_tor` MUST be a boolean, populated by the operator in the log format (typically via a lookup against a TOR exit-node
-  list). For the file format, lines without this field default to `false` for backward compatibility.
- **FR-1.7** `asn` MUST be an int32 decimal value sourced from MaxMind GeoIP2 (or equivalent). For the file format, lines without
-  this field default to `0`.
- **FR-1.8** `ipng_source_tag` MUST be a short string identifying which attribution tag the request arrived under. For records from
-  on-disk log files, the collector MUST assign the tag `"direct"` (mirroring `nginx-ipng-stats-plugin`'s default-source convention). For
-  records from the UDP stream, the tag is taken from the log line as emitted by the plugin.
+  list). Operators without TOR data MUST emit literal `0`.
+- **FR-1.7** `asn` MUST be an int32 decimal value sourced from MaxMind GeoIP2 (or equivalent). Operators without GeoIP data MUST
+  emit literal `0`.
+- **FR-1.8** `ipng_source_tag` MUST be a short string identifying which attribution tag the request arrived under. The tag is
+  always taken verbatim from the log line; the collector does NOT synthesise a fallback. Operators not running
+  `nginx-ipng-stats-plugin` MUST emit a literal value (typically `"direct"`).

 **FR-2 Log formats**

- **FR-2.1 File format.** The collector MUST accept nginx access logs in the following tab-separated layout, with the last two fields
-  (`is_tor`, `asn`) optional for backward compatibility:
+- **FR-2.1 Versioned dispatch.** Both the file tailer and the UDP listener MUST funnel every input line through a single
+  parser that switches on a leading `v<N>\t` version tag. Lines without a recognised tag — including the legacy
+  positional file format — MUST be rejected and counted as parse failures. Two versions are defined: `v1` (FR-2.2) and
+  `v2` (FR-2.3). Both ingest paths accept both versions; downstream processing is identical regardless of which path the
+  line came in over. `$server_addr` and `$scheme` are parsed but discarded — they are reserved for future use.
+
+- **FR-2.2 v1 format.** The v1 payload MUST be exactly 12 tab-separated fields after the `v1` tag (13 fields total).

  ```nginx
-  log_format logtail '$host\t$remote_addr\t$msec\t$request_method\t$request_uri\t$status\t$body_bytes_sent\t$request_time\t$is_tor\t$asn';
+  log_format ipng_stats_logtail
+    'v1\t$host\t$remote_addr\t$request_method\t$request_uri\t$status\t'
+    '$body_bytes_sent\t$request_time\t$is_tor\t$asn\t$ipng_source_tag\t$server_addr\t$scheme';
  ```

  | # | Field             | Ingested into                       |
-  |---|-------------------|----------------------------|
-  | 0 | `$host`           | `website`                  |
-  | 1 | `$remote_addr`    | `client_prefix` (truncated)|
-  | 2 | `$msec`           | (discarded)                |
+  |---|-------------------|-------------------------------------|
+  | 0 | `v1`              | version tag                         |
+  | 1 | `$host`           | `website`                           |
+  | 2 | `$remote_addr`    | `client_prefix` (truncated)         |
  | 3 | `$request_method` | Prom `method` label                 |
-  | 4 | `$request_uri`    | `http_request_uri`         |
+  | 4 | `$request_uri`    | `http_request_uri` (query stripped) |
  | 5 | `$status`         | `http_response`                     |
-  | 6 | `$body_bytes_sent`| Prom body histogram        |
-  | 7 | `$request_time`   | Prom duration histogram    |
-  | 8 | `$is_tor`         | `is_tor` (optional)        |
-  | 9 | `$asn`            | `asn` (optional)           |
+  | 6 | `$body_bytes_sent`| Prom `nginx_http_bytes_sent`        |
+  | 7 | `$request_time`   | Prom `nginx_http_request_duration_seconds` |
+  | 8 | `$is_tor`         | `is_tor`                            |
+  | 9 | `$asn`            | `asn`                               |
+  | 10| `$ipng_source_tag`| `source_tag`                        |
+  | 11| `$server_addr`    | *(parsed and discarded)*            |
+  | 12| `$scheme`         | *(parsed and discarded)*            |

