collector/netdev_linux.go: Use netlink to get stats

Instead of parsing `/proc/net/dev` to get network interface statistics, get
them from a netlink call.

Internally, both come from the [rtnl_link_stats64] struct, but with
`/proc/net/dev`, some of the values are aggregated together in
[dev_seq_printf_stats], so we get less information out of them.

This commit maintains compatibility by aggregating those stats back into the
same metrics.

[rtnl_link_stats64]:    https://github.com/torvalds/linux/blob/master/include/uapi/linux/if_link.h#L42-L246
[dev_seq_printf_stats]: https://github.com/torvalds/linux/blob/master/net/core/net-procfs.c#L75-L97

Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>

collector/netdev_linux.go

@@ -17,86 +17,60 @@
 package collector
 
 import (
-	"bufio"
-	"fmt"
-	"io"
-	"os"
-	"regexp"
-	"strconv"
-	"strings"
-
 	"github.com/go-kit/log"
 	"github.com/go-kit/log/level"
-)
 
-var (
+	"github.com/jsimonetti/rtnetlink"
-	procNetDevInterfaceRE = regexp.MustCompile(`^(.+): *(.+)$`)
-	procNetDevFieldSep    = regexp.MustCompile(` +`)
 )
 
 func getNetDevStats(filter *deviceFilter, logger log.Logger) (netDevStats, error) {
-	file, err := os.Open(procFilePath("net/dev"))
+	conn, err := rtnetlink.Dial(nil)
 	if err != nil {
 		return nil, err
 	}
-	defer file.Close()
+	defer conn.Close()
 
-	return parseNetDevStats(file, filter, logger)
+	links, err := conn.Link.List()
+	if err != nil {
+		return nil, err
 	}
 
-func parseNetDevStats(r io.Reader, filter *deviceFilter, logger log.Logger) (netDevStats, error) {
+	return netlinkStats(links, filter, logger), nil
-	scanner := bufio.NewScanner(r)
-	scanner.Scan() // skip first header
-	scanner.Scan()
-	parts := strings.Split(scanner.Text(), "|")
-	if len(parts) != 3 { // interface + receive + transmit
-		return nil, fmt.Errorf("invalid header line in net/dev: %s",
-			scanner.Text())
 }
 
-	receiveHeader := strings.Fields(parts[1])
+func netlinkStats(links []rtnetlink.LinkMessage, filter *deviceFilter, logger log.Logger) netDevStats {
-	transmitHeader := strings.Fields(parts[2])
+	metrics := netDevStats{}
-	headerLength := len(receiveHeader) + len(transmitHeader)
 
-	netDev := netDevStats{}
+	for _, msg := range links {
-	for scanner.Scan() {
+		name := msg.Attributes.Name
-		line := strings.TrimLeft(scanner.Text(), " ")
+		stats := msg.Attributes.Stats64
-		parts := procNetDevInterfaceRE.FindStringSubmatch(line)
-		if len(parts) != 3 {
-			return nil, fmt.Errorf("couldn't get interface name, invalid line in net/dev: %q", line)
-		}
 
-		dev := parts[1]
+		if filter.ignored(name) {
-		if filter.ignored(dev) {
+			level.Debug(logger).Log("msg", "Ignoring device", "device", name)
-			level.Debug(logger).Log("msg", "Ignoring device", "device", dev)
 			continue
 		}
 
-		values := procNetDevFieldSep.Split(strings.TrimLeft(parts[2], " "), -1)
+		// https://github.com/torvalds/linux/blob/master/include/uapi/linux/if_link.h#L42-L246
-		if len(values) != headerLength {
+		// https://github.com/torvalds/linux/blob/master/net/core/net-procfs.c#L75-L97
-			return nil, fmt.Errorf("couldn't get values, invalid line in net/dev: %q", parts[2])
+		metrics[name] = map[string]uint64{
+			"receive_packets":     stats.RXPackets,
+			"transmit_packets":    stats.TXPackets,
+			"receive_bytes":       stats.RXBytes,
+			"transmit_bytes":      stats.TXBytes,
+			"receive_errs":        stats.RXErrors,
+			"transmit_errs":       stats.TXErrors,
+			"receive_drop":        stats.RXDropped + stats.RXMissedErrors,
+			"transmit_drop":       stats.TXDropped,
+			"receive_multicast":   stats.Multicast,
+			"transmit_colls":      stats.Collisions,
+			"receive_frame":       stats.RXLengthErrors + stats.RXOverErrors + stats.RXCRCErrors + stats.RXFrameErrors,
+			"receive_fifo":        stats.RXFIFOErrors,
+			"transmit_carrier":    stats.TXAbortedErrors + stats.TXCarrierErrors + stats.TXHeartbeatErrors + stats.TXWindowErrors,
+			"transmit_fifo":       stats.TXFIFOErrors,
+			"receive_compressed":  stats.RXCompressed,
+			"transmit_compressed": stats.TXCompressed,
+		}
 	}
 
