ethtool: Expose node_ethtool_info metric

Add a `node_ethtool_info` metric to all ethtool devices to expose driver
information with following labels:

 * bus_info
 * driver
 * expansion_rom_version
 * firmware_version
 * version

This metric is useful to monitor the firmware version to be up-to-date.

Note: The version label might be malformed due to bug #39 in ethtool:
https://github.com/safchain/ethtool/issues/39

Signed-off-by: Benjamin Drung <benjamin.drung@ionos.com>
This commit is contained in:
Benjamin Drung 2021-08-16 15:33:25 +02:00 committed by Johannes 'fish' Ziemke
parent 6ac6ea2d13
commit 26ca609183
6 changed files with 99 additions and 11 deletions

View File

@ -192,7 +192,7 @@ Name | Description | OS
buddyinfo | Exposes statistics of memory fragments as reported by /proc/buddyinfo. | Linux
devstat | Exposes device statistics | Dragonfly, FreeBSD
drbd | Exposes Distributed Replicated Block Device statistics (to version 8.4) | Linux
ethtool | Exposes network interface and network driver statistics equivalent to `ethtool -S`. | Linux
ethtool | Exposes network interface and network driver statistics equivalent to `ethtool -S` and `ethtool -i`. | Linux
interrupts | Exposes detailed interrupts statistics. | Linux, OpenBSD
ksmd | Exposes kernel and system statistics from `/sys/kernel/mm/ksm`. | Linux
logind | Exposes session counts from [logind](http://www.freedesktop.org/wiki/Software/systemd/logind/). | Linux

View File

@ -45,28 +45,35 @@ var (
transmittedRegex = regexp.MustCompile(`(^|_)tx(_|$)`)
)
type EthtoolStats interface {
type Ethtool interface {
DriverInfo(string) (ethtool.DrvInfo, error)
Stats(string) (map[string]uint64, error)
}
type ethtoolStats struct {
type ethtoolLibrary struct {
ethtool *ethtool.Ethtool
}
func (e *ethtoolStats) Stats(intf string) (map[string]uint64, error) {
return ethtool.Stats(intf)
func (e *ethtoolLibrary) DriverInfo(intf string) (ethtool.DrvInfo, error) {
return e.ethtool.DriverInfo(intf)
}
func (e *ethtoolLibrary) Stats(intf string) (map[string]uint64, error) {
return e.ethtool.Stats(intf)
}
type ethtoolCollector struct {
fs sysfs.FS
entries map[string]*prometheus.Desc
ethtool Ethtool
ignoredDevicesPattern *regexp.Regexp
infoDesc *prometheus.Desc
metricsPattern *regexp.Regexp
logger log.Logger
stats EthtoolStats
}
// makeEthtoolCollector is the internal constructor for EthtoolCollector.
// This allows NewEthtoolTestCollector to override its .stats interface
// This allows NewEthtoolTestCollector to override its .ethtool interface
// for testing.
func makeEthtoolCollector(logger log.Logger) (*ethtoolCollector, error) {
fs, err := sysfs.NewFS(*sysPath)
@ -74,13 +81,18 @@ func makeEthtoolCollector(logger log.Logger) (*ethtoolCollector, error) {
return nil, fmt.Errorf("failed to open sysfs: %w", err)
}
e, err := ethtool.NewEthtool()
if err != nil {
return nil, fmt.Errorf("failed to initialize ethtool library: %w", err)
}
// Pre-populate some common ethtool metrics.
return &ethtoolCollector{
fs: fs,
ethtool: &ethtoolLibrary{e},
ignoredDevicesPattern: regexp.MustCompile(*ethtoolIgnoredDevices),
metricsPattern: regexp.MustCompile(*ethtoolIncludedMetrics),
logger: logger,
stats: &ethtoolStats{},
entries: map[string]*prometheus.Desc{
"rx_bytes": prometheus.NewDesc(
prometheus.BuildFQName(namespace, "ethtool", "received_bytes_total"),
@ -118,6 +130,11 @@ func makeEthtoolCollector(logger log.Logger) (*ethtoolCollector, error) {
[]string{"device"}, nil,
),
},
infoDesc: prometheus.NewDesc(
prometheus.BuildFQName(namespace, "ethtool", "info"),
"A metric with a constant '1' value labeled by bus_info, device, driver, expansion_rom_version, firmware_version, version.",
[]string{"bus_info", "device", "driver", "expansion_rom_version", "firmware_version", "version"}, nil,
),
}, nil
}
@ -175,7 +192,24 @@ func (c *ethtoolCollector) Update(ch chan<- prometheus.Metric) error {
continue
}
stats, err = c.stats.Stats(device)
drvInfo, err := c.ethtool.DriverInfo(device)
if err == nil {
ch <- prometheus.MustNewConstMetric(c.infoDesc, prometheus.GaugeValue, 1.0,
drvInfo.BusInfo, device, drvInfo.Driver, drvInfo.EromVersion, drvInfo.FwVersion, drvInfo.Version)
} else {
if errno, ok := err.(syscall.Errno); ok {
if err == unix.EOPNOTSUPP {
level.Debug(c.logger).Log("msg", "ethtool driver info error", "err", err, "device", device, "errno", uint(errno))
} else if errno != 0 {
level.Error(c.logger).Log("msg", "ethtool driver info error", "err", err, "device", device, "errno", uint(errno))
}
} else {
level.Error(c.logger).Log("msg", "ethtool driver info error", "err", err, "device", device)
}
}
stats, err = c.ethtool.Stats(device)
// If Stats() returns EOPNOTSUPP it doesn't support ethtool stats. Log that only at Debug level.
// Otherwise log it at Error level.

