fingerprint: add config option to disable dmidecode (#25108)

.changelog/25108.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+cpustats: Add config "cpu_disable_dmidecode" to disable cpu detection using dmidecode
+```

client/config/config.go

@@ -123,6 +123,9 @@ type Config struct {
 	// be determined dynamically.
 	NetworkSpeed int
 
+	// CpuDisableDmidecode disables cpu fingerprinting using dmidecode on Linux
+	CpuDisableDmidecode bool
+
 	// CpuCompute is the default total CPU compute if they can not be determined
 	// dynamically. It should be given as Cores * MHz (2 Cores * 2 Ghz = 4000)
 	CpuCompute int

client/driver_manager_test.go

@@ -19,7 +19,7 @@ import (
 )
 
 var (
-	topology = numalib.Scan(numalib.PlatformScanners())
+	topology = numalib.Scan(numalib.PlatformScanners(false))
 )
 
 // TestDriverManager_Fingerprint_Run asserts that node is populated with

client/fingerprint/cpu.go

@@ -88,7 +88,7 @@ func (f *CPUFingerprint) initialize(request *FingerprintRequest) {
 	}
 
 	f.top = numalib.Scan(append(
-		numalib.PlatformScanners(),
+		numalib.PlatformScanners(request.Config.CpuDisableDmidecode),
 		&numalib.ConfigScanner{
 			ReservableCores: reservableCores,
 			ReservedCores:   reservedCores,

client/lib/numalib/detect_darwin.go

@@ -13,7 +13,7 @@ import (
 )
 
 // PlatformScanners returns the set of SystemScanner for macOS.
-func PlatformScanners() []SystemScanner {
+func PlatformScanners(_ bool) []SystemScanner {
 	return []SystemScanner{
 		new(MacOS),
 	}

client/lib/numalib/detect_default.go

@@ -7,7 +7,7 @@ package numalib
 
 // PlatformScanners returns the set of SystemScanner for systems without a
 // specific implementation.
-func PlatformScanners() []SystemScanner {
+func PlatformScanners(_ bool) []SystemScanner {
 	return []SystemScanner{
 		new(Generic),
 	}

client/lib/numalib/detect_linux.go

@@ -19,14 +19,16 @@ import (
 )
 
 // PlatformScanners returns the set of SystemScanner for Linux.
-func PlatformScanners() []SystemScanner {
-	return []SystemScanner{
-		new(Sysfs),
-		new(Smbios),
-		new(Cgroups1),
-		new(Cgroups2),
-		new(Fallback),
+func PlatformScanners(cpuDisableDmidecode bool) []SystemScanner {
+	scanners := []SystemScanner{new(Sysfs)}
+	if !cpuDisableDmidecode {
+		scanners = append(scanners, new(Smbios))
 	}
+	scanners = append(scanners, new(Cgroups1))
+	scanners = append(scanners, new(Cgroups2))
+	scanners = append(scanners, new(Fallback))
+
+	return scanners
 }
 
 const (

client/lib/numalib/detect_noimpl_test.go

@@ -17,7 +17,7 @@ func Test_NoImpl_yes(t *testing.T) {
 }
 
 func Test_NoImpl_no(t *testing.T) {
-	original := Scan(PlatformScanners())
+	original := Scan(PlatformScanners(false))
 	fallback := NoImpl(original)
 	must.EqOp(t, original, fallback) // pointer is same
 }

client/lib/numalib/detect_test.go

@@ -12,7 +12,7 @@ import (
 // TestScanTopology is going to be different on every machine; even the CI
 // systems change sometimes so it's hard to make good assertions here.
 func TestScanTopology(t *testing.T) {
-	top := Scan(PlatformScanners())
+	top := Scan(PlatformScanners(false))
 	must.Positive(t, top.UsableCompute())
 	must.Positive(t, top.TotalCompute())
 	must.Positive(t, top.NumCores())

client/pluginmanager/drivermanager/testing.go

@@ -27,7 +27,7 @@ type testManager struct {
 }
 
 func TestDriverManager(t *testing.T) Manager {
-	topology := numalib.Scan(numalib.PlatformScanners())
+	topology := numalib.Scan(numalib.PlatformScanners(false))
 	logger := testlog.HCLogger(t).Named("driver_mgr")
 	pluginLoader := catalog.TestPluginLoader(t)
 	return &testManager{

command/agent/agent.go

@@ -755,6 +755,9 @@ func convertClientConfig(agentConfig *Config) (*clientconfig.Config, error) {
 	if agentConfig.Client.NetworkSpeed != 0 {
 		conf.NetworkSpeed = agentConfig.Client.NetworkSpeed
 	}
+	if agentConfig.Client.CpuDisableDmidecode {
+		conf.CpuDisableDmidecode = agentConfig.Client.CpuDisableDmidecode
+	}
 	if agentConfig.Client.CpuCompute != 0 {
 		conf.CpuCompute = agentConfig.Client.CpuCompute
 	}

command/agent/config.go

@@ -275,6 +275,9 @@ type ClientConfig struct {
 	// speed.
 	NetworkSpeed int `hcl:"network_speed"`
 
+	// CpuDisableDmidecode is used to disable dmidecode usage for CPU calculation
+	CpuDisableDmidecode bool `hcl:"cpu_disable_dmidecode"`
+
 	// CpuCompute is used to override any detected or default total CPU compute.
 	CpuCompute int `hcl:"cpu_total_compute"`
 