- **FR-2.2 UDP format.** The collector MUST accept datagrams in a versioned tab-separated layout, as emitted by
-  `nginx-ipng-stats-plugin`'s `ipng_stats_logtail` directive. Every datagram MUST begin with a literal version tag
-  (`v<N>\t`) so the collector can route each packet to the appropriate parser. Only `v1` is defined in this revision;
-  unknown versions MUST be counted as parse failures and dropped.
+- **FR-2.3 v2 format.** The v2 payload MUST be exactly 15 tab-separated fields after the `v2` tag (16 fields total).
+  v2 replaces `$body_bytes_sent` with `$bytes_sent` (full wire bytes including headers) and adds four operationally
+  important fields: `$request_length` (request size including headers), `$upstream_response_time`, `$upstream_status`,
+  and the existing v1 fields rearranged for clarity.

  ```nginx
-  log_format ipng_stats_logtail 'v1\t$host\t$remote_addr\t$request_method\t$request_uri\t$status\t$body_bytes_sent\t$request_time\t$is_tor\t$asn\t$ipng_source_tag\t$server_addr\t$scheme';
+  log_format ipng_stats_logtail
+    'v2\t$host\t$remote_addr\t$request_method\t$request_uri\t$status\t'
+    '$bytes_sent\t$request_length\t$request_time\t$upstream_response_time\t$upstream_status\t'
+    '$is_tor\t$asn\t$ipng_source_tag\t$server_addr\t$scheme';
  ```

-  The v1 payload MUST have exactly 12 tab-separated fields after the `v1` tag (13 fields total). `$server_addr` and
-  `$scheme` MUST be parsed but dropped; they are reserved for future use. Malformed datagrams (wrong version, wrong
-  field count, bad IP) MUST be counted (FR-8.5) and silently dropped.
+  | # | Field                     | Ingested into                                |
+  |---|---------------------------|----------------------------------------------|
+  | 0 | `v2`                      | version tag                                  |
+  | 1 | `$host`                   | `website`                                    |
+  | 2 | `$remote_addr`            | `client_prefix` (truncated)                  |
+  | 3 | `$request_method`         | Prom `method` label                          |
+  | 4 | `$request_uri`            | `http_request_uri` (query stripped)          |
+  | 5 | `$status`                 | `http_response`                              |
+  | 6 | `$bytes_sent`             | Prom `nginx_http_bytes_sent`                 |
+  | 7 | `$request_length`         | Prom `nginx_http_request_bytes` (v2-only)    |
+  | 8 | `$request_time`           | Prom `nginx_http_request_duration_seconds`   |
+  | 9 | `$upstream_response_time` | Prom `nginx_http_upstream_duration_seconds` (v2-only) |
+  | 10| `$upstream_status`        | Prom `nginx_http_upstream_requests_total` (v2-only) |
+  | 11| `$is_tor`                 | `is_tor`                                     |
+  | 12| `$asn`                    | `asn`                                        |
+  | 13| `$ipng_source_tag`        | `source_tag`                                 |
+  | 14| `$server_addr`            | *(parsed and discarded)*                     |
+  | 15| `$scheme`                 | *(parsed and discarded)*                     |

