storj/satellite/orders/service.go

581 lines
21 KiB
Go
Raw Normal View History

2019-03-28 20:09:23 +00:00
// Copyright (C) 2019 Storj Labs, Inc.
// See LICENSE for copying information.
package orders
import (
"context"
[V3-1927] Repairer uploads to max threshold instead of success… (#2423) * pkg/datarepair: Add test to check num upload pieces Add a new test for ensuring the number of pieces that the repair process upload when a segment is injured. * satellite/orders: Don't create "put order limits" over total Repair must not create "put order limits" more than the total count. * pkg/datarepair: Update upload repair pieces test Update the test which checks the number of pieces which are uploaded during a repair for using the same excess over the success threshold value than the implementation. * satellites/orders: Limit repair put order for not being total Limit the number of put orders to be used by repair for only uploading pieces to a % excess over the successful threshold. * pkg/datarepair: Change DataRepair test to pass again Make some changes in the DataRepair test to make pass again after the repair upload repaired pieces only until a % excess over success threshold. Also update the steps description of the DataRepair test after it has been changed, to match on what's now, besides to leave it more generic for avoiding having to update it on minimal future refactorings. * satellite: Make repair excess optimal threshold configurable Add a new configuration parameter to the satellite for being able to configure the percentage excess over the optimal threshold, used for determining how many pieces should be repaired/uploaded, rather than having the value hard coded. * repairer: Add configurable param to segments/repairer Add a new parameters to the segment/repairer to calculate the maximum number of excess nodes, based on the optimal threshold, that repaired pieces can be uploaded. This new parameter has been added for not returning more nodes than the number of upload orders for data repair satellite service calculate for repairing pieces. * pkg/storage/ec: Update log message in clien.Repair * satellite: Update configuration lock file
2019-07-11 23:44:47 +01:00
"math"
mathrand "math/rand"
"sync"
2019-03-28 20:09:23 +00:00
"time"
"github.com/zeebo/errs"
"go.uber.org/zap"
"storj.io/common/pb"
"storj.io/common/signing"
"storj.io/common/storj"
"storj.io/common/uuid"
"storj.io/storj/satellite/metainfo/metabase"
"storj.io/storj/satellite/overlay"
"storj.io/uplink/private/eestream"
2019-03-28 20:09:23 +00:00
)
// ErrDownloadFailedNotEnoughPieces is returned when download failed due to missing pieces.
var ErrDownloadFailedNotEnoughPieces = errs.Class("not enough pieces for download")
// Config is a configuration struct for orders Service.
type Config struct {
Expiration time.Duration `help:"how long until an order expires" default:"48h"` // 2 days
SettlementBatchSize int `help:"how many orders to batch per transaction" default:"250"`
FlushBatchSize int `help:"how many items in the rollups write cache before they are flushed to the database" devDefault:"20" releaseDefault:"10000"`
FlushInterval time.Duration `help:"how often to flush the rollups write cache to the database" devDefault:"30s" releaseDefault:"1m"`
ReportedRollupsReadBatchSize int `help:"how many records to read in a single transaction when calculating billable bandwidth" default:"1000"`
NodeStatusLogging bool `hidden:"true" help:"deprecated, log the offline/disqualification status of nodes" default:"false"`
WindowEndpointRolloutPhase WindowEndpointRolloutPhase `help:"rollout phase for the windowed endpoint" default:"phase1"`
}
// BucketsDB returns information about buckets.
type BucketsDB interface {
// GetBucketID returns an existing bucket id.
GetBucketID(ctx context.Context, bucketName []byte, projectID uuid.UUID) (id uuid.UUID, err error)
}
2019-03-28 20:09:23 +00:00
// Service for creating order limits.
2019-09-10 14:24:16 +01:00
//
// architecture: Service
2019-03-28 20:09:23 +00:00
type Service struct {
log *zap.Logger
satellite signing.Signer
overlay *overlay.Service
orders DB
buckets BucketsDB
satelliteAddress *pb.NodeAddress
orderExpiration time.Duration
rngMu sync.Mutex
rng *mathrand.Rand
2019-03-28 20:09:23 +00:00
}
// NewService creates new service for creating order limits.