@@ -2346,6 +2349,9 @@ func (a *ClientConfig) Merge(b *ClientConfig) *ClientConfig {
 	if b.NetworkSpeed != 0 {
 		result.NetworkSpeed = b.NetworkSpeed
 	}
+	if b.CpuDisableDmidecode {
+		result.CpuDisableDmidecode = b.CpuDisableDmidecode
+	}
 	if b.CpuCompute != 0 {
 		result.CpuCompute = b.CpuCompute
 	}

drivers/docker/driver_test.go

@@ -68,7 +68,7 @@ var (
 )
 
 var (
-	top = numalib.Scan(numalib.PlatformScanners())
+	top = numalib.Scan(numalib.PlatformScanners(false))
 )
 
 func dockerIsRemote() bool {

drivers/docker/fingerprint_test.go

@@ -16,7 +16,7 @@ import (
 )
 
 var (
-	topology = numalib.Scan(numalib.PlatformScanners())
+	topology = numalib.Scan(numalib.PlatformScanners(false))
 )
 
 // TestDockerDriver_FingerprintHealth asserts that docker reports healthy

drivers/exec/driver_test.go

@@ -75,7 +75,7 @@ func testResources(allocID, task string) *drivers.Resources {
 }
 
 func newExecDriverTest(t *testing.T, ctx context.Context) drivers.DriverPlugin {
-	topology := numalib.Scan(numalib.PlatformScanners())
+	topology := numalib.Scan(numalib.PlatformScanners(false))
 	d := NewExecDriver(ctx, testlog.HCLogger(t))
 	d.(*Driver).nomadConfig = &base.ClientDriverConfig{Topology: topology}
 	d.(*Driver).userIDValidator = &mockIDValidator{}

drivers/java/driver_test.go

@@ -39,7 +39,7 @@ func javaCompatible(t *testing.T) {
 }
 
 func newJavaDriverTest(t *testing.T, ctx context.Context) drivers.DriverPlugin {
-	topology := numalib.Scan(numalib.PlatformScanners())
+	topology := numalib.Scan(numalib.PlatformScanners(false))
 	d := NewDriver(ctx, testlog.HCLogger(t))
 	d.(*Driver).nomadConfig = &base.ClientDriverConfig{Topology: topology}
 	return d

drivers/qemu/driver_test.go

@@ -38,7 +38,7 @@ func TestQemuDriver_Start_Wait_Stop(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
-	topology := numalib.Scan(numalib.PlatformScanners())
+	topology := numalib.Scan(numalib.PlatformScanners(false))
 	d := NewQemuDriver(ctx, testlog.HCLogger(t))
 	d.(*Driver).nomadConfig = &base.ClientDriverConfig{Topology: topology}
 	harness := dtestutil.NewDriverHarness(t, d)
@@ -116,7 +116,7 @@ func TestQemuDriver_User(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
-	topology := numalib.Scan(numalib.PlatformScanners())
+	topology := numalib.Scan(numalib.PlatformScanners(false))
 	d := NewQemuDriver(ctx, testlog.HCLogger(t))
 	d.(*Driver).nomadConfig = &base.ClientDriverConfig{Topology: topology}
 	harness := dtestutil.NewDriverHarness(t, d)
@@ -161,7 +161,7 @@ func TestQemuDriver_Stats(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
-	topology := numalib.Scan(numalib.PlatformScanners())
+	topology := numalib.Scan(numalib.PlatformScanners(false))
 	d := NewQemuDriver(ctx, testlog.HCLogger(t))
 	d.(*Driver).nomadConfig = &base.ClientDriverConfig{Topology: topology}
 	harness := dtestutil.NewDriverHarness(t, d)

drivers/rawexec/driver_test.go

@@ -73,7 +73,7 @@ func TestMain(m *testing.M) {
 }
 
 var (
-	topology = numalib.Scan(numalib.PlatformScanners())
+	topology = numalib.Scan(numalib.PlatformScanners(false))
 )
 
 type mockIDValidator struct{}

drivers/shared/executor/executor_test.go

@@ -55,7 +55,7 @@ func init() {
 }
 
 var (
-	topology = numalib.Scan(numalib.PlatformScanners())
+	topology = numalib.Scan(numalib.PlatformScanners(false))
 	compute  = topology.Compute()
 )
 
@@ -341,7 +341,7 @@ func TestExecutor_Shutdown_Exit(t *testing.T) {
 	}
 
 	driverCfg := &base.ClientDriverConfig{
-		Topology: numalib.Scan(numalib.PlatformScanners()),
+		Topology: numalib.Scan(numalib.PlatformScanners(false)),
 	}
 
 	executor, pluginClient, err := CreateExecutor(testlog.HCLogger(t), driverCfg, cfg)

drivers/shared/executor/executor_windows_test.go

@@ -61,7 +61,7 @@ func testExecutorCommand(t *testing.T) *testExecCmd {
 func TestExecutor_ProcessExit(t *testing.T) {
 	ci.Parallel(t)
 
-	topology := numalib.Scan(numalib.PlatformScanners())
+	topology := numalib.Scan(numalib.PlatformScanners(false))
 	compute := topology.Compute()
 
 	cmd := testExecutorCommand(t)

website/content/docs/configuration/client.mdx

@@ -78,6 +78,9 @@ client {
   the preferred family. When the option is not specified, the current behavior is conserved:
   the first IP address is selected no matter the family.
 
+- `cpu_disable_dmidecode` `(bool: false)` - Specifies the client should not use dmidecode
+  as a method of cpu detection. Nomad ignores this field on all platforms except Linux.
+
 - `cpu_total_compute` `(int: 0)` - Specifies an override for the total CPU
   compute. This value should be set to `# Cores * Core MHz`. For example, a
   quad-core running at 2 GHz would have a total compute of 8000 (4 \* 2000). Most