- **FR-2.3** The file tailer MUST set `source_tag="direct"` on every record it parses. The UDP listener MUST propagate
-  `$ipng_source_tag` verbatim. This is the only difference in downstream processing between the two ingest paths.
+  When nginx serves the response without an upstream (static files, redirects, errors), nginx emits literal `-` for
+  `$upstream_response_time` and `$upstream_status`. The parser MUST treat that as "no upstream", skip the upstream
+  histograms, and not increment the upstream counter. When nginx retries across multiple upstreams, both fields are
+  comma-separated; the parser MUST keep the last entry, since that is the upstream that ultimately served the response.
+
+- **FR-2.4 Semantic shift on v2 rollout.** v1 fills `nginx_http_bytes_sent` from `$body_bytes_sent`; v2 fills it from
+  `$bytes_sent`. Operators MUST expect a small step up in the metric when emitters move from v1 to v2 (header overhead;
+  typically a few hundred bytes per response).
+
+- **FR-2.5 Malformed input.** Lines with an unknown version, the wrong field count for the claimed version, or an
+  unparsable IP MUST be silently dropped. UDP drops MUST be counted via FR-8.6; file-path drops are implicit (the tailer
+  falls behind the file).
+
+- **FR-2.6 Unknown `$is_tor` / `$asn`.** Operators without TOR or GeoIP data MUST emit literal `0` for both fields. A
+  literal `0` in `$is_tor` parses as `false`; a literal `0` in `$asn` parses as ASN `0`, filterable at query time with
+  `--asn '!=0'`.

 **FR-3 Ring buffers and time windows**

@@ -242,15 +285,21 @@ Each requirement carries a unique identifier (`FR-X.Y` or `NFR-X.Y`) so that lat
 - **FR-8.2** The collector MUST expose a per-request counter `nginx_http_requests_total{host, method, status}` capped at
  `promCounterCap = 250 000` distinct label sets. When the cap is reached, further new label sets MUST be dropped (existing series
  keep incrementing) until the map is rolled over.
- **FR-8.3** The collector MUST expose per-host histograms
-  `nginx_http_response_body_bytes{host, le}` (body-size distribution) and
-  `nginx_http_request_duration_seconds{host, le}` (request-time distribution). The duration histogram MUST NOT be split by
-  `source_tag` — its bucket count would multiply without operational benefit.
- **FR-8.4** The collector MUST expose two parallel roll-ups labeled by `source_tag` only (not cross-producted with host):
-  `nginx_http_requests_by_source_total{source_tag}` and
-  `nginx_http_response_body_bytes_by_source{source_tag, le}`. These are separate metric names to avoid inconsistent label sets
-  under a single name.
- **FR-8.5** The collector MUST expose three counters that let operators distinguish UDP parse failures from back-pressure drops:
+- **FR-8.3** The collector MUST expose two histograms keyed by `{host, source_tag}`:
+  `nginx_http_bytes_sent{host, source_tag, le}` (response wire-bytes distribution; FR-2.4) and
+  `nginx_http_request_duration_seconds{host, source_tag, le}` (end-to-end request time distribution).
+  Cardinality is bounded by `host × source_tag × bucket_count`, which is small enough that no explicit cap is required.
+- **FR-8.4** The collector MUST expose three v2-only metrics that are populated only when v2 records arrive (and, for the
+  upstream metrics, only when nginx involved an upstream):
+  `nginx_http_request_bytes{host, source_tag, le}` from `$request_length`,
+  `nginx_http_upstream_duration_seconds{host, source_tag, le}` from `$upstream_response_time`, and
+  `nginx_http_upstream_requests_total{host, source_tag, status_class}` from `$upstream_status`. `status_class` is the
+  HTTP class of the upstream's status code, folded to `2xx`/`3xx`/`4xx`/`5xx`/`other`.
+- **FR-8.5** The collector MUST expose a source-tag rollup counter
+  `nginx_http_requests_by_source_total{source_tag, status_class}`. `status_class` is the HTTP class of `$status`, folded
+  the same way as in FR-8.4. This rollup is intentionally not cross-producted with `host` — its purpose is fleet-wide
+  source-attribution health, not per-host detail.
+- **FR-8.6** The collector MUST expose three counters that let operators distinguish UDP parse failures from back-pressure drops:
  `logtail_udp_packets_received_total` (datagrams off the socket, one increment per `recvfrom`),
  `logtail_udp_loglines_success_total` (log lines that parsed OK, incremented once per log line — a single batched datagram from
  the nginx plugin may contribute many), and
@@ -279,13 +328,13 @@ Each requirement carries a unique identifier (`FR-X.Y` or `NFR-X.Y`) so that lat
  for a collector is therefore approximately 845 MB (live map ~19 MB + fine ring ~558 MB + coarse ring ~268 MB).
 - **NFR-2.3** The aggregator MUST apply the same tier caps as the collector. Its steady-state memory is roughly equivalent to one
  collector regardless of the number of collectors subscribed.
- **NFR-2.4** The Prometheus counter map (FR-8.2) MUST be capped at `promCounterCap = 250 000` entries. The per-host and per-source
-  histograms MUST NOT be capped explicitly — they grow only with the distinct host count, which is bounded by the operator's vhost
-  configuration.
+- **NFR-2.4** The Prometheus counter map (FR-8.2) MUST be capped at `promCounterCap = 250 000` entries. The dual-labeled
+  `{host, source_tag}` histograms MUST NOT be capped explicitly — they grow only with the cross-product of distinct
+  hosts and distinct source tags, both bounded by the operator's nginx configuration.

 **NFR-3 Performance**

- **NFR-3.1** `ParseLine` and `ParseUDPLine` MUST use `strings.Split` / `strings.SplitN` (no regex), so that per-line cost stays
+- **NFR-3.1** `ParseLine` MUST use `strings.Split` / `strings.IndexByte` (no regex), so that per-line cost stays
  around 50 ns on commodity hardware.
 - **NFR-3.2** `TopN` and `Trend` queries across the full 24-hour coarse ring MUST complete in well under 250 ms at the 50 000-entry
  fine cap, for fully-specified filters.
@@ -417,8 +466,10 @@ connected.

 #### Key data types

- `LogRecord` — ten fields (website, client_prefix, URI, status, is_tor, asn, method, body_bytes_sent, request_time, source_tag).
-  Produced by `ParseLine` or `ParseUDPLine` and consumed by the store goroutine.
+- `LogRecord` — fourteen fields (website, client_prefix, URI, status, is_tor, asn, method, bytes_sent, request_length,
+  request_time, upstream_response_time, upstream_status, has_upstream, source_tag). Produced by `ParseLine` (which
+  dispatches on the `v<N>\t` prefix) and consumed by the store goroutine. v1 records leave the v2-only fields
+  (`request_length`, upstream_*) at zero / false.
 - `Tuple6` (historical name; carries seven fields now) — the aggregation key. NUL-separated when encoded as a map key for snapshots.
  The code name is intentionally stable so downstream tests and consumers are not churned.
 - `Snapshot` — `(timestamp, []Entry)` where `Entry = (label, count)` and `label` is an encoded `Tuple6`.
@@ -559,9 +610,10 @@ transitions. No per-request logging.
 - **Frontend crash.** Stateless. Operator restarts.
 - **UDP datagram loss.** Any datagram dropped in-kernel (socket buffer full, network drop) does not register as a parse failure; it
  is simply invisible. Operators should size `SO_RCVBUF` appropriately; the collector already requests 4 MiB.
- **Malformed log lines.** File format: lines with <8 tab-separated fields are silently skipped; an invalid IP also drops the line.
-  UDP: packets without a recognised `v<N>\t` prefix, or with the wrong field count for the claimed version, or with a bad IP, are
-  counted as received-but-not-success and dropped.
+- **Malformed log lines.** Both ingest paths use the versioned `v<N>\t` parser (FR-2). Lines without a recognised version
+  prefix, with the wrong field count for the claimed version, or with a bad IP are silently dropped. UDP drops are
+  visible as `packets_received_total - loglines_success_total`; file-path drops are implicit (the tailer simply moves
+  past them).
 - **Clock skew between collectors.** Trend sparklines derived from merged data assume collectors are roughly NTP-synced. Per-bucket
  alignment is to the local minute / 5-minute boundary of each collector.
 - **gRPC traffic over untrusted links.** The system does not ship TLS; operators should front the gRPC ports with a TLS-terminating
@@ -588,11 +640,11 @@ transitions. No per-request logging.
 - **pull-based collector polling (aggregator polls collectors every second).** Rejected in favor of push. Polling multiplies query
  latency and makes the aggregator's cache stale by the poll interval. Push-stream with delta merge keeps the cache within seconds
  of real time.
- **One metric name for both per-host and per-source_tag roll-ups.** Rejected for Prometheus hygiene. Mixing different label sets
-  under one metric name breaks aggregation rules; separate metric names (`_by_source`) are clearer and easier to query.
- **Cross-product of `host × source_tag` for every counter and histogram.** Rejected. With ~20 tags and ~50 hosts the cardinality
-  explodes quickly on the duration histogram without operational benefit. The duration histogram stays per-host; requests and body
-  size get a parallel `_by_source` rollup.
+- **Separate `_by_source` metric names with a single label.** The original v0.2 layout exposed `_by_source` siblings to
+  avoid mixing label sets under one metric name. Superseded by the v0.3 layout: histograms now carry both `host` and
+  `source_tag` directly, and the source-tag rollup counter gains a `status_class` label. Cardinality stays bounded
+  (~7 hosts × ~6 tags × 11 buckets ≈ 460 series per histogram), and Grafana queries become simpler (`sum by(source_tag)`
+  rather than picking a different metric name).
 - **Writing every `snapshot` to disk for restart recovery.** Rejected in favor of `DumpSnapshots` RPC backfill. Disk-backed
  persistence would multiply operational surface (rotation, fsck, permissions) for a feature that needs to survive only an
  aggregator restart.
@@ -131,101 +131,98 @@ or for temporary overrides, without editing the unit.
 The file is **not a dpkg conffile**: postinst writes it only when absent, so operator edits
 survive upgrades, and `dpkg --purge` removes it.

