seaweedfs/weed/wdclient/masterclient.go

328 lines
11 KiB
Go
Raw Normal View History

2018-07-28 09:10:32 +00:00
package wdclient
import (
"context"
2022-07-15 10:29:15 +00:00
"fmt"
2019-02-18 20:11:52 +00:00
"math/rand"
"sync"
2018-07-29 04:02:56 +00:00
"time"
2018-07-28 09:10:32 +00:00
"github.com/seaweedfs/seaweedfs/weed/stats"
"github.com/seaweedfs/seaweedfs/weed/util"
"google.golang.org/grpc"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/pb"
"github.com/seaweedfs/seaweedfs/weed/pb/master_pb"
)
2018-07-28 09:10:32 +00:00
type MasterClient struct {
2022-08-30 07:07:15 +00:00
FilerGroup string
clientType string
clientHost pb.ServerAddress
rack string
currentMaster pb.ServerAddress
currentMasterLock sync.RWMutex
masters map[string]pb.ServerAddress
grpcDialOption grpc.DialOption
*vidMap
vidMapCacheSize int
2022-08-26 17:23:42 +00:00
OnPeerUpdate func(update *master_pb.ClusterNodeUpdate, startFrom time.Time)
OnPeerUpdateLock sync.RWMutex
2018-07-28 09:10:32 +00:00
}
2022-07-03 07:29:25 +00:00
func NewMasterClient(grpcDialOption grpc.DialOption, filerGroup string, clientType string, clientHost pb.ServerAddress, clientDataCenter string, rack string, masters map[string]pb.ServerAddress) *MasterClient {
2018-07-28 09:10:32 +00:00
return &MasterClient{
FilerGroup: filerGroup,
clientType: clientType,
clientHost: clientHost,
2022-07-29 08:15:19 +00:00
rack: rack,
masters: masters,
grpcDialOption: grpcDialOption,
vidMap: newVidMap(clientDataCenter),
vidMapCacheSize: 5,
2018-07-28 09:10:32 +00:00
}
}
func (mc *MasterClient) SetOnPeerUpdateFn(onPeerUpdate func(update *master_pb.ClusterNodeUpdate, startFrom time.Time)) {
2022-08-26 17:23:42 +00:00
mc.OnPeerUpdateLock.Lock()
mc.OnPeerUpdate = onPeerUpdate
2022-08-26 17:23:42 +00:00
mc.OnPeerUpdateLock.Unlock()
}
func (mc *MasterClient) GetLookupFileIdFunction() LookupFileIdFunctionType {
return mc.LookupFileIdWithFallback
}
func (mc *MasterClient) LookupFileIdWithFallback(fileId string) (fullUrls []string, err error) {
fullUrls, err = mc.vidMap.LookupFileId(fileId)
2022-07-15 10:29:15 +00:00
if err == nil && len(fullUrls) > 0 {
2022-06-25 16:22:49 +00:00
return
}
err = pb.WithMasterClient(false, mc.GetMaster(), mc.grpcDialOption, false, func(client master_pb.SeaweedClient) error {
resp, err := client.LookupVolume(context.Background(), &master_pb.LookupVolumeRequest{
VolumeOrFileIds: []string{fileId},
})
if err != nil {
2022-08-30 16:43:17 +00:00
return fmt.Errorf("LookupVolume %s failed: %v", fileId, err)
}
for vid, vidLocation := range resp.VolumeIdLocations {
for _, vidLoc := range vidLocation.Locations {
loc := Location{
Url: vidLoc.Url,
PublicUrl: vidLoc.PublicUrl,
GrpcPort: int(vidLoc.GrpcPort),
DataCenter: vidLoc.DataCenter,
}
mc.vidMap.addLocation(uint32(vid), loc)
httpUrl := "http://" + loc.Url + "/" + fileId
// Prefer same data center
if mc.DataCenter != "" && mc.DataCenter == loc.DataCenter {
fullUrls = append([]string{httpUrl}, fullUrls...)