-		devStats := map[string]uint64{}
+	return metrics
-		addStats := func(key, value string) {
-			v, err := strconv.ParseUint(value, 0, 64)
-			if err != nil {
-				level.Debug(logger).Log("msg", "invalid value in netstats", "key", key, "value", value, "err", err)
-				return
-			}
-
-			devStats[key] = v
-		}
-
-		for i := 0; i < len(receiveHeader); i++ {
-			addStats("receive_"+receiveHeader[i], values[i])
-		}
-
-		for i := 0; i < len(transmitHeader); i++ {
-			addStats("transmit_"+transmitHeader[i], values[i+len(receiveHeader)])
-		}
-
-		netDev[dev] = devStats
-	}
-	return netDev, scanner.Err()
 }

collector/netdev_linux_test.go

@@ -14,25 +14,125 @@
 package collector
 
 import (
-	"os"
 	"testing"
 
 	"github.com/go-kit/log"
+
+	"github.com/jsimonetti/rtnetlink"
 )
 
-func TestNetDevStatsIgnore(t *testing.T) {
+var links = []rtnetlink.LinkMessage{
-	file, err := os.Open("fixtures/proc/net/dev")
+	{
-	if err != nil {
+		Attributes: &rtnetlink.LinkAttributes{
-		t.Fatal(err)
+			Name: "tun0",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 24,
+				TXPackets: 934,
+				RXBytes:   1888,
+				TXBytes:   67120,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "veth4B09XN",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 8,
+				TXPackets: 10640,
+				RXBytes:   648,
+				TXBytes:   1943284,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "lo",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 1832522,
+				TXPackets: 1832522,
+				RXBytes:   435303245,
+				TXBytes:   435303245,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "eth0",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 520993275,
+				TXPackets: 43451486,
+				RXBytes:   68210035552,
+				TXBytes:   9315587528,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "lxcbr0",
+			Stats64: &rtnetlink.LinkStats64{
+				TXPackets: 28339,
+				TXBytes:   2630299,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "wlan0",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 13899359,
+				TXPackets: 11726200,
+				RXBytes:   10437182923,
+				TXBytes:   2851649360,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "docker0",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 1065585,
+				TXPackets: 1929779,
+				RXBytes:   64910168,
+				TXBytes:   2681662018,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name:    "ibr10:30",
+			Stats64: &rtnetlink.LinkStats64{},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "flannel.1",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 228499337,
+				TXPackets: 258369223,
+				RXBytes:   18144009813,
+				TXBytes:   20758990068,
+				TXDropped: 64,
+			},
+		},
+	},
+	{
+		Attributes: &rtnetlink.LinkAttributes{
+			Name: "💩0",
+			Stats64: &rtnetlink.LinkStats64{
+				RXPackets: 105557,
+				TXPackets: 304261,
+				RXBytes:   57750104,
+				TXBytes:   404570255,
+				Multicast: 72,
+			},
+		},
+	},
 }
-	defer file.Close()
 
+func TestNetDevStatsIgnore(t *testing.T) {
 	filter := newDeviceFilter("^veth", "")
 
-	netStats, err := parseNetDevStats(file, &filter, log.NewNopLogger())
+	netStats := netlinkStats(links, &filter, log.NewNopLogger())
-	if err != nil {
-		t.Fatal(err)
-	}
 
 	if want, got := uint64(10437182923), netStats["wlan0"]["receive_bytes"]; want != got {
 		t.Errorf("want netstat wlan0 bytes %v, got %v", want, got)
@@ -64,17 +164,8 @@ func TestNetDevStatsIgnore(t *testing.T) {
 }
 
 func TestNetDevStatsAccept(t *testing.T) {
-	file, err := os.Open("fixtures/proc/net/dev")
-	if err != nil {
-		t.Fatal(err)
-	}
-	defer file.Close()
-
 	filter := newDeviceFilter("", "^💩0$")
-	netStats, err := parseNetDevStats(file, &filter, log.NewNopLogger())
+	netStats := netlinkStats(links, &filter, log.NewNopLogger())
-	if err != nil {
-		t.Fatal(err)
-	}
 
 	if want, got := 1, len(netStats); want != got {
 		t.Errorf("want count of devices to be %d, got %d", want, got)