feat(config): allow config dir to be passed as argument (#1850)

This commit now allows default hard-coded config directory to be passed
as an argument. This allow quickly changing between different
configurations to be stored separately and to switch between them
(especially during development).

The commit also
* simplifies global config initialization by ensuring  it is initialised
  at the time kepler's main function is executed and fail with error that
  step fails.

* It also cleans up use of config object to read CGroup info by creating
  a `realSystem` struct that handles this functionality.

Signed-off-by: Sunil Thaha <[email protected]>

cmd/exporter/exporter.go

@@ -53,6 +53,7 @@ const (
 
 // AppConfig holds the configuration info for the application.
 type AppConfig struct {
+	BaseDir                      string
 	Address                      string
 	MetricsPath                  string
 	EnableGPU                    bool
@@ -69,21 +70,22 @@ type AppConfig struct {
 
 func newAppConfig() *AppConfig {
 	// Initialize flags
-	_config := &AppConfig{}
-	flag.StringVar(&_config.Address, "address", "0.0.0.0:8888", "bind address")
-	flag.StringVar(&_config.MetricsPath, "metrics-path", "/metrics", "metrics path")
-	flag.BoolVar(&_config.EnableGPU, "enable-gpu", false, "whether enable gpu (need to have libnvidia-ml installed)")
-	flag.BoolVar(&_config.EnableEBPFCgroupID, "enable-cgroup-id", true, "whether enable eBPF to collect cgroup id")
-	flag.BoolVar(&_config.ExposeHardwareCounterMetrics, "expose-hardware-counter-metrics", true, "whether expose hardware counter as prometheus metrics")
-	flag.BoolVar(&_config.EnableMSR, "enable-msr", false, "whether MSR is allowed to obtain energy data")
-	flag.StringVar(&_config.Kubeconfig, "kubeconfig", "", "absolute path to the kubeconfig file, if empty we use the in-cluster configuration")
-	flag.BoolVar(&_config.ApiserverEnabled, "apiserver", true, "if apiserver is disabled, we collect pod information from kubelet")
-	flag.StringVar(&_config.RedfishCredFilePath, "redfish-cred-file-path", "", "path to the redfish credential file")
-	flag.BoolVar(&_config.ExposeEstimatedIdlePower, "expose-estimated-idle-power", false, "Whether to expose the estimated idle power as a metric")
-	flag.StringVar(&_config.MachineSpecFilePath, "machine-spec", "", "path to the machine spec file in json format")
-	flag.BoolVar(&_config.DisablePowerMeter, "disable-power-meter", false, "whether manually disable power meter read and forcefully apply the estimator for node powers")
-
-	return _config
+	cfg := &AppConfig{}
+	flag.StringVar(&cfg.BaseDir, "config-dir", config.BaseDir, "path to config base directory")
+	flag.StringVar(&cfg.Address, "address", "0.0.0.0:8888", "bind address")
+	flag.StringVar(&cfg.MetricsPath, "metrics-path", "/metrics", "metrics path")
+	flag.BoolVar(&cfg.EnableGPU, "enable-gpu", false, "whether enable gpu (need to have libnvidia-ml installed)")
+	flag.BoolVar(&cfg.EnableEBPFCgroupID, "enable-cgroup-id", true, "whether enable eBPF to collect cgroup id")
+	flag.BoolVar(&cfg.ExposeHardwareCounterMetrics, "expose-hardware-counter-metrics", true, "whether expose hardware counter as prometheus metrics")
+	flag.BoolVar(&cfg.EnableMSR, "enable-msr", false, "whether MSR is allowed to obtain energy data")
+	flag.StringVar(&cfg.Kubeconfig, "kubeconfig", "", "absolute path to the kubeconfig file, if empty we use the in-cluster configuration")
+	flag.BoolVar(&cfg.ApiserverEnabled, "apiserver", true, "if apiserver is disabled, we collect pod information from kubelet")
+	flag.StringVar(&cfg.RedfishCredFilePath, "redfish-cred-file-path", "", "path to the redfish credential file")
+	flag.BoolVar(&cfg.ExposeEstimatedIdlePower, "expose-estimated-idle-power", false, "Whether to expose the estimated idle power as a metric")
+	flag.StringVar(&cfg.MachineSpecFilePath, "machine-spec", "", "path to the machine spec file in json format")
+	flag.BoolVar(&cfg.DisablePowerMeter, "disable-power-meter", false, "whether manually disable power meter read and forcefully apply the estimator for node powers")
+
+	return cfg
 }
 
