Merge pull request #127910 from leonzz/leonzz-tmp

fix node start time inconsistency in kubelet

pkg/kubelet/util/boottime_util_linux.go

@@ -21,17 +21,54 @@ package util
 
 import (
 	"fmt"
+	"os"
+	"strconv"
+	"strings"
 	"time"
 
 	"golang.org/x/sys/unix"
+	"k8s.io/klog/v2"
 )
 
-// GetBootTime returns the time at which the machine was started, truncated to the nearest second
+// GetBootTime returns the time at which the machine was started, truncated to the nearest second.
+// It uses /proc/stat first, which is more accurate, and falls back to the less accurate
+// unix.Sysinfo if /proc/stat failed.
 func GetBootTime() (time.Time, error) {
+	bootTime, err := getBootTimeWithProcStat()
+	if err != nil {
+		klog.InfoS("Failed to get boot time from /proc/uptime. Will retry with unix.Sysinfo.", "error", err)
+		return getBootTimeWithSysinfo()
+	}
+	return bootTime, nil
+}
+
+func getBootTimeWithProcStat() (time.Time, error) {
+	raw, err := os.ReadFile("/proc/stat")
+	if err != nil {
+		return time.Time{}, fmt.Errorf("error getting boot time: %w", err)
+	}
+	rawFields := strings.Fields(string(raw))
+	for i, v := range rawFields {
+		if v == "btime" {
+			if len(rawFields) > i+1 {
+				sec, err := strconv.ParseInt(rawFields[i+1], 10, 64)
+				if err != nil {
+					return time.Time{}, fmt.Errorf("error parsing boot time %s: %w", rawFields[i+1], err)
+				}
+				return time.Unix(sec, 0), nil
+			}
+			break
+		}
+	}
+
+	return time.Time{}, fmt.Errorf("can not find btime from /proc/stat: %s", raw)
+}
+
+func getBootTimeWithSysinfo() (time.Time, error) {
 	currentTime := time.Now()
 	var info unix.Sysinfo_t
 	if err := unix.Sysinfo(&info); err != nil {
-		return time.Time{}, fmt.Errorf("error getting system uptime: %s", err)
+		return time.Time{}, fmt.Errorf("error getting system uptime: %w", err)
 	}
 	return currentTime.Add(-time.Duration(info.Uptime) * time.Second).Truncate(time.Second), nil
 }

pkg/kubelet/util/boottime_util_linux_test.go

@@ -35,3 +35,18 @@ func TestGetBootTime(t *testing.T) {
 		t.Errorf("Invalid system uptime")
 	}
 }
+
+func TestVariationBetweenGetBootTimeMethods(t *testing.T) {
+	boottime1, err := getBootTimeWithProcStat()
+	if err != nil {
+		t.Errorf("Unable to get boot time from /proc/uptime")
+	}
+	boottime2, err := getBootTimeWithSysinfo()
+	if err != nil {
+		t.Errorf("Unable to get boot time from unix.Sysinfo")
+	}
+	diff := boottime1.Sub(boottime2)
+	if diff > time.Second || diff < -time.Second {
+		t.Errorf("boot time produced by 2 methods should not vary more than a second")
+	}
+}