Merge pull request #1454 from tkatila/xpum-sidecar-updates

Xpum sidecar updates

Author: tkatila

cmd/xpumanager_sidecar/README.md

@@ -21,6 +21,8 @@ Intel GPUs can be interconnected via an XeLink. In some workloads it is benefici
 | -interval | int | 10 | Interval for XeLink topology fetching and label writing (seconds, >= 1) |
 | -startup-delay | int | 10 | Startup delay before the first topology fetching (seconds, >= 0) |
 | -label-namespace | string | gpu.intel.com | Namespace or prefix for the labels. i.e. **gpu.intel.com**/xe-links |
+| -allow-subdeviceless-links | bool | false | Include xelinks that are not on subdevices |
+| -use-https | bool | false | Use HTTPS protocol when connecting to XPU-Manager |
 
 The sidecar also accepts a number of other arguments. Please use the -h option to see the complete list of options.
 

cmd/xpumanager_sidecar/main.go

@@ -18,6 +18,7 @@ import (
 	"bufio"
 	"bytes"
 	"context"
+	"crypto/tls"
 	"flag"
 	"fmt"
 	"io"
@@ -55,15 +56,17 @@ type xpuManagerTopologyMatrixCell struct {
 }
 
 type xpuManagerSidecar struct {
-	getMetricsData func() []byte
-	tmpDirPrefix   string
-	dstFilePath    string
-	labelNamespace string
-	url            string
-	interval       uint64
-	startDelay     uint64
-	xpumPort       uint64
-	laneCount      uint64
+	getMetricsData          func() []byte
+	tmpDirPrefix            string
+	dstFilePath             string
+	labelNamespace          string
+	url                     string
+	interval                uint64
+	startDelay              uint64
+	xpumPort                uint64
+	laneCount               uint64
+	allowSubdevicelessLinks bool
+	useHTTPS                bool
 }
 
 func (e *invalidEntryErr) Error() string {
@@ -75,6 +78,14 @@ func (xms *xpuManagerSidecar) getMetricsDataFromXPUM() []byte {
 		Timeout: 5 * time.Second,
 	}
 
+	if xms.useHTTPS {
+		customTransport := http.DefaultTransport.(*http.Transport).Clone()
+		//#nosec
+		customTransport.TLSClientConfig = &tls.Config{InsecureSkipVerify: true}
+
+		client.Transport = customTransport
+	}
+
 	ctx := context.Background()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, xms.url, http.NoBody)
@@ -108,7 +119,7 @@ func (xms *xpuManagerSidecar) getMetricsDataFromXPUM() []byte {
 	return resBody
 }
 
-func processMetricsLabels(labels []*io_prometheus_client.LabelPair) (xpuManagerTopologyMatrixCell, error) {
+func processMetricsLabels(labels []*io_prometheus_client.LabelPair, allowNonSubdeviceLinks bool) (xpuManagerTopologyMatrixCell, error) {
 	cell := createInvalidTopologyCell()
 
 	for _, label := range labels {
@@ -118,7 +129,7 @@ func processMetricsLabels(labels []*io_prometheus_client.LabelPair) (xpuManagerT
 		klog.V(5).Info(name, " ", strVal)
 
 		// xelinks should always be on subdevices
-		if name == "local_on_subdevice" && strVal != "true" {
+		if !allowNonSubdeviceLinks && name == "local_on_subdevice" && strVal != "true" {
 			return cell, &invalidEntryErr{}
 		}
 
@@ -193,7 +204,7 @@ func (xms *xpuManagerSidecar) GetTopologyFromXPUMMetrics(data []byte) (topologyI
 				continue
 			}
 
-			cell, err := processMetricsLabels(metric.Label)
+			cell, err := processMetricsLabels(metric.Label, xms.allowSubdevicelessLinks)
 			if err == nil {
 				klog.V(5).Info("topology entry: ", cell)
 				topologyInfos = append(topologyInfos, cell)
@@ -367,6 +378,8 @@ func main() {
 	flag.StringVar(&xms.dstFilePath, "dst-file-path", "/etc/kubernetes/node-feature-discovery/features.d/xpum-sidecar-labels.txt", "label file destination")
 	flag.Uint64Var(&xms.laneCount, "lane-count", 4, "minimum lane count for xelink")
 	flag.StringVar(&xms.labelNamespace, "label-namespace", "gpu.intel.com", "namespace for the labels")
+	flag.BoolVar(&xms.allowSubdevicelessLinks, "allow-subdeviceless-links", false, "allow xelinks that are not tied to subdevices (=1 tile GPUs)")
+	flag.BoolVar(&xms.useHTTPS, "use-https", false, "Use HTTPS protocol to connect to xpumanager")
 	klog.InitFlags(nil)
 
 	flag.Parse()
@@ -375,7 +388,12 @@ func main() {
 		klog.Fatal("zero interval won't work, set it to at least 1")
 	}
 