 func healthProbe(w http.ResponseWriter, req *http.Request) {
@@ -99,7 +101,11 @@ func main() {
 	klog.InitFlags(nil)
 	appConfig := newAppConfig() // Initialize appConfig and define flags
 	flag.Parse()                // Parse command-line flags
-	config.GetConfig()          // Initialize the configuration
+
+	if _, err := config.Initialize(appConfig.BaseDir); err != nil {
+		klog.Fatalf("Failed to initialize config: %v", err)
+	}
+
 	klog.Infof("Kepler running on version: %s", build.Version)
 
 	registry := metrics.GetRegistry()

pkg/collector/metric_collector_test.go

@@ -40,7 +40,13 @@ func newMockCollector(mockAttacher bpf.Exporter) *Collector {
 
 var _ = Describe("Test Collector Unit", func() {
 
+	BeforeEach(func() {
+		_, err := config.Initialize(".")
+		Expect(err).NotTo(HaveOccurred())
+	})
+
 	It("Get container power", func() {
+
 		bpfExporter := bpf.NewMockExporter(bpf.DefaultSupportedMetrics())
 		metricCollector := newMockCollector(bpfExporter)
 		// The default estimator model is the ratio

pkg/collector/stats/benchmark_test.go

@@ -21,10 +21,12 @@ import (
 	"github.com/sustainable-computing-io/kepler/pkg/bpf"
 	"github.com/sustainable-computing-io/kepler/pkg/collector"
 	"github.com/sustainable-computing-io/kepler/pkg/collector/stats"
+	"github.com/sustainable-computing-io/kepler/pkg/config"
 	"github.com/sustainable-computing-io/kepler/pkg/model"
 )
 
 func benchmarkNtesting(b *testing.B, processNumber int) {
+	_, _ = config.Initialize(".")
 	// enable metrics
 	stats.SetMockedCollectorMetrics()
 	// create node node metrics

pkg/collector/stats/container_stats_test.go

@@ -24,6 +24,11 @@ import (
 
 var _ = Describe("Test Container Metric", func() {
 
+	BeforeEach(func() {
+		_, err := config.Initialize(".")
+		Expect(err).NotTo(HaveOccurred())
+	})
+
 	It("Test ResetDeltaValues", func() {
 		SetMockedCollectorMetrics()
 		c := NewContainerStats("containerA", "podA", "test", "containerIDA")

pkg/collector/stats/node_stats_test.go

@@ -31,6 +31,9 @@ var _ = Describe("Test Node Metric", func() {
 	)
 
 	BeforeEach(func() {
+		_, err := config.Initialize(".")
+		Expect(err).NotTo(HaveOccurred())
+
 		SetMockedCollectorMetrics()
 		processMetrics = CreateMockedProcessStats(2)
 		nodeMetrics = CreateMockedNodeStats()

pkg/collector/stats/process_stats_test.go

@@ -8,6 +8,11 @@ import (
 
 var _ = Describe("ProcessMetric", func() {
 
+	BeforeEach(func() {
+		_, err := config.Initialize(".")
+		Expect(err).NotTo(HaveOccurred())
+	})
+
 	It("Test ResetDeltaValues", func() {
 		SetMockedCollectorMetrics()
 		metrics := CreateMockedProcessStats(1)

pkg/collector/stats/stats_test.go

@@ -8,7 +8,9 @@ import (
 
 var _ = Describe("Stats", func() {
 	It("Test InitAvailableParamAndMetrics", func() {
-		config.GetConfig()
+		_, err := config.Initialize(".")
+		Expect(err).NotTo(HaveOccurred())
+
 		config.SetEnabledHardwareCounterMetrics(false)
 		exp := []string{}
 		Expect(len(GetProcessFeatureNames()) >= len(exp)).To(BeTrue())

pkg/collector/stats/test_utils.go

@@ -31,7 +31,6 @@ const (
 // SetMockedCollectorMetrics adds all metric to a process, otherwise it will not create the right usageMetric with all elements. The usageMetric is used in the Prediction Power Models
 // TODO: do not use a fixed usageMetric array in the power models, a structured data is more disarable.
 func SetMockedCollectorMetrics() {
-	config.GetConfig()
 	if gpu := acc.GetActiveAcceleratorByType(config.GPU); gpu != nil {
 		err := gpu.Device().Init() // create structure instances that will be accessed to create a processMetric
 		klog.Fatalln(err)