[V3-1927] Repairer uploads to max threshold instead of success… (#2423) * pkg/datarepair: Add test to check num upload pieces Add a new test for ensuring the number of pieces that the repair process upload when a segment is injured. * satellite/orders: Don't create "put order limits" over total Repair must not create "put order limits" more than the total count. * pkg/datarepair: Update upload repair pieces test Update the test which checks the number of pieces which are uploaded during a repair for using the same excess over the success threshold value than the implementation. * satellites/orders: Limit repair put order for not being total Limit the number of put orders to be used by repair for only uploading pieces to a % excess over the successful threshold. * pkg/datarepair: Change DataRepair test to pass again Make some changes in the DataRepair test to make pass again after the repair upload repaired pieces only until a % excess over success threshold. Also update the steps description of the DataRepair test after it has been changed, to match on what's now, besides to leave it more generic for avoiding having to update it on minimal future refactorings. * satellite: Make repair excess optimal threshold configurable Add a new configuration parameter to the satellite for being able to configure the percentage excess over the optimal threshold, used for determining how many pieces should be repaired/uploaded, rather than having the value hard coded. * repairer: Add configurable param to segments/repairer Add a new parameters to the segment/repairer to calculate the maximum number of excess nodes, based on the optimal threshold, that repaired pieces can be uploaded. This new parameter has been added for not returning more nodes than the number of upload orders for data repair satellite service calculate for repairing pieces. * pkg/storage/ec: Update log message in clien.Repair * satellite: Update configuration lock file
2019-07-11 23:44:47 +01:00
func NewService(
log *zap.Logger, satellite signing.Signer, overlay *overlay.Service,
orders DB, buckets BucketsDB,
orderExpiration time.Duration, satelliteAddress *pb.NodeAddress,
[V3-1927] Repairer uploads to max threshold instead of success… (#2423) * pkg/datarepair: Add test to check num upload pieces Add a new test for ensuring the number of pieces that the repair process upload when a segment is injured. * satellite/orders: Don't create "put order limits" over total Repair must not create "put order limits" more than the total count. * pkg/datarepair: Update upload repair pieces test Update the test which checks the number of pieces which are uploaded during a repair for using the same excess over the success threshold value than the implementation. * satellites/orders: Limit repair put order for not being total Limit the number of put orders to be used by repair for only uploading pieces to a % excess over the successful threshold. * pkg/datarepair: Change DataRepair test to pass again Make some changes in the DataRepair test to make pass again after the repair upload repaired pieces only until a % excess over success threshold. Also update the steps description of the DataRepair test after it has been changed, to match on what's now, besides to leave it more generic for avoiding having to update it on minimal future refactorings. * satellite: Make repair excess optimal threshold configurable Add a new configuration parameter to the satellite for being able to configure the percentage excess over the optimal threshold, used for determining how many pieces should be repaired/uploaded, rather than having the value hard coded. * repairer: Add configurable param to segments/repairer Add a new parameters to the segment/repairer to calculate the maximum number of excess nodes, based on the optimal threshold, that repaired pieces can be uploaded. This new parameter has been added for not returning more nodes than the number of upload orders for data repair satellite service calculate for repairing pieces. * pkg/storage/ec: Update log message in clien.Repair * satellite: Update configuration lock file
2019-07-11 23:44:47 +01:00
) *Service {
2019-03-28 20:09:23 +00:00
return &Service{
log: log,
satellite: satellite,
overlay: overlay,
orders: orders,
buckets: buckets,
satelliteAddress: satelliteAddress,
orderExpiration: orderExpiration,
rng: mathrand.New(mathrand.NewSource(time.Now().UnixNano())),
2019-03-28 20:09:23 +00:00
}
}
// VerifyOrderLimitSignature verifies that the signature inside order limit belongs to the satellite.
