Merge branch 'main' into issue-fix-6693

This commit is contained in:
Lyndon-Li
2023-11-10 10:04:27 +08:00
27 changed files with 1140 additions and 51 deletions
+76 -4
View File
@@ -19,6 +19,7 @@ package nodeagent
import (
"context"
"fmt"
"math"
"net/http"
"os"
"strings"
@@ -32,6 +33,7 @@ import (
storagev1api "k8s.io/api/storage/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/fields"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/apimachinery/pkg/runtime"
"k8s.io/apimachinery/pkg/types"
"k8s.io/apimachinery/pkg/util/sets"
@@ -53,7 +55,9 @@ import (
"github.com/vmware-tanzu/velero/pkg/cmd"
"github.com/vmware-tanzu/velero/pkg/cmd/util/signals"
"github.com/vmware-tanzu/velero/pkg/controller"
"github.com/vmware-tanzu/velero/pkg/datapath"
"github.com/vmware-tanzu/velero/pkg/metrics"
"github.com/vmware-tanzu/velero/pkg/nodeagent"
"github.com/vmware-tanzu/velero/pkg/repository"
"github.com/vmware-tanzu/velero/pkg/util/filesystem"
"github.com/vmware-tanzu/velero/pkg/util/logging"
@@ -73,6 +77,7 @@ const (
defaultResourceTimeout = 10 * time.Minute
defaultDataMoverPrepareTimeout = 30 * time.Minute
defaultDataPathConcurrentNum = 1
)
type nodeAgentServerConfig struct {
@@ -132,6 +137,7 @@ type nodeAgentServer struct {
config nodeAgentServerConfig
kubeClient kubernetes.Interface
csiSnapshotClient *snapshotv1client.Clientset
dataPathMgr *datapath.Manager
}
func newNodeAgentServer(logger logrus.FieldLogger, factory client.Factory, config nodeAgentServerConfig) (*nodeAgentServer, error) {
@@ -219,6 +225,10 @@ func newNodeAgentServer(logger logrus.FieldLogger, factory client.Factory, confi
if err != nil {
return nil, err
}
dataPathConcurrentNum := s.getDataPathConcurrentNum(defaultDataPathConcurrentNum)
s.dataPathMgr = datapath.NewManager(dataPathConcurrentNum)
return s, nil
}
@@ -263,24 +273,24 @@ func (s *nodeAgentServer) run() {
credentialGetter := &credentials.CredentialGetter{FromFile: credentialFileStore, FromSecret: credSecretStore}
repoEnsurer := repository.NewEnsurer(s.mgr.GetClient(), s.logger, s.config.resourceTimeout)
pvbReconciler := controller.NewPodVolumeBackupReconciler(s.mgr.GetClient(), repoEnsurer,
pvbReconciler := controller.NewPodVolumeBackupReconciler(s.mgr.GetClient(), s.dataPathMgr, repoEnsurer,
credentialGetter, s.nodeName, s.mgr.GetScheme(), s.metrics, s.logger)
if err := pvbReconciler.SetupWithManager(s.mgr); err != nil {
s.logger.Fatal(err, "unable to create controller", "controller", controller.PodVolumeBackup)
}
if err = controller.NewPodVolumeRestoreReconciler(s.mgr.GetClient(), repoEnsurer, credentialGetter, s.logger).SetupWithManager(s.mgr); err != nil {
if err = controller.NewPodVolumeRestoreReconciler(s.mgr.GetClient(), s.dataPathMgr, repoEnsurer, credentialGetter, s.logger).SetupWithManager(s.mgr); err != nil {
s.logger.WithError(err).Fatal("Unable to create the pod volume restore controller")
}
dataUploadReconciler := controller.NewDataUploadReconciler(s.mgr.GetClient(), s.kubeClient, s.csiSnapshotClient.SnapshotV1(), repoEnsurer, clock.RealClock{}, credentialGetter, s.nodeName, s.fileSystem, s.config.dataMoverPrepareTimeout, s.logger, s.metrics)
dataUploadReconciler := controller.NewDataUploadReconciler(s.mgr.GetClient(), s.kubeClient, s.csiSnapshotClient.SnapshotV1(), s.dataPathMgr, repoEnsurer, clock.RealClock{}, credentialGetter, s.nodeName, s.fileSystem, s.config.dataMoverPrepareTimeout, s.logger, s.metrics)
s.markDataUploadsCancel(dataUploadReconciler)
if err = dataUploadReconciler.SetupWithManager(s.mgr); err != nil {
s.logger.WithError(err).Fatal("Unable to create the data upload controller")
}
dataDownloadReconciler := controller.NewDataDownloadReconciler(s.mgr.GetClient(), s.kubeClient, repoEnsurer, credentialGetter, s.nodeName, s.config.dataMoverPrepareTimeout, s.logger, s.metrics)
dataDownloadReconciler := controller.NewDataDownloadReconciler(s.mgr.GetClient(), s.kubeClient, s.dataPathMgr, repoEnsurer, credentialGetter, s.nodeName, s.config.dataMoverPrepareTimeout, s.logger, s.metrics)