-	xms.url = fmt.Sprintf("http://127.0.0.1:%d/metrics", xms.xpumPort)
+	protocol := "http"
+	if xms.useHTTPS {
+		protocol = "https"
+	}
+
+	xms.url = fmt.Sprintf("%s://127.0.0.1:%d/metrics", protocol, xms.xpumPort)
 
 	keepIterating := true
 

cmd/xpumanager_sidecar/main_test.go

@@ -23,10 +23,11 @@ import (
 )
 
 type testCase struct {
-	name           string
-	metricsData    []string
-	expectedLabels []string
-	minLaneCount   int
+	name               string
+	metricsData        []string
+	expectedLabels     []string
+	minLaneCount       int
+	allowSubdeviceless bool
 }
 
 func createTestCases() []testCase {
@@ -59,12 +60,25 @@ func createTestCases() []testCase {
 			metricsData: []string{
 				`# HELP xpum_topology_link Connection type fo two GPU tiles`,
 				`# TYPE xpum_topology_link gauge`,
-				`xpum_topology_link{dev_file="card1",dev_name="Intel(R) Graphics [0x0bdb]",pci_bdf="0000:51:00.0",pci_dev="0xbdb",src="direct",uuid="01000000-0000-0000-0000-000000510000",vendor="Intel(R) Corporation",local_cpu_affinity="0-23,48-71",local_device_id="0",local_numa_index="0",local_on_subdevice="false",local_subdevice_id="0",remote_device_id="0",remote_subdevice_id="0"} 1`,
-				`xpum_topology_link{dev_file="card1",dev_name="Intel(R) Graphics [0x0bdb]",pci_bdf="0000:51:00.0",pci_dev="0xbdb",src="direct",uuid="01000000-0000-0000-0000-000000510000",vendor="Intel(R) Corporation",local_cpu_affinity="0-23,48-71",local_device_id="0",local_numa_index="0",local_on_subdevice="false",local_subdevice_id="0",remote_device_id="1",remote_subdevice_id="0"} 1`,
+				`xpum_topology_link{dev_file="card1",dev_name="Intel(R) Graphics [0x0bdb]",pci_bdf="0000:51:00.0",pci_dev="0xbdb",src="direct",uuid="01000000-0000-0000-0000-000000510000",vendor="Intel(R) Corporation",local_cpu_affinity="0-23,48-71",local_device_id="0",local_numa_index="0",local_on_subdevice="false",local_subdevice_id="0",remote_device_id="1",remote_subdevice_id="0",lane_count="4"} 1`,
+				`xpum_topology_link{dev_file="card1",dev_name="Intel(R) Graphics [0x0bdb]",pci_bdf="0000:51:00.0",pci_dev="0xbdb",src="direct",uuid="01000000-0000-0000-0000-000000510000",vendor="Intel(R) Corporation",local_cpu_affinity="0-23,48-71",local_device_id="0",local_numa_index="0",local_on_subdevice="false",local_subdevice_id="0",remote_device_id="1",remote_subdevice_id="1",lane_count="4"} 1`,
 				"",
 			},
 			expectedLabels: []string{"xpumanager.intel.com/xe-links="},
 		},
+		{
+			name:         "Xelinks not on sub devices when it's allowed",
+			minLaneCount: 4,
+			metricsData: []string{
+				`# HELP xpum_topology_link Connection type fo two GPU tiles`,
+				`# TYPE xpum_topology_link gauge`,
+				`xpum_topology_link{dev_file="card1",dev_name="Intel(R) Graphics [0x0bdb]",pci_bdf="0000:51:00.0",pci_dev="0xbdb",src="direct",uuid="01000000-0000-0000-0000-000000510000",vendor="Intel(R) Corporation",local_cpu_affinity="0-23,48-71",local_device_id="0",local_numa_index="0",local_on_subdevice="false",local_subdevice_id="0",remote_device_id="1",remote_subdevice_id="0",lane_count="4"} 1`,
+				`xpum_topology_link{dev_file="card1",dev_name="Intel(R) Graphics [0x0bdb]",pci_bdf="0000:51:00.0",pci_dev="0xbdb",src="direct",uuid="01000000-0000-0000-0000-000000510000",vendor="Intel(R) Corporation",local_cpu_affinity="0-23,48-71",local_device_id="0",local_numa_index="0",local_on_subdevice="false",local_subdevice_id="0",remote_device_id="1",remote_subdevice_id="1",lane_count="4"} 1`,
+				"",
+			},
+			expectedLabels:     []string{"xpumanager.intel.com/xe-links=0.0-1.0_0.0-1.1"},
+			allowSubdeviceless: true,
+		},
 		{
 			name:         "Xelinks without lan counts",
 			minLaneCount: 4,
@@ -208,6 +222,9 @@ func TestLabeling(t *testing.T) {
 	for _, tc := range tcs {
 		print("Testcase (labeling): ", tc.name, "\n")
 		xms := tc.createFakeXMS(tc.metricsData, tc.minLaneCount)
+
+		xms.allowSubdevicelessLinks = tc.allowSubdeviceless
+
 		topologyInfos := xms.GetTopologyFromXPUMMetrics([]byte(strings.Join(tc.metricsData, "\n")))
 
 		labels := xms.createLabels(topologyInfos)
@@ -224,6 +241,8 @@ func TestIterate(t *testing.T) {
 		print("Testcase (iterate): ", tc.name, "\n")
 		xms := tc.createFakeXMS(tc.metricsData, tc.minLaneCount)
 
+		xms.allowSubdevicelessLinks = tc.allowSubdeviceless
+
 		root, err := os.MkdirTemp("", "test_new_xms")
 		if err != nil {
 			t.Fatalf("can't create temporary directory: %+v", err)