func (service *Service) VerifyOrderLimitSignature(ctx context.Context, signed *pb.OrderLimit) (err error) {
defer mon.Task()(&ctx)(&err)
return signing.VerifyOrderLimitSignature(ctx, service.satellite, signed)
2019-03-28 20:09:23 +00:00
}
func (service *Service) saveSerial(ctx context.Context, serialNumber storj.SerialNumber, bucket metabase.BucketLocation, expiresAt time.Time) (err error) {
defer mon.Task()(&ctx)(&err)
return service.orders.CreateSerialInfo(ctx, serialNumber, []byte(bucket.Prefix()), expiresAt)
2019-03-28 20:09:23 +00:00
}
func (service *Service) updateBandwidth(ctx context.Context, bucket metabase.BucketLocation, addressedOrderLimits ...*pb.AddressedOrderLimit) (err error) {
defer mon.Task()(&ctx)(&err)
if len(addressedOrderLimits) == 0 {
return nil
}
var action pb.PieceAction
var bucketAllocation int64
for _, addressedOrderLimit := range addressedOrderLimits {
if addressedOrderLimit != nil && addressedOrderLimit.Limit != nil {
orderLimit := addressedOrderLimit.Limit
action = orderLimit.Action
bucketAllocation += orderLimit.Limit
}
}
now := time.Now().UTC()
intervalStart := time.Date(now.Year(), now.Month(), now.Day(), now.Hour(), 0, 0, 0, now.Location())
// TODO: all of this below should be a single db transaction. in fact, this whole function should probably be part of an existing transaction
if err := service.orders.UpdateBucketBandwidthAllocation(ctx, bucket.ProjectID, []byte(bucket.BucketName), action, bucketAllocation, intervalStart); err != nil {
return Error.Wrap(err)
}
return nil
}
2019-03-28 20:09:23 +00:00
// CreateGetOrderLimits creates the order limits for downloading the pieces of pointer.
func (service *Service) CreateGetOrderLimits(ctx context.Context, bucket metabase.BucketLocation, pointer *pb.Pointer) (_ []*pb.AddressedOrderLimit, privateKey storj.PiecePrivateKey, err error) {
defer mon.Task()(&ctx)(&err)
2019-03-28 20:09:23 +00:00
redundancy, err := eestream.NewRedundancyStrategyFromProto(pointer.GetRemote().GetRedundancy())
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
pieceSize := eestream.CalcPieceSize(pointer.GetSegmentSize(), redundancy)
nodeIDs := make([]storj.NodeID, len(pointer.GetRemote().GetRemotePieces()))
for i, piece := range pointer.GetRemote().GetRemotePieces() {
nodeIDs[i] = piece.NodeId
}
nodes, err := service.overlay.GetOnlineNodesForGetDelete(ctx, nodeIDs)
if err != nil {
service.log.Debug("error getting nodes from overlay", zap.Error(err))
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
signer, err := NewSignerGet(service, pointer.GetRemote().RootPieceId, time.Now(), pieceSize)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
neededLimits := pb.NewRedundancySchemeToStorj(pointer.GetRemote().GetRedundancy()).DownloadNodes()
pieces := pointer.GetRemote().GetRemotePieces()
for _, pieceIndex := range service.perm(len(pieces)) {
piece := pieces[pieceIndex]
node, ok := nodes[piece.NodeId]
if !ok {
continue
}
address := node.Address.Address
if node.LastIPPort != "" {
address = node.LastIPPort
2019-03-28 20:09:23 +00:00
}
_, err := signer.Sign(ctx, storj.NodeURL{
ID: piece.NodeId,
Address: address,
}, piece.PieceNum)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
2019-03-28 20:09:23 +00:00
if len(signer.AddressedLimits) >= int(neededLimits) {
break
}
}
if len(signer.AddressedLimits) < redundancy.RequiredCount() {
mon.Meter("download_failed_not_enough_pieces_uplink").Mark(1) //locked
return nil, storj.PiecePrivateKey{}, ErrDownloadFailedNotEnoughPieces.New("not enough orderlimits: got %d, required %d", len(signer.AddressedLimits), redundancy.RequiredCount())
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if err := service.updateBandwidth(ctx, bucket, signer.AddressedLimits...); err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return signer.AddressedLimits, signer.PrivateKey, nil
2019-03-28 20:09:23 +00:00
}
func (service *Service) perm(n int) []int {
service.rngMu.Lock()
defer service.rngMu.Unlock()
return service.rng.Perm(n)
}
2019-03-28 20:09:23 +00:00
// CreatePutOrderLimits creates the order limits for uploading pieces to nodes.