s.markDataDownloadsCancel(dataDownloadReconciler)
if err = dataDownloadReconciler.SetupWithManager(s.mgr); err != nil {
s.logger.WithError(err).Fatal("Unable to create the data download controller")
@@ -478,3 +488,65 @@ func (s *nodeAgentServer) markInProgressPVRsFailed(client ctrlclient.Client) {
s.logger.WithField("podvolumerestore", pvr.GetName()).Warn(pvr.Status.Message)
}
}
var getConfigsFunc = nodeagent.GetConfigs
func (s *nodeAgentServer) getDataPathConcurrentNum(defaultNum int) int {
configs, err := getConfigsFunc(s.ctx, s.namespace, s.kubeClient)
if err != nil {
s.logger.WithError(err).Warn("Failed to get node agent configs")
return defaultNum
}
if configs == nil || configs.DataPathConcurrency == nil {
s.logger.Infof("Concurrency configs are not found, use the default number %v", defaultNum)
return defaultNum
}
globalNum := configs.DataPathConcurrency.GlobalConfig
if globalNum <= 0 {
s.logger.Warnf("Global number %v is invalid, use the default value %v", globalNum, defaultNum)
globalNum = defaultNum
}
if len(configs.DataPathConcurrency.PerNodeConfig) == 0 {
return globalNum
}
curNode, err := s.kubeClient.CoreV1().Nodes().Get(s.ctx, s.nodeName, metav1.GetOptions{})
if err != nil {
s.logger.WithError(err).Warnf("Failed to get node info for %s, use the global number %v", s.nodeName, globalNum)
return globalNum
}
concurrentNum := math.MaxInt32
for _, rule := range configs.DataPathConcurrency.PerNodeConfig {
selector, err := metav1.LabelSelectorAsSelector(&rule.NodeSelector)
if err != nil {
s.logger.WithError(err).Warnf("Failed to parse rule with label selector %s, skip it", rule.NodeSelector.String())
continue
}
if rule.Number <= 0 {
s.logger.Warnf("Rule with label selector %s is with an invalid number %v, skip it", rule.NodeSelector.String(), rule.Number)
continue
}
if selector.Matches(labels.Set(curNode.GetLabels())) {
if concurrentNum > rule.Number {
concurrentNum = rule.Number
}
}
}
if concurrentNum == math.MaxInt32 {
s.logger.Infof("Per node number for node %s is not found, use the global number %v", s.nodeName, globalNum)
concurrentNum = globalNum
} else {
s.logger.Infof("Use the per node number %v over global number %v for node %s", concurrentNum, globalNum, s.nodeName)
}
return concurrentNum
}
+262
View File
@@ -17,16 +17,22 @@ package nodeagent
import (
"context"
"fmt"
"os"
"path/filepath"
"strings"
"testing"
"github.com/pkg/errors"
"github.com/stretchr/testify/assert"
corev1 "k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/runtime"
"k8s.io/client-go/kubernetes"
"k8s.io/client-go/kubernetes/fake"
"github.com/vmware-tanzu/velero/pkg/builder"
"github.com/vmware-tanzu/velero/pkg/nodeagent"
testutil "github.com/vmware-tanzu/velero/pkg/test"
)
@@ -107,3 +113,259 @@ func Test_validatePodVolumesHostPath(t *testing.T) {
})
}
}
func Test_getDataPathConcurrentNum(t *testing.T) {
defaultNum := 100001
globalNum := 6
nodeName := "node-agent-node"
node1 := builder.ForNode("node-agent-node").Result()
node2 := builder.ForNode("node-agent-node").Labels(map[string]string{
"host-name": "node-1",
"xxxx": "yyyyy",
}).Result()
invalidLabelSelector := metav1.LabelSelector{
MatchLabels: map[string]string{
"inva/lid": "inva/lid",
},
}
validLabelSelector1 := metav1.LabelSelector{
MatchLabels: map[string]string{
"host-name": "node-1",
},
}
validLabelSelector2 := metav1.LabelSelector{
MatchLabels: map[string]string{
"xxxx": "yyyyy",
},
}
tests := []struct {
name string
getFunc func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error)
setKubeClient bool
kubeClientObj []runtime.Object
expectNum int
expectLog string
}{
{
name: "failed to get configs",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return nil, errors.New("fake-get-error")
},
expectLog: "Failed to get node agent configs",
expectNum: defaultNum,
},
{
name: "configs cm not found",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return nil, nil
},
expectLog: fmt.Sprintf("Concurrency configs are not found, use the default number %v", defaultNum),
expectNum: defaultNum,
},
{