} else {
fullUrls = append(fullUrls, httpUrl)
}
}
}
return nil
})
return
}
func (mc *MasterClient) getCurrentMaster() pb.ServerAddress {
2022-08-30 07:07:15 +00:00
mc.currentMasterLock.RLock()
defer mc.currentMasterLock.RUnlock()
return mc.currentMaster
}
func (mc *MasterClient) setCurrentMaster(master pb.ServerAddress) {
2022-08-30 07:07:15 +00:00
mc.currentMasterLock.Lock()
mc.currentMaster = master
2022-08-30 07:07:15 +00:00
mc.currentMasterLock.Unlock()
}
func (mc *MasterClient) GetMaster() pb.ServerAddress {
2021-05-21 09:10:44 +00:00
mc.WaitUntilConnected()
return mc.getCurrentMaster()
}
2022-04-02 00:27:49 +00:00
func (mc *MasterClient) GetMasters() map[string]pb.ServerAddress {
mc.WaitUntilConnected()
return mc.masters
}
func (mc *MasterClient) WaitUntilConnected() {
for {
if mc.getCurrentMaster() != "" {
return
}
time.Sleep(time.Duration(rand.Int31n(200)) * time.Millisecond)
2023-03-26 19:22:48 +00:00
print(".")
}
}
func (mc *MasterClient) KeepConnectedToMaster() {
2022-05-02 04:59:16 +00:00
glog.V(1).Infof("%s.%s masterClient bootstraps with masters %v", mc.FilerGroup, mc.clientType, mc.masters)
for {
2018-07-28 09:10:32 +00:00
mc.tryAllMasters()
time.Sleep(time.Second)
}
}
func (mc *MasterClient) FindLeaderFromOtherPeers(myMasterAddress pb.ServerAddress) (leader string) {
for _, master := range mc.masters {
if master == myMasterAddress {
continue
}
if grpcErr := pb.WithMasterClient(false, master, mc.grpcDialOption, false, func(client master_pb.SeaweedClient) error {
ctx, cancel := context.WithTimeout(context.Background(), 120*time.Millisecond)
defer cancel()
resp, err := client.GetMasterConfiguration(ctx, &master_pb.GetMasterConfigurationRequest{})
if err != nil {
return err
}
leader = resp.Leader
return nil
}); grpcErr != nil {
glog.V(0).Infof("connect to %s: %v", master, grpcErr)
}
if leader != "" {
glog.V(0).Infof("existing leader is %s", leader)
return
}
}
glog.V(0).Infof("No existing leader found!")
return
}
2018-07-28 09:10:32 +00:00
func (mc *MasterClient) tryAllMasters() {
var nextHintedLeader pb.ServerAddress
2018-07-28 09:10:32 +00:00
for _, master := range mc.masters {
nextHintedLeader = mc.tryConnectToMaster(master)
for nextHintedLeader != "" {
nextHintedLeader = mc.tryConnectToMaster(nextHintedLeader)
}
mc.setCurrentMaster("")
}
}
func (mc *MasterClient) tryConnectToMaster(master pb.ServerAddress) (nextHintedLeader pb.ServerAddress) {
2022-05-02 04:59:16 +00:00
glog.V(1).Infof("%s.%s masterClient Connecting to master %v", mc.FilerGroup, mc.clientType, master)
2022-01-24 14:09:43 +00:00
stats.MasterClientConnectCounter.WithLabelValues("total").Inc()
gprcErr := pb.WithMasterClient(true, master, mc.grpcDialOption, false, func(client master_pb.SeaweedClient) error {
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
stream, err := client.KeepConnected(ctx)
if err != nil {
2022-05-02 04:59:16 +00:00