func (service *Service) CreatePutOrderLimits(ctx context.Context, bucket metabase.BucketLocation, nodes []*overlay.SelectedNode, pieceExpiration time.Time, maxPieceSize int64) (_ storj.PieceID, _ []*pb.AddressedOrderLimit, privateKey storj.PiecePrivateKey, err error) {
defer mon.Task()(&ctx)(&err)
signer, err := NewSignerPut(service, pieceExpiration, time.Now(), maxPieceSize)
2019-03-28 20:09:23 +00:00
if err != nil {
return storj.PieceID{}, nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
for pieceNum, node := range nodes {
address := node.Address.Address
if node.LastIPPort != "" {
address = node.LastIPPort
}
_, err := signer.Sign(ctx, storj.NodeURL{ID: node.ID, Address: address}, int32(pieceNum))
if err != nil {
return storj.PieceID{}, nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return storj.PieceID{}, nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if err := service.updateBandwidth(ctx, bucket, signer.AddressedLimits...); err != nil {
return storj.PieceID{}, nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return signer.RootPieceID, signer.AddressedLimits, signer.PrivateKey, nil
2019-03-28 20:09:23 +00:00
}
// CreateDeleteOrderLimits creates the order limits for deleting the pieces of pointer.
func (service *Service) CreateDeleteOrderLimits(ctx context.Context, bucket metabase.BucketLocation, pointer *pb.Pointer) (_ []*pb.AddressedOrderLimit, _ storj.PiecePrivateKey, err error) {
defer mon.Task()(&ctx)(&err)
nodeIDs := make([]storj.NodeID, len(pointer.GetRemote().GetRemotePieces()))
for i, piece := range pointer.GetRemote().GetRemotePieces() {
nodeIDs[i] = piece.NodeId
}
nodes, err := service.overlay.GetOnlineNodesForGetDelete(ctx, nodeIDs)
if err != nil {
service.log.Debug("error getting nodes from overlay", zap.Error(err))
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
signer, err := NewSignerDelete(service, pointer.GetRemote().RootPieceId, time.Now())
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
var nodeErrors errs.Group
2019-03-28 20:09:23 +00:00
for _, piece := range pointer.GetRemote().GetRemotePieces() {
node, ok := nodes[piece.NodeId]
if !ok {
nodeErrors.Add(errs.New("node %q is not reliable", piece.NodeId))
continue
}
address := node.Address.Address
if node.LastIPPort != "" {
address = node.LastIPPort
}
_, err := signer.Sign(ctx, storj.NodeURL{ID: piece.NodeId, Address: address}, piece.PieceNum)
2019-03-28 20:09:23 +00:00
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
}
if len(signer.AddressedLimits) == 0 {
return nil, storj.PiecePrivateKey{}, Error.New("failed creating order limits: %w", nodeErrors.Err())
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return signer.AddressedLimits, signer.PrivateKey, nil
2019-03-28 20:09:23 +00:00
}
// CreateAuditOrderLimits creates the order limits for auditing the pieces of pointer.