name: "configs cm's data path concurrency is nil",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{}, nil
},
expectLog: fmt.Sprintf("Concurrency configs are not found, use the default number %v", defaultNum),
expectNum: defaultNum,
},
{
name: "global number is invalid",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: -1,
},
}, nil
},
expectLog: fmt.Sprintf("Global number %v is invalid, use the default value %v", -1, defaultNum),
expectNum: defaultNum,
},
{
name: "global number is valid",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
},
}, nil
},
expectNum: globalNum,
},
{
name: "node is not found",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
PerNodeConfig: []nodeagent.RuledConfigs{
{
Number: 100,
},
},
},
}, nil
},
setKubeClient: true,
expectLog: fmt.Sprintf("Failed to get node info for %s, use the global number %v", nodeName, globalNum),
expectNum: globalNum,
},
{
name: "failed to get selector",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
PerNodeConfig: []nodeagent.RuledConfigs{
{
NodeSelector: invalidLabelSelector,
Number: 100,
},
},
},
}, nil
},
setKubeClient: true,
kubeClientObj: []runtime.Object{node1},
expectLog: fmt.Sprintf("Failed to parse rule with label selector %s, skip it", invalidLabelSelector.String()),
expectNum: globalNum,
},
{
name: "rule number is invalid",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
PerNodeConfig: []nodeagent.RuledConfigs{
{
NodeSelector: validLabelSelector1,
Number: -1,
},
},
},
}, nil
},
setKubeClient: true,
kubeClientObj: []runtime.Object{node1},
expectLog: fmt.Sprintf("Rule with label selector %s is with an invalid number %v, skip it", validLabelSelector1.String(), -1),
expectNum: globalNum,
},
{
name: "label doesn't match",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
PerNodeConfig: []nodeagent.RuledConfigs{
{
NodeSelector: validLabelSelector1,
Number: -1,
},
},
},
}, nil
},
setKubeClient: true,
kubeClientObj: []runtime.Object{node1},
expectLog: fmt.Sprintf("Per node number for node %s is not found, use the global number %v", nodeName, globalNum),
expectNum: globalNum,
},
{
name: "match one rule",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
PerNodeConfig: []nodeagent.RuledConfigs{
{
NodeSelector: validLabelSelector1,
Number: 66,
},
},
},
}, nil
},
setKubeClient: true,
kubeClientObj: []runtime.Object{node2},
expectLog: fmt.Sprintf("Use the per node number %v over global number %v for node %s", 66, globalNum, nodeName),
expectNum: 66,
},
{
name: "match multiple rules",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
PerNodeConfig: []nodeagent.RuledConfigs{
{
NodeSelector: validLabelSelector1,
Number: 66,
},
{
NodeSelector: validLabelSelector2,
Number: 36,
},
},
},
}, nil
},
setKubeClient: true,
kubeClientObj: []runtime.Object{node2},
expectLog: fmt.Sprintf("Use the per node number %v over global number %v for node %s", 36, globalNum, nodeName),
expectNum: 36,
},
{
name: "match multiple rules 2",
getFunc: func(context.Context, string, kubernetes.Interface) (*nodeagent.Configs, error) {
return &nodeagent.Configs{
DataPathConcurrency: &nodeagent.DataPathConcurrency{
GlobalConfig: globalNum,
PerNodeConfig: []nodeagent.RuledConfigs{
{
NodeSelector: validLabelSelector1,
Number: 36,
},
{
NodeSelector: validLabelSelector2,
Number: 66,
},
},
},
}, nil
},
setKubeClient: true,
kubeClientObj: []runtime.Object{node2},
expectLog: fmt.Sprintf("Use the per node number %v over global number %v for node %s", 36, globalNum, nodeName),
expectNum: 36,
},
}
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
fakeKubeClient := fake.NewSimpleClientset(test.kubeClientObj...)
logBuffer := ""
s := &nodeAgentServer{
nodeName: nodeName,
logger: testutil.NewSingleLogger(&logBuffer),
}
if test.setKubeClient {
s.kubeClient = fakeKubeClient
}
getConfigsFunc = test.getFunc
num := s.getDataPathConcurrentNum(defaultNum)
assert.Equal(t, test.expectNum, num)
if test.expectLog == "" {
assert.Equal(t, "", logBuffer)
} else {
assert.True(t, strings.Contains(logBuffer, test.expectLog))
}
})
}
}