run ut parallelly to make it fast (#141)

.github/workflows/ci.yml

@@ -7,8 +7,8 @@ on:
   workflow_dispatch: {}
 
 jobs:
-  lint:
-    name: Lint
+  fmt:
+    name: Format
     runs-on: ubuntu-20.04
     steps:
       - uses: actions/checkout@v3
@@ -24,10 +24,30 @@ jobs:
         run: |
           make fmt && git add pkg cmd &&
           git diff --cached --exit-code || (echo 'Please run "make fmt" to verify gofmt' && exit 1);
+
+  vet:
+    name: Vet
+    runs-on: ubuntu-20.04
+    steps:
+      - uses: actions/checkout@v3
+      - uses: actions/setup-go@v3
+        with:
+          go-version-file: go.mod
+          cache: false
       - name: Verify govet
         run: |
           make vet && git add pkg cmd &&
           git diff --cached --exit-code || (echo 'Please run "make vet" to verify govet' && exit 1);
+
+  lint:
+    name: Lint
+    runs-on: ubuntu-20.04
+    steps:
+      - uses: actions/checkout@v3
+      - uses: actions/setup-go@v3
+        with:
+          go-version-file: go.mod
+          cache: false
       - uses: golangci/golangci-lint-action@v3
         with:
           args: --verbose --out-${NO_FUTURE}format colored-line-number --config .golangci.yml

.golangci.yml

@@ -1,7 +1,7 @@
 # options for analysis running
 run:
-  # default concurrency is a available CPU number
-  concurrency: 4
+  # default concurrency is the available CPU number
+  concurrency: 16
 
   # timeout for analysis, e.g. 30s, 5m, default is 1m
   timeout: 10m

Makefile

@@ -112,7 +112,7 @@ vet: ## Run go vet against code.
 
 .PHONY: test
 test: ## Run go test against code.
-	go test -v -coverprofile=coverage.txt -covermode=atomic -race -coverpkg=./... ./...
+	go test -v -coverprofile=coverage.txt -parallel=16 -p=16 -covermode=atomic -race -coverpkg=./... ./pkg/...
 
 .PHONY: license
 license:

cmd/katalyst-agent/app/options/eviction/memory_pressure_eviction_test.go

@@ -26,6 +26,8 @@ import (
 )
 
 func TestMemoryPressureEvictionPluginOptions_ApplyTo(t *testing.T) {
+	t.Parallel()
+
 	options := NewMemoryPressureEvictionOptions()
 	configuration := eviction.NewMemoryPressureEvictionPluginConfiguration()
 

pkg/agent/evictionmanager/manager_test.go

@@ -228,6 +228,8 @@ func makeEvictionManager() *EvictionManger {
 }
 
 func TestEvictionManger_collectEvictionResult(t *testing.T) {
+	t.Parallel()
+
 	mgr := makeEvictionManager()
 	tests := []struct {
 		name               string

pkg/agent/evictionmanager/plugin/memory/helper_test.go

@@ -56,6 +56,8 @@ func makeHelper() (*EvictionHelper, error) {
 }
 
 func TestEvictionHelper_getEvictionCmpFuncs(t *testing.T) {
+	t.Parallel()
+
 	helper, err := makeHelper()
 	assert.NoError(t, err)
 	assert.NotNil(t, helper)

pkg/agent/evictionmanager/plugin/memory/numa_pressure_test.go

@@ -62,6 +62,8 @@ func makeNumaPressureEvictionPlugin(conf *config.Configuration) (*NumaMemoryPres
 }
 
 func TestNewNumaPressureEvictionPlugin(t *testing.T) {
+	t.Parallel()
+
 	plugin, err := makeNumaPressureEvictionPlugin(makeConf())
 	assert.NoError(t, err)
 	assert.NotNil(t, plugin)
@@ -74,6 +76,8 @@ func TestNewNumaPressureEvictionPlugin(t *testing.T) {
 }
 
 func TestNumaMemoryPressurePlugin_ThresholdMet(t *testing.T) {
+	t.Parallel()
+
 	plugin, err := makeNumaPressureEvictionPlugin(makeConf())
 	assert.NoError(t, err)
 	assert.NotNil(t, plugin)
@@ -213,6 +217,8 @@ func TestNumaMemoryPressurePlugin_ThresholdMet(t *testing.T) {
 }
 
 func TestNumaMemoryPressurePlugin_GetTopEvictionPods(t *testing.T) {
+	t.Parallel()
+
 	plugin, err := makeNumaPressureEvictionPlugin(makeConf())
 	assert.NoError(t, err)
 	assert.NotNil(t, plugin)

pkg/agent/evictionmanager/plugin/memory/rss_overuse_test.go

@@ -54,6 +54,8 @@ func makeRssOverusePlugin(conf *config.Configuration) (*RssOveruseEvictionPlugin
 }
 
 func TestRssOveruseEvictionPlugin_GetEvictPods(t *testing.T) {
+	t.Parallel()
+
 	plugin, err := makeRssOverusePlugin(makeConf())
 	assert.NoError(t, err)
 	assert.NotNil(t, plugin)

pkg/agent/evictionmanager/plugin/memory/system_pressure_test.go

@@ -97,6 +97,8 @@ func makeSystemPressureEvictionPlugin(conf *config.Configuration) (*SystemPressu
 }
 
 func TestNewSystemPressureEvictionPlugin(t *testing.T) {
+	t.Parallel()
+
 	plugin, err := makeSystemPressureEvictionPlugin(makeConf())
 	assert.NoError(t, err)
 	assert.NotNil(t, plugin)
@@ -111,6 +113,8 @@ func TestNewSystemPressureEvictionPlugin(t *testing.T) {
 }
 
 func TestSystemPressureEvictionPlugin_ThresholdMet(t *testing.T) {
+	t.Parallel()
+
 	plugin, err := makeSystemPressureEvictionPlugin(makeConf())
 	assert.NoError(t, err)
 	assert.NotNil(t, plugin)
@@ -290,6 +294,8 @@ func TestSystemPressureEvictionPlugin_ThresholdMet(t *testing.T) {
 }
 
 func TestSystemPressureEvictionPlugin_GetTopEvictionPods(t *testing.T) {
+	t.Parallel()
+
 	plugin, err := makeSystemPressureEvictionPlugin(makeConf())
 	assert.NoError(t, err)
 	assert.NotNil(t, plugin)

pkg/agent/evictionmanager/plugin/reclaimed_resources_test.go

@@ -63,6 +63,8 @@ func generateTestMetaServer(clientSet *client.GenericClientSet, conf *config.Con
 }
 
 func TestNewReclaimedResourcesEvictionPlugin(t *testing.T) {
+	t.Parallel()
+
 	testNodeName := "test-node"
 	testConf := generateTestConfiguration(t, testNodeName)
 	pods := []*corev1.Pod{

pkg/agent/evictionmanager/podkiller/killer_test.go

@@ -32,6 +32,8 @@ import (
 )
 
 func TestEvictionQueue(t *testing.T) {
+	t.Parallel()
+
 	pods := []*v1.Pod{
 		{
 			ObjectMeta: metav1.ObjectMeta{Name: "pod-1"},

pkg/agent/evictionmanager/rule/queue_test.go

@@ -43,6 +43,8 @@ func makeRuledEvictPod(name, scope string) *RuledEvictPod {
 }
 
 func TestEvictionQueue(t *testing.T) {
+	t.Parallel()
+
 	for _, tc := range []struct {
 		comment string
 		q       EvictionQueue

pkg/agent/evictionmanager/rule/rule_test.go

@@ -39,6 +39,8 @@ func makeRuledEvictPodForSort(name, scope string, annotations map[string]string,
 }
 
 func TestEvictionStrategyImp(t *testing.T) {
+	t.Parallel()
+
 	testConf, _ := options.NewOptions().Config()
 	s := NewEvictionStrategyImpl(testConf)
 

pkg/agent/qrm-plugins/cpu/dynamicpolicy/cpuadvisor/qos_aware_client_stub_test.go

@@ -24,16 +24,22 @@ import (
 )
 
 func TestClientAddContainer(t *testing.T) {
+	t.Parallel()
+
 	client := NewCPUAdvisorClientStub()
 	_, _ = client.AddContainer(context.Background(), &advisorsvc.AddContainerRequest{})
 }
 
 func TestClientRemovePod(t *testing.T) {
+	t.Parallel()
+
 	client := NewCPUAdvisorClientStub()
 	_, _ = client.RemovePod(context.Background(), &advisorsvc.RemovePodRequest{})
 }
 
 func TestClientListAndWatch(t *testing.T) {
+	t.Parallel()
+
 	client := NewCPUAdvisorClientStub()
 	_, _ = client.ListAndWatch(context.Background(), &advisorsvc.Empty{})
 }

pkg/agent/qrm-plugins/cpu/dynamicpolicy/cpueviction/strategy/pressure_load_test.go

@@ -78,14 +78,16 @@ func makeConf(metricRingSize int, gracePeriod int64, loadUpperBoundRatio,
 }
 
 func makeState(topo *machine.CPUTopology) (qrmstate.State, error) {
-	tmpDir, err := os.MkdirTemp("", "checkpoint")
+	tmpDir, err := os.MkdirTemp("", "checkpoint-makeState")
 	if err != nil {
 		return nil, fmt.Errorf("make tmp dir for checkpoint failed with error: %v", err)
 	}
 	return qrmstate.NewCheckpointState(tmpDir, "test", "test", topo, false)
 }
 
 func TestNewCPUPressureLoadEviction(t *testing.T) {
+	t.Parallel()
+
 	as := require.New(t)
 
 	cpuTopology, err := machine.GenerateDummyCPUTopology(16, 2, 4)
@@ -102,6 +104,8 @@ func TestNewCPUPressureLoadEviction(t *testing.T) {
 }
 
 func TestThresholdMet(t *testing.T) {
+	t.Parallel()
+
 	as := require.New(t)
 
 	cpuTopology, err := machine.GenerateDummyCPUTopology(16, 2, 4)
@@ -385,6 +389,8 @@ func TestThresholdMet(t *testing.T) {
 }
 
 func TestGetTopEvictionPods(t *testing.T) {
+	t.Parallel()
+
 	as := require.New(t)
 
 	cpuTopology, err := machine.GenerateDummyCPUTopology(16, 2, 4)

pkg/agent/qrm-plugins/cpu/dynamicpolicy/cpueviction/strategy/pressure_suppression_test.go

@@ -43,7 +43,7 @@ import (
 
 const (
 	defaultCPUMaxSuppressionToleranceRate     = 5.0
-	defaultCPUMinSuppressionToleranceDuration = 1 * time.Second
+	defaultCPUMinSuppressionToleranceDuration = 10 * time.Millisecond
 )
 
 func makeSuppressionEvictionConf(cpuMaxSuppressionToleranceRate float64,
@@ -56,6 +56,8 @@ func makeSuppressionEvictionConf(cpuMaxSuppressionToleranceRate float64,
 }
 
 func TestNewCPUPressureSuppressionEviction(t *testing.T) {
+	t.Parallel()
+
 	as := require.New(t)
 
 	cpuTopology, err := machine.GenerateDummyCPUTopology(16, 2, 4)
@@ -71,6 +73,8 @@ func TestNewCPUPressureSuppressionEviction(t *testing.T) {
 }
 
 func TestCPUPressureSuppression_GetEvictPods(t *testing.T) {
+	t.Parallel()
+
 	as := require.New(t)
 
 	cpuTopology, err := machine.GenerateDummyCPUTopology(16, 2, 4)
@@ -304,7 +308,7 @@ func TestCPUPressureSuppression_GetEvictPods(t *testing.T) {
 			assert.NoError(t, err)
 			assert.NotNil(t, resp)
 
-			time.Sleep(1 * time.Second)
+			time.Sleep(defaultCPUMinSuppressionToleranceDuration)
 
 			resp, err = plugin.GetEvictPods(context.TODO(), &evictionpluginapi.GetEvictPodsRequest{
 				ActivePods: pods,

pkg/agent/qrm-plugins/cpu/dynamicpolicy/policy.go

@@ -65,10 +65,6 @@ const (
 	syncCPUIdlePeriod = 30 * time.Second
 )
 
-var (
-	transitionPeriod = 30 * time.Second
-)
-
 var (
 	readonlyStateLock sync.RWMutex
 	readonlyState     state.ReadonlyState
@@ -125,6 +121,7 @@ type DynamicPolicy struct {
 	qosConfig                     *generic.QoSConfiguration
 	dynamicConfig                 *dynamicconfig.DynamicAgentConfiguration
 	podDebugAnnoKeys              []string
+	transitionPeriod              time.Duration
 }
 
 func NewDynamicPolicy(agentCtx *agent.GenericContext, conf *config.Configuration,
@@ -196,6 +193,7 @@ func NewDynamicPolicy(agentCtx *agent.GenericContext, conf *config.Configuration
 		enableCPUIdle:                 conf.CPUQRMPluginConfig.EnableCPUIdle,
 		reclaimRelativeRootCgroupPath: conf.ReclaimRelativeRootCgroupPath,
 		podDebugAnnoKeys:              conf.PodDebugAnnoKeys,
+		transitionPeriod:              time.Second,
 	}
 
 	// register allocation behaviors for pods with different QoS level
@@ -212,7 +210,7 @@ func NewDynamicPolicy(agentCtx *agent.GenericContext, conf *config.Configuration
 		consts.PodAnnotationQoSLevelReclaimedCores: policyImplement.reclaimedCoresHintHandler,
 	}
 
-	state.GetContainerRequestedCores = policyImplement.getContainerRequestedCores
+	state.SetContainerRequestedCores(policyImplement.getContainerRequestedCores)
 
 	if err := policyImplement.cleanPools(); err != nil {
 		return false, agent.ComponentStub{}, fmt.Errorf("cleanPools failed with error: %v", err)
@@ -324,7 +322,7 @@ func (p *DynamicPolicy) Start() (err error) {
 		general.Infof("sync existing containers to cpu advisor successfully")
 
 		// call lw of CPUAdvisorServer and do allocation
-		if err := p.lwCPUAdvisorServer(p.stopCh); err != nil {
+		if err = p.lwCPUAdvisorServer(p.stopCh); err != nil {
 			general.Errorf("lwCPUAdvisorServer failed with error: %v", err)
 		} else {
 			general.Infof("lwCPUAdvisorServer finished")
@@ -420,7 +418,7 @@ func (p *DynamicPolicy) GetResourcesAllocation(_ context.Context,
 
 				allocationInfo.InitTimestamp = time.Now().Format(util.QRMTimeFormat)
 				p.state.SetAllocationInfo(podUID, containerName, allocationInfo)
-			} else if allocationInfo.RampUp && time.Now().After(initTs.Add(transitionPeriod)) {
+			} else if allocationInfo.RampUp && time.Now().After(initTs.Add(p.transitionPeriod)) {
 				general.Infof("pod: %s/%s, container: %s ramp up finished", allocationInfo.PodNamespace, allocationInfo.PodName, allocationInfo.ContainerName)
 				allocationInfo.RampUp = false
 				p.state.SetAllocationInfo(podUID, containerName, allocationInfo)

pkg/agent/qrm-plugins/cpu/dynamicpolicy/policy_advisor_handler.go

@@ -521,7 +521,7 @@ func (p *DynamicPolicy) applyBlocks(blockCPUSet advisorapi.BlockCPUSet, resp *ad
 
 			// adapt to old checkpoint without RequestQuantity property
 			if newEntries[podUID][containerName] != nil {
-				newEntries[podUID][containerName].RequestQuantity = state.GetContainerRequestedCores(allocationInfo)
+				newEntries[podUID][containerName].RequestQuantity = state.GetContainerRequestedCores()(allocationInfo)
 				continue
 			}