func (service *Service) CreateAuditOrderLimits(ctx context.Context, bucket metabase.BucketLocation, pointer *pb.Pointer, skip map[storj.NodeID]bool) (_ []*pb.AddressedOrderLimit, _ storj.PiecePrivateKey, err error) {
defer mon.Task()(&ctx)(&err)
2019-03-28 20:09:23 +00:00
redundancy := pointer.GetRemote().GetRedundancy()
shareSize := redundancy.GetErasureShareSize()
totalPieces := redundancy.GetTotal()
nodeIDs := make([]storj.NodeID, len(pointer.GetRemote().GetRemotePieces()))
for i, piece := range pointer.GetRemote().GetRemotePieces() {
nodeIDs[i] = piece.NodeId
}
nodes, err := service.overlay.GetOnlineNodesForGetDelete(ctx, nodeIDs)
if err != nil {
service.log.Debug("error getting nodes from overlay", zap.Error(err))
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
signer, err := NewSignerAudit(service, pointer.GetRemote().RootPieceId, time.Now(), int64(shareSize))
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
var nodeErrors errs.Group
2019-03-28 20:09:23 +00:00
var limitsCount int32
limits := make([]*pb.AddressedOrderLimit, totalPieces)
for _, piece := range pointer.GetRemote().GetRemotePieces() {
if skip[piece.NodeId] {
continue
}
node, ok := nodes[piece.NodeId]
if !ok {
nodeErrors.Add(errs.New("node %q is not reliable", piece.NodeId))
continue
}
limit, err := signer.Sign(ctx, storj.NodeURL{
ID: piece.NodeId,
Address: node.Address.Address,
}, piece.PieceNum)
2019-03-28 20:09:23 +00:00
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
limits[piece.GetPieceNum()] = limit
2019-03-28 20:09:23 +00:00
limitsCount++
}
2019-03-28 20:09:23 +00:00
if limitsCount < redundancy.GetMinReq() {
err = Error.New("not enough nodes available: got %d, required %d", limitsCount, redundancy.GetMinReq())
return nil, storj.PiecePrivateKey{}, errs.Combine(err, nodeErrors.Err())
2019-03-28 20:09:23 +00:00
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if err := service.updateBandwidth(ctx, bucket, limits...); err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return limits, signer.PrivateKey, nil
2019-03-28 20:09:23 +00:00
}
// CreateAuditOrderLimit creates an order limit for auditing a single the piece from a pointer.
func (service *Service) CreateAuditOrderLimit(ctx context.Context, bucket metabase.BucketLocation, nodeID storj.NodeID, pieceNum int32, rootPieceID storj.PieceID, shareSize int32) (limit *pb.AddressedOrderLimit, _ storj.PiecePrivateKey, err error) {
// TODO reduce number of params ?
defer mon.Task()(&ctx)(&err)
node, err := service.overlay.Get(ctx, nodeID)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if node.Disqualified != nil {
2019-08-21 17:30:29 +01:00
return nil, storj.PiecePrivateKey{}, overlay.ErrNodeDisqualified.New("%v", nodeID)
}
if node.ExitStatus.ExitFinishedAt != nil {
return nil, storj.PiecePrivateKey{}, overlay.ErrNodeFinishedGE.New("%v", nodeID)
}
if !service.overlay.IsOnline(node) {
2019-08-21 17:30:29 +01:00
return nil, storj.PiecePrivateKey{}, overlay.ErrNodeOffline.New("%v", nodeID)
}
signer, err := NewSignerAudit(service, rootPieceID, time.Now(), int64(shareSize))
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
orderLimit, err := signer.Sign(ctx, storj.NodeURL{
ID: nodeID,
Address: node.Address.Address,
}, pieceNum)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if err := service.updateBandwidth(ctx, bucket, limit); err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return orderLimit, signer.PrivateKey, nil
}
[V3-1927] Repairer uploads to max threshold instead of success… (#2423) * pkg/datarepair: Add test to check num upload pieces Add a new test for ensuring the number of pieces that the repair process upload when a segment is injured. * satellite/orders: Don't create "put order limits" over total Repair must not create "put order limits" more than the total count. * pkg/datarepair: Update upload repair pieces test Update the test which checks the number of pieces which are uploaded during a repair for using the same excess over the success threshold value than the implementation. * satellites/orders: Limit repair put order for not being