Read drive IO stats from sysfs instead of procfs (#19131)

Currently, we read from `/proc/diskstats` which is found to be
un-reliable in k8s environments. We can read from `sysfs` instead.

Also, cache the latest drive io stats to find the diff and update
the metrics.
This commit is contained in:
Praveen raj Mani
2024-02-27 01:04:50 +05:30
committed by GitHub
parent 2b5e4b853c
commit 30c2596512
13 changed files with 243 additions and 192 deletions

View File

@@ -22,7 +22,9 @@ package disk
import (
"bufio"
"errors"
"fmt"
"io"
"os"
"path/filepath"
"strconv"
@@ -109,125 +111,63 @@ func GetInfo(path string, firstTime bool) (info Info, err error) {
return info, nil
}
const (
statsPath = "/proc/diskstats"
)
// GetDriveStats returns IO stats of the drive by its major:minor
func GetDriveStats(major, minor uint32) (iostats IOStats, err error) {
return readDriveStats(fmt.Sprintf("/sys/dev/block/%v:%v/stat", major, minor))
}
// GetAllDrivesIOStats returns IO stats of all drives found in the machine
func GetAllDrivesIOStats() (info AllDrivesIOStats, err error) {
proc, err := os.Open(statsPath)
func readDriveStats(statsFile string) (iostats IOStats, err error) {
stats, err := readStat(statsFile)
if err != nil {
return IOStats{}, err
}
if len(stats) < 11 {
return IOStats{}, fmt.Errorf("found invalid format while reading %v", statsFile)
}
// refer https://www.kernel.org/doc/Documentation/block/stat.txt
iostats = IOStats{
ReadIOs: stats[0],
ReadMerges: stats[1],
ReadSectors: stats[2],
ReadTicks: stats[3],
WriteIOs: stats[4],
WriteMerges: stats[5],
WriteSectors: stats[6],
WriteTicks: stats[7],
CurrentIOs: stats[8],
TotalTicks: stats[9],
ReqTicks: stats[10],
}
// as per the doc, only 11 fields are guaranteed
// only set if available
if len(stats) > 14 {
iostats.DiscardIOs = stats[11]
iostats.DiscardMerges = stats[12]
iostats.DiscardSectors = stats[13]
iostats.DiscardTicks = stats[14]
}
return
}
func readStat(fileName string) (stats []uint64, err error) {
file, err := os.Open(fileName)
if err != nil {
return nil, err
}
defer proc.Close()
defer file.Close()
ret := make(AllDrivesIOStats)
sc := bufio.NewScanner(proc)
for sc.Scan() {
line := sc.Text()
fields := strings.Fields(line)
if len(fields) < 11 {
continue
}
var err error
var ds IOStats
ds.ReadIOs, err = strconv.ParseUint((fields[3]), 10, 64)
if err != nil {
return ret, err
}
ds.ReadMerges, err = strconv.ParseUint((fields[4]), 10, 64)
if err != nil {
return ret, err
}
ds.ReadSectors, err = strconv.ParseUint((fields[5]), 10, 64)
if err != nil {
return ret, err
}
ds.ReadTicks, err = strconv.ParseUint((fields[6]), 10, 64)
if err != nil {
return ret, err
}
ds.WriteIOs, err = strconv.ParseUint((fields[7]), 10, 64)
if err != nil {
return ret, err
}
ds.WriteMerges, err = strconv.ParseUint((fields[8]), 10, 64)
if err != nil {
return ret, err
}
ds.WriteSectors, err = strconv.ParseUint((fields[9]), 10, 64)
if err != nil {
return ret, err
}
ds.WriteTicks, err = strconv.ParseUint((fields[10]), 10, 64)
if err != nil {
return ret, err
}
if len(fields) > 11 {
ds.CurrentIOs, err = strconv.ParseUint((fields[11]), 10, 64)
if err != nil {
return ret, err
}
ds.TotalTicks, err = strconv.ParseUint((fields[12]), 10, 64)
if err != nil {
return ret, err
}
ds.ReqTicks, err = strconv.ParseUint((fields[13]), 10, 64)
if err != nil {
return ret, err
}
}
if len(fields) > 14 {
ds.DiscardIOs, err = strconv.ParseUint((fields[14]), 10, 64)
if err != nil {
return ret, err
}
ds.DiscardMerges, err = strconv.ParseUint((fields[15]), 10, 64)
if err != nil {
return ret, err
}
ds.DiscardSectors, err = strconv.ParseUint((fields[16]), 10, 64)
if err != nil {
return ret, err
}
ds.DiscardTicks, err = strconv.ParseUint((fields[17]), 10, 64)
if err != nil {
return ret, err
}
}
if len(fields) > 18 {
ds.FlushIOs, err = strconv.ParseUint((fields[18]), 10, 64)
if err != nil {
return ret, err
}
ds.FlushTicks, err = strconv.ParseUint((fields[19]), 10, 64)
if err != nil {
return ret, err
}
}
major, err := strconv.ParseUint((fields[0]), 10, 32)
if err != nil {
return ret, err
}
minor, err := strconv.ParseUint((fields[1]), 10, 32)
if err != nil {
return ret, err
}
ret[DevID{uint32(major), uint32(minor)}] = ds
}
if err := sc.Err(); err != nil {
s, err := bufio.NewReader(file).ReadString('\n')
if err != nil && !errors.Is(err, io.EOF) {
return nil, err
}
statLine := strings.TrimSpace(s)
for _, token := range strings.Fields(statLine) {
ui64, err := strconv.ParseUint(token, 10, 64)
if err != nil {
return nil, err
}
stats = append(stats, ui64)
}
return ret, nil
return stats, nil
}