View File

@ -25,6 +25,7 @@ import (
"github.com/go-kit/log"
"github.com/prometheus/client_golang/prometheus"
"github.com/safchain/ethtool"
"golang.org/x/sys/unix"
)
@ -32,10 +33,49 @@ type EthtoolFixture struct {
fixturePath string
}
func (e *EthtoolFixture) DriverInfo(intf string) (ethtool.DrvInfo, error) {
res := ethtool.DrvInfo{}
fixtureFile, err := os.Open(filepath.Join(e.fixturePath, intf, "driver"))
if e, ok := err.(*os.PathError); ok && e.Err == syscall.ENOENT {
// The fixture for this interface doesn't exist. Translate that to unix.EOPNOTSUPP
// to replicate an interface that doesn't support ethtool driver info
return res, unix.EOPNOTSUPP
}
if err != nil {
return res, err
}
defer fixtureFile.Close()
scanner := bufio.NewScanner(fixtureFile)
for scanner.Scan() {
line := scanner.Text()
if strings.HasPrefix(line, "#") {
continue
}
line = strings.Trim(line, " ")
items := strings.Split(line, ": ")
switch items[0] {
case "driver":
res.Driver = items[1]
case "version":
res.Version = items[1]
case "firmware-version":
res.FwVersion = items[1]
case "bus-info":
res.BusInfo = items[1]
case "expansion-rom-version":
res.EromVersion = items[1]
}
}
return res, err
}
func (e *EthtoolFixture) Stats(intf string) (map[string]uint64, error) {
res := make(map[string]uint64)
fixtureFile, err := os.Open(filepath.Join(e.fixturePath, intf))
fixtureFile, err := os.Open(filepath.Join(e.fixturePath, intf, "statistics"))
if e, ok := err.(*os.PathError); ok && e.Err == syscall.ENOENT {
// The fixture for this interface doesn't exist. Translate that to unix.EOPNOTSUPP
// to replicate an interface that doesn't support ethtool stats
@ -72,7 +112,7 @@ func (e *EthtoolFixture) Stats(intf string) (map[string]uint64, error) {
func NewEthtoolTestCollector(logger log.Logger) (Collector, error) {
collector, err := makeEthtoolCollector(logger)
collector.stats = &EthtoolFixture{
collector.ethtool = &EthtoolFixture{
fixturePath: "fixtures/ethtool/",
}
if err != nil {
@ -118,6 +158,9 @@ func TestBuildEthtoolFQName(t *testing.T) {
func TestEthtoolCollector(t *testing.T) {
testcases := []string{
prometheus.NewDesc("node_ethtool_info",
"A metric with a constant '1' value labeled by bus_info, device, driver, expansion_rom_version, firmware_version, version.",
[]string{"bus_info", "device", "driver", "expansion_rom_version", "firmware_version", "version"}, nil).String(),
prometheus.NewDesc("node_ethtool_align_errors", "Network interface align_errors", []string{"device"}, nil).String(),
prometheus.NewDesc("node_ethtool_received_broadcast", "Network interface rx_broadcast", []string{"device"}, nil).String(),
prometheus.NewDesc("node_ethtool_received_errors_total", "Number of received frames with errors", []string{"device"}, nil).String(),

View File

@ -0,0 +1,11 @@
# ethtool -i eth0
driver: e1000e
version: 5.11.0-22-generic
firmware-version: 0.5-4
expansion-rom-version:
bus-info: 0000:00:1f.6
supports-statistics: yes
supports-test: yes
supports-eeprom-access: yes
supports-register-dump: yes
supports-priv-flags: yes