glog.V(1).Infof("%s.%s masterClient failed to keep connected to %s: %v", mc.FilerGroup, mc.clientType, master, err)
stats.MasterClientConnectCounter.WithLabelValues(stats.FailedToKeepConnected).Inc()
return err
}
if err = stream.Send(&master_pb.KeepConnectedRequest{
2022-05-02 04:59:16 +00:00
FilerGroup: mc.FilerGroup,
2022-07-03 07:29:25 +00:00
DataCenter: mc.DataCenter,
Rack: mc.rack,
ClientType: mc.clientType,
ClientAddress: string(mc.clientHost),
Version: util.Version(),
}); err != nil {
2022-05-02 04:59:16 +00:00
glog.V(0).Infof("%s.%s masterClient failed to send to %s: %v", mc.FilerGroup, mc.clientType, master, err)
stats.MasterClientConnectCounter.WithLabelValues(stats.FailedToSend).Inc()
return err
}
2022-05-02 04:59:16 +00:00
glog.V(1).Infof("%s.%s masterClient Connected to %v", mc.FilerGroup, mc.clientType, master)
resp, err := stream.Recv()
if err != nil {
glog.V(0).Infof("%s.%s masterClient failed to receive from %s: %v", mc.FilerGroup, mc.clientType, master, err)
stats.MasterClientConnectCounter.WithLabelValues(stats.FailedToReceive).Inc()
return err
}
2022-06-25 03:56:09 +00:00
// check if it is the leader to determine whether to reset the vidMap
if resp.VolumeLocation != nil {
if resp.VolumeLocation.Leader != "" && string(master) != resp.VolumeLocation.Leader {
glog.V(0).Infof("master %v redirected to leader %v", master, resp.VolumeLocation.Leader)
nextHintedLeader = pb.ServerAddress(resp.VolumeLocation.Leader)
stats.MasterClientConnectCounter.WithLabelValues(stats.RedirectedToLeader).Inc()
return nil
}
mc.resetVidMap()
mc.updateVidMap(resp)
} else {
mc.resetVidMap()
}
mc.setCurrentMaster(master)
for {
2021-11-06 01:11:40 +00:00
resp, err := stream.Recv()
if err != nil {
2022-05-02 04:59:16 +00:00
glog.V(0).Infof("%s.%s masterClient failed to receive from %s: %v", mc.FilerGroup, mc.clientType, master, err)
stats.MasterClientConnectCounter.WithLabelValues(stats.FailedToReceive).Inc()
return err
}
2021-11-06 01:11:40 +00:00
if resp.VolumeLocation != nil {
// maybe the leader is changed
if resp.VolumeLocation.Leader != "" && string(mc.GetMaster()) != resp.VolumeLocation.Leader {
glog.V(0).Infof("currentMaster %v redirected to leader %v", mc.GetMaster(), resp.VolumeLocation.Leader)
2021-11-06 01:11:40 +00:00
nextHintedLeader = pb.ServerAddress(resp.VolumeLocation.Leader)
stats.MasterClientConnectCounter.WithLabelValues(stats.RedirectedToLeader).Inc()
2021-11-06 01:11:40 +00:00
return nil
}
mc.updateVidMap(resp)
}
2021-11-06 11:07:38 +00:00
if resp.ClusterNodeUpdate != nil {
update := resp.ClusterNodeUpdate
2022-08-26 17:23:42 +00:00
mc.OnPeerUpdateLock.RLock()
2021-11-06 21:23:35 +00:00
if mc.OnPeerUpdate != nil {
2022-05-02 04:59:16 +00:00
if update.FilerGroup == mc.FilerGroup {
if update.IsAdd {
2023-06-20 01:19:13 +00:00
glog.V(0).Infof("+ %s.%s %s\n", update.FilerGroup, update.NodeType, update.Address)
2022-05-02 04:59:16 +00:00
} else {
2023-06-20 01:19:13 +00:00
glog.V(0).Infof("- %s.%s %s\n", update.FilerGroup, update.NodeType, update.Address)
2022-05-02 04:59:16 +00:00
}
stats.MasterClientConnectCounter.WithLabelValues(stats.OnPeerUpdate).Inc()
mc.OnPeerUpdate(update, time.Now())
2021-11-06 21:23:35 +00:00
}
2021-11-06 11:07:38 +00:00
}
2022-08-26 17:23:42 +00:00
mc.OnPeerUpdateLock.RUnlock()
2021-11-06 11:07:38 +00:00
}
}
})
if gprcErr != nil {
stats.MasterClientConnectCounter.WithLabelValues(stats.Failed).Inc()
2022-05-02 04:59:16 +00:00
glog.V(1).Infof("%s.%s masterClient failed to connect with master %v: %v", mc.FilerGroup, mc.clientType, master, gprcErr)
}
return
}
func (mc *MasterClient) updateVidMap(resp *master_pb.KeepConnectedResponse) {
if resp.VolumeLocation.IsEmptyUrl() {
glog.V(0).Infof("updateVidMap ignore short heartbeat: %+v", resp)
return
}
// process new volume location
loc := Location{
Url: resp.VolumeLocation.Url,
PublicUrl: resp.VolumeLocation.PublicUrl,
DataCenter: resp.VolumeLocation.DataCenter,
GrpcPort: int(resp.VolumeLocation.GrpcPort),
}
for _, newVid := range resp.VolumeLocation.NewVids {
glog.V(2).Infof("%s.%s: %s masterClient adds volume %d", mc.FilerGroup, mc.clientType, loc.Url, newVid)
mc.addLocation(newVid, loc)
}
for _, deletedVid := range resp.VolumeLocation.DeletedVids {
glog.V(2).Infof("%s.%s: %s masterClient removes volume %d", mc.FilerGroup, mc.clientType, loc.Url, deletedVid)
mc.deleteLocation(deletedVid, loc)
}
for _, newEcVid := range resp.VolumeLocation.NewEcVids {
glog.V(2).Infof("%s.%s: %s masterClient adds ec volume %d", mc.FilerGroup, mc.clientType, loc.Url, newEcVid)
mc.addEcLocation(newEcVid, loc)
}
for _, deletedEcVid := range resp.VolumeLocation.DeletedEcVids {
glog.V(2).Infof("%s.%s: %s masterClient removes ec volume %d", mc.FilerGroup, mc.clientType, loc.Url, deletedEcVid)
mc.deleteEcLocation(deletedEcVid, loc)
}
glog.V(1).Infof("updateVidMap(%s) %s.%s: %s volume add: %d, del: %d, add ec: %d del ec: %d",
resp.VolumeLocation.DataCenter, mc.FilerGroup, mc.clientType, loc.Url,
len(resp.VolumeLocation.NewVids), len(resp.VolumeLocation.DeletedVids),
len(resp.VolumeLocation.NewEcVids), len(resp.VolumeLocation.DeletedEcVids))
}
func (mc *MasterClient) WithClient(streamingMode bool, fn func(client master_pb.SeaweedClient) error) error {
2020-11-01 10:36:43 +00:00
return util.Retry("master grpc", func() error {
return pb.WithMasterClient(streamingMode, mc.GetMaster(), mc.grpcDialOption, false, func(client master_pb.SeaweedClient) error {
2020-11-01 08:40:16 +00:00
return fn(client)
})
2019-03-19 12:19:37 +00:00
})
}
func (mc *MasterClient) resetVidMap() {
tail := &vidMap{
vid2Locations: mc.vid2Locations,
ecVid2Locations: mc.ecVid2Locations,
DataCenter: mc.DataCenter,
cache: mc.cache,
}
nvm := newVidMap(mc.DataCenter)
nvm.cache = tail
mc.vidMap = nvm
//trim
for i := 0; i < mc.vidMapCacheSize && tail.cache != nil; i++ {
if i == mc.vidMapCacheSize-1 {
tail.cache = nil
} else {
tail = tail.cache
}
}
}