-### nginx — file-based ingest
+### nginx — log format

-Add the `logtail` format and attach it to whichever `server` blocks you want tracked:
+Both ingest paths (file and UDP) use the same versioned tab-separated format. Every line MUST
+begin with a literal `v1\t` or `v2\t` prefix; lines without a recognised prefix are dropped.
+Two versions are defined; you can mix them across a fleet during a rollout (the collector
+parses both).

-```nginx
-http {
-    log_format logtail '$host\t$remote_addr\t$msec\t$request_method\t$request_uri\t$status\t$body_bytes_sent\t$request_time\t$is_tor\t$asn';
+#### v2 (recommended)

-    server {
-        access_log /var/log/nginx/access.log logtail;
-        # or per-vhost:
-        access_log /var/log/nginx/www.example.com.access.log logtail;
-    }
-}
-```
-
-Tab-separated, fixed field order, ten fields. The precise layout:
-
-| # | Field             | Ingested into            |
-|---|-------------------|--------------------------|
-| 0 | `$host`           | `website`                |
-| 1 | `$remote_addr`    | `client_prefix` (truncated) |
-| 2 | `$msec`           | *(discarded)*            |
-| 3 | `$request_method` | Prom `method` label      |
-| 4 | `$request_uri`    | `http_request_uri` (query stripped) |
-| 5 | `$status`         | `http_response`          |
-| 6 | `$body_bytes_sent`| Prom body histogram      |
-| 7 | `$request_time`   | Prom duration histogram  |
-| 8 | `$is_tor`         | `is_tor` (optional)      |
-| 9 | `$asn`            | `asn` (optional)         |
-
-`$is_tor` is `1` if the client IP is a TOR exit node and `0` otherwise (typically populated
-via a Lua script or `$geoip2_data_*`). `$asn` is the client AS number as a decimal integer
-(e.g. MaxMind GeoIP2's `$geoip2_data_autonomous_system_number`).
-
-**If either is unknown, emit `0`.** A literal `0` in `$is_tor` parses as `false`; a literal
-`0` in `$asn` parses as ASN `0`, which you can exclude at query time with `--asn '!=0'` / the
-`asn!=0` filter expression. Operators who don't have TOR or GeoIP data can simply emit `0` for
-both columns and everything works.
-
-Both fields are also **positionally optional** for backward compatibility — older 8-field
-lines are accepted and default to `false` / `0`. Records from the file tailer are always
-tagged `source_tag="direct"`.
-
-Then point the collector at the log files via `COLLECTOR_LOGS` — comma-separated paths or
-glob patterns. Make sure the files are group-readable by `www-data` (the collector's primary
-group in the systemd unit).
-
-### nginx — UDP ingest (`nginx-ipng-stats-plugin`)
-
-If the nginx host runs [`nginx-ipng-stats-plugin`](https://git.ipng.ch/ipng/nginx-ipng-stats-plugin),
-the plugin's `ipng_stats_logtail` directive emits one UDP datagram per request directly to
-the collector, no log file involved. The wire format is **versioned** — every datagram starts
-with a literal `v1\t` prefix so the collector can ship new parser versions (v2, v3, …) before
-emitters are upgraded and route each packet accordingly.
+v2 carries five operationally important fields v1 lacks: `$bytes_sent` (full wire bytes,