total Limit the number of put orders to be used by repair for only uploading pieces to a % excess over the successful threshold. * pkg/datarepair: Change DataRepair test to pass again Make some changes in the DataRepair test to make pass again after the repair upload repaired pieces only until a % excess over success threshold. Also update the steps description of the DataRepair test after it has been changed, to match on what's now, besides to leave it more generic for avoiding having to update it on minimal future refactorings. * satellite: Make repair excess optimal threshold configurable Add a new configuration parameter to the satellite for being able to configure the percentage excess over the optimal threshold, used for determining how many pieces should be repaired/uploaded, rather than having the value hard coded. * repairer: Add configurable param to segments/repairer Add a new parameters to the segment/repairer to calculate the maximum number of excess nodes, based on the optimal threshold, that repaired pieces can be uploaded. This new parameter has been added for not returning more nodes than the number of upload orders for data repair satellite service calculate for repairing pieces. * pkg/storage/ec: Update log message in clien.Repair * satellite: Update configuration lock file
2019-07-11 23:44:47 +01:00
// CreateGetRepairOrderLimits creates the order limits for downloading the
// healthy pieces of pointer as the source for repair.
//
// The length of the returned orders slice is the total number of pieces of the
// segment, setting to null the ones which don't correspond to a healthy piece.
2019-03-28 20:09:23 +00:00
// CreateGetRepairOrderLimits creates the order limits for downloading the healthy pieces of pointer as the source for repair.
func (service *Service) CreateGetRepairOrderLimits(ctx context.Context, bucket metabase.BucketLocation, pointer *pb.Pointer, healthy []*pb.RemotePiece) (_ []*pb.AddressedOrderLimit, _ storj.PiecePrivateKey, err error) {
defer mon.Task()(&ctx)(&err)
redundancy, err := eestream.NewRedundancyStrategyFromProto(pointer.GetRemote().GetRedundancy())
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
pieceSize := eestream.CalcPieceSize(pointer.GetSegmentSize(), redundancy)
totalPieces := redundancy.TotalCount()
2019-03-28 20:09:23 +00:00
nodeIDs := make([]storj.NodeID, len(pointer.GetRemote().GetRemotePieces()))
for i, piece := range pointer.GetRemote().GetRemotePieces() {
nodeIDs[i] = piece.NodeId
}
nodes, err := service.overlay.GetOnlineNodesForGetDelete(ctx, nodeIDs)
if err != nil {
service.log.Debug("error getting nodes from overlay", zap.Error(err))
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
signer, err := NewSignerRepairGet(service, pointer.GetRemote().RootPieceId, time.Now(), pieceSize)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
var nodeErrors errs.Group
var limitsCount int
2019-03-28 20:09:23 +00:00
limits := make([]*pb.AddressedOrderLimit, totalPieces)
for _, piece := range healthy {
node, ok := nodes[piece.NodeId]
if !ok {
nodeErrors.Add(errs.New("node %q is not reliable", piece.NodeId))
continue
}
limit, err := signer.Sign(ctx, storj.NodeURL{
ID: piece.NodeId,
Address: node.Address.Address,
}, piece.PieceNum)
2019-03-28 20:09:23 +00:00
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
limits[piece.GetPieceNum()] = limit
limitsCount++
2019-03-28 20:09:23 +00:00
}
if limitsCount < redundancy.RequiredCount() {
err = Error.New("not enough nodes available: got %d, required %d", limitsCount, redundancy.RequiredCount())
return nil, storj.PiecePrivateKey{}, errs.Combine(err, nodeErrors.Err())
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if err := service.updateBandwidth(ctx, bucket, limits...); err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return limits, signer.PrivateKey, nil
2019-03-28 20:09:23 +00:00
}
// CreatePutRepairOrderLimits creates the order limits for uploading the repaired pieces of pointer to newNodes.
func (service *Service) CreatePutRepairOrderLimits(ctx context.Context, bucket metabase.BucketