Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 21 additions & 4 deletions internal/collector/cgroup_node_metrics_collector_linux.go
Original file line number Diff line number Diff line change
Expand Up @@ -174,12 +174,29 @@ func (c *cgroupNodeMetricsCollector) resolveNodeName(nodes []*corev1.Node) (stri

// readNodeMemoryUsage helper
func readNodeMemoryUsage(logger *slog.Logger, memPath string) int64 {
// 1. Try cgroup first
val, err := readCgroupMemory(logger, memPath)
if err == nil && val > 0 {
if val > 9223372036854775807 { // MaxInt64
return 9223372036854775807
}
return int64(val)
}

// 2. Fallback to /proc/meminfo
// This is often necessary in containers where /sys/fs/cgroup/memory may not reflect the node's
// full usage or permissions are restricted.
if logger != nil && logger.Enabled(context.Background(), slog.LevelDebug) {
logger.Debug("cgroup memory read failed or zero; falling back to /proc/meminfo", slog.Any("error", err))
}

memInfo, err := readProcMemInfo("/proc/meminfo")
if err != nil {
if logger != nil {
logger.Warn("failed to read node memory from both cgroup and /proc/meminfo", slog.String("error", err.Error()))
}
return 0
}
if val > 9223372036854775807 { // MaxInt64
return 9223372036854775807
}
return int64(val)

return memInfo
}
65 changes: 65 additions & 0 deletions internal/collector/cgroup_node_metrics_test.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
package collector

import (
"os"
"path/filepath"
"testing"
)

func TestReadProcMemInfo(t *testing.T) {
// Create a temporary file mimicking /proc/meminfo
tmpDir := t.TempDir()
memInfoFile := filepath.Join(tmpDir, "meminfo")

content := `MemTotal: 16306544 kB
MemFree: 4354264 kB
MemAvailable: 8992348 kB
Buffers: 345124 kB
Cached: 4567232 kB
SwapCached: 0 kB
`
if err := os.WriteFile(memInfoFile, []byte(content), 0o644); err != nil {
t.Fatalf("failed to write meminfo mock: %v", err)
}

// Test Case 1: Standard MemAvailable present
usage, err := readProcMemInfo(memInfoFile)
if err != nil {
t.Fatalf("readProcMemInfo failed: %v", err)
}

// Expected: Total (16306544) - Available (8992348) = 7314196 kB
// In bytes: 7314196 * 1024 = 7,489,736,704
expectedKB := int64(16306544 - 8992348)
expectedBytes := expectedKB * 1024

if usage != expectedBytes {
t.Errorf("expected usage %d bytes, got %d", expectedBytes, usage)
}

// Test Case 2: No MemAvailable (older kernels)
oldContent := `MemTotal: 16306544 kB
MemFree: 4354264 kB
Buffers: 345124 kB
Cached: 4567232 kB
`
memInfoFileOld := filepath.Join(tmpDir, "meminfo_old")
if err := os.WriteFile(memInfoFileOld, []byte(oldContent), 0o644); err != nil {
t.Fatalf("failed to write meminfo old mock: %v", err)
}

usageOld, err := readProcMemInfo(memInfoFileOld)
if err != nil {
t.Fatalf("readProcMemInfo (old) failed: %v", err)
}

// Expected: Total - Free - Buffers - Cached
// 16306544 - 4354264 - 345124 - 4567232 = 7039924 kB
// 7039924 * 1024 = 7,208,882,176
expectedKBOld := int64(16306544 - 4354264 - 345124 - 4567232)
expectedBytesOld := expectedKBOld * 1024

if usageOld != expectedBytesOld {
t.Errorf("expected usage (old) %d bytes, got %d", expectedBytesOld, usageOld)
}
}
64 changes: 64 additions & 0 deletions internal/collector/node_memory.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,64 @@
package collector

import (
"fmt"
"os"
"strconv"
"strings"
)

// readProcMemInfo parses /proc/meminfo to calculate used memory (Total - Available)
// This function is kept platform-independent for easier testing, although /proc/meminfo is Linux-specific.
func readProcMemInfo(path string) (int64, error) {
data, err := os.ReadFile(path) // #nosec G304
if err != nil {
return 0, err
}

var memTotal, memAvailable, memFree, buffers, cached uint64
var foundTotal, foundAvailable bool

lines := strings.Split(string(data), "\n")
for _, line := range lines {
fields := strings.Fields(line)
if len(fields) < 2 {
continue
}
key := strings.TrimSuffix(fields[0], ":")
val, err := strconv.ParseUint(fields[1], 10, 64)
if err != nil {
continue
}
// /proc/meminfo values are in kB
valBytes := val * 1024

switch key {
case "MemTotal":
memTotal = valBytes
foundTotal = true
case "MemAvailable":
memAvailable = valBytes
foundAvailable = true
case "MemFree":
memFree = valBytes
case "Buffers":
buffers = valBytes
case "Cached":
cached = valBytes
}
}

if !foundTotal {
return 0, fmt.Errorf("MemTotal not found in %s", path)
}

// Ideally use MemAvailable (kernels 3.14+)
if foundAvailable {
// Used = Total - Available
return int64(memTotal - memAvailable), nil
}

// Fallback for older kernels: Used = Total - Free - Buffers - Cached
used := memTotal - memFree - buffers - cached
return int64(used), nil
}
Loading