+replaces `$body_bytes_sent`), `$request_length` (request size including headers),
+`$upstream_response_time`, and `$upstream_status`. Together they let dashboards split
+end-to-end latency into upstream vs. nginx overhead, attribute errors to the upstream vs. the
+edge, and report ingress bandwidth.

 ```nginx
 http {
    log_format ipng_stats_logtail
-        'v1\t$host\t$remote_addr\t$request_method\t$request_uri\t$status\t$body_bytes_sent\t$request_time\t$is_tor\t$asn\t$ipng_source_tag\t$server_addr\t$scheme';
+        'v2\t$host\t$remote_addr\t$request_method\t$request_uri\t$status\t'
+        '$bytes_sent\t$request_length\t$request_time\t$upstream_response_time\t$upstream_status\t'
+        '$is_tor\t$asn\t$ipng_source_tag\t$server_addr\t$scheme';

+    # File ingest:
+    server {
+        access_log /var/log/nginx/access.log ipng_stats_logtail;
+    }
+    # UDP ingest (nginx-ipng-stats-plugin):
    ipng_stats_logtail ipng_stats_logtail udp://127.0.0.1:9514 buffer=64k flush=1s;
 }
 ```

-Precise v1 layout — 13 tab-separated fields total (version prefix + 12 payload fields):
-
 | # | Field                     | Ingested into                                     |
-|---|-------------------|------------------------------|
-| 0 | `v1`              | version tag                  |
+|---|---------------------------|---------------------------------------------------|
+| 0 | `v2`                      | version tag                                       |
 | 1 | `$host`                   | `website`                                         |
 | 2 | `$remote_addr`            | `client_prefix` (truncated)                       |
 | 3 | `$request_method`         | Prom `method` label                               |
 | 4 | `$request_uri`            | `http_request_uri` (query stripped)               |
 | 5 | `$status`                 | `http_response`                                   |
-| 6 | `$body_bytes_sent`| Prom body histogram          |
-| 7 | `$request_time`   | Prom duration histogram      |
-| 8 | `$is_tor`         | `is_tor`                     |
-| 9 | `$asn`            | `asn`                        |
-| 10| `$ipng_source_tag`| `source_tag`                 |
-| 11| `$server_addr`    | *(parsed and discarded)*     |
-| 12| `$scheme`         | *(parsed and discarded)*     |
+| 6 | `$bytes_sent`             | Prom `nginx_http_bytes_sent`                      |
+| 7 | `$request_length`         | Prom `nginx_http_request_bytes`                   |
+| 8 | `$request_time`           | Prom `nginx_http_request_duration_seconds`        |
+| 9 | `$upstream_response_time` | Prom `nginx_http_upstream_duration_seconds`       |
+| 10| `$upstream_status`        | Prom `nginx_http_upstream_requests_total`         |
+| 11| `$is_tor`                 | `is_tor`                                          |
+| 12| `$asn`                    | `asn`                                             |
+| 13| `$ipng_source_tag`        | `source_tag`                                      |
+| 14| `$server_addr`            | *(parsed and discarded)*                          |
+| 15| `$scheme`                 | *(parsed and discarded)*                          |

-Compared to the file format: the version tag is added, `$msec` is dropped, and three fields
-are appended — `$ipng_source_tag` (propagated into the data model), `$server_addr` and
-`$scheme` (reserved for future use).
+For requests served without an upstream (static files, redirects, errors), nginx emits
+literal `-` for `$upstream_response_time` and `$upstream_status`; the parser treats those as
+"no upstream" and skips the upstream metrics rather than counting them as zeros. When nginx
+retries across multiple upstreams, both fields are comma-separated and the parser keeps the
+last value (the upstream that ultimately served the response).

-**Unknown `$is_tor` / `$asn`: emit `0`.** Same convention as the file format — operators
-without TOR or GeoIP data can emit `0` for both columns and everything works. A literal `0`
-in `$is_tor` is `false`; a literal `0` in `$asn` is ASN `0`, filterable at query time.
+#### v1 (legacy)

-All 13 fields are required for v1 — malformed packets (wrong version, wrong field count, bad
-IP) are silently dropped and counted via `logtail_udp_packets_received_total` minus
-`logtail_udp_loglines_success_total`. Both paths (file + UDP) can feed the same collector
-simultaneously; they converge on the same aggregation pipeline.
+v1 is preserved unchanged so existing emitters can be upgraded after the collector. Layout:
+
+```nginx
+log_format ipng_stats_logtail
+    'v1\t$host\t$remote_addr\t$request_method\t$request_uri\t$status\t'
+    '$body_bytes_sent\t$request_time\t$is_tor\t$asn\t$ipng_source_tag\t$server_addr\t$scheme';
+```
+
+12 tab-separated payload fields after the `v1` prefix. v1 fills `nginx_http_bytes_sent` from
+`$body_bytes_sent`; v2 fills it from `$bytes_sent`. Operators will see a small step up in
+that metric (header overhead, typically a few hundred bytes per response) when emitters move
+to v2.
+
+#### Required values
+
+`$is_tor` is `1` if the client IP is a TOR exit node and `0` otherwise (typically populated
+via a Lua script or `$geoip2_data_*`). `$asn` is the client AS number as a decimal integer
+(e.g. MaxMind GeoIP2's `$geoip2_data_autonomous_system_number`). Operators without TOR or
+GeoIP data MUST emit literal `0` for both — a literal `0` in `$is_tor` parses as `false`; a
+literal `0` in `$asn` is ASN `0`, filterable at query time with `--asn '!=0'`.
+
+`$ipng_source_tag` is provided by [`nginx-ipng-stats-plugin`](https://git.ipng.ch/ipng/nginx-ipng-stats-plugin).
+Operators not running the plugin SHOULD declare a constant via `set $ipng_source_tag direct;`
+in their `server` block — there is no synthesised fallback in the collector.
+
+#### Pointing the collector at logs
+
+For file ingest, set `COLLECTOR_LOGS` to comma-separated paths or glob patterns. Make sure
+the files are group-readable by `www-data` (the collector's primary group in the systemd
+unit). For UDP ingest, the plugin's `ipng_stats_logtail udp://127.0.0.1:9514` line above is
+sufficient. Both paths can feed the same collector simultaneously and converge on the same
+aggregation pipeline. Malformed lines (wrong version, wrong field count, bad IP) are silently
+dropped; for UDP they show up as `logtail_udp_packets_received_total` minus
+`logtail_udp_loglines_success_total`.

 ---

@@ -303,21 +300,33 @@ the new file appears. No restart or SIGHUP required.
 The collector exposes a Prometheus-compatible `/metrics` endpoint on `--prom-listen` (default
 `:9100`). Set `--prom-listen ""` to disable it entirely.

-**Per-host series:**
+**Per-{host,source_tag} series** (both v1 and v2):

 - `nginx_http_requests_total{host, method, status}` — counter. Map capped at 250 000 distinct
  label sets; new entries beyond the cap are dropped until the map is rolled over.
- `nginx_http_response_body_bytes_{bucket,count,sum}{host, le}` — histogram of
-  `$body_bytes_sent`. Buckets (bytes): `256, 1024, 4096, 16384, 65536, 262144, 1048576, +Inf`.
- `nginx_http_request_duration_seconds_{bucket,count,sum}{host, le}` — histogram of
-  `$request_time`. Buckets (seconds): `0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5,
-  10, +Inf`. Not split by `source_tag` (duration histogram stays per-host to avoid cardinality
-  blow-up).
+- `nginx_http_bytes_sent_{bucket,count,sum}{host, source_tag, le}` — histogram of response
+  size. v1 fills from `$body_bytes_sent`; v2 fills from `$bytes_sent`. Buckets (bytes):
+  `256, 1024, 4096, 16384, 65536, 262144, 1048576, +Inf`.
+- `nginx_http_request_duration_seconds_{bucket,count,sum}{host, source_tag, le}` — histogram
+  of `$request_time`. Buckets (seconds): `0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5,
+  5, 10, +Inf`.

-**Per-`source_tag` roll-ups** (parallel series, not a cross-product with `host`):
+**v2-only series** (populated only when v2 emitters are running, and the upstream histograms
+only when nginx involved an upstream):

- `nginx_http_requests_by_source_total{source_tag}` — counter.
- `nginx_http_response_body_bytes_by_source_{bucket,count,sum}{source_tag, le}` — histogram.
+- `nginx_http_request_bytes_{bucket,count,sum}{host, source_tag, le}` — histogram of
+  `$request_length` (ingress, headers + body). Same byte buckets as `bytes_sent`.
+- `nginx_http_upstream_duration_seconds_{bucket,count,sum}{host, source_tag, le}` —
+  histogram of `$upstream_response_time`. Same time buckets as `request_duration`. Lets you
+  split end-to-end latency into upstream vs. nginx overhead.
+- `nginx_http_upstream_requests_total{host, source_tag, status_class}` — counter incremented
+  once per upstream-served request, classed by `$upstream_status` (`2xx`/`3xx`/`4xx`/`5xx`/`other`).
+  Lets you spot upstream errors masked at the edge (e.g. nginx 502 because origin 504).
+
+**Source-tag rollup** (fleet-wide attribution health, intentionally not crossed with host):
+
+- `nginx_http_requests_by_source_total{source_tag, status_class}` — counter classed by
+  `$status`. Use it to spot per-source error spikes without exploding cardinality.

 **UDP ingest counters** — lets operators distinguish parse failures from back-pressure drops:

@@ -365,10 +374,23 @@ histogram_quantile(0.95,
  sum by (host, le) (rate(nginx_http_request_duration_seconds_bucket[5m]))
 )

-# Median response body size per host
-histogram_quantile(0.50,
-  sum by (host, le) (rate(nginx_http_response_body_bytes_bucket[5m]))
+# 95th percentile response time per source_tag (drill in further as needed)
+histogram_quantile(0.95,
+  sum by (source_tag, le) (rate(nginx_http_request_duration_seconds_bucket[5m]))
 )
+
+# Median response size per host
+histogram_quantile(0.50,
+  sum by (host, le) (rate(nginx_http_bytes_sent_bucket[5m]))
+)
+
+# v2-only: upstream P95, split out from nginx overhead
+histogram_quantile(0.95,
+  sum by (host, le) (rate(nginx_http_upstream_duration_seconds_bucket[5m]))
+)
+
+# v2-only: upstream 5xx rate per source_tag
+sum by (source_tag) (rate(nginx_http_upstream_requests_total{status_class="5xx"}[5m]))
 ```

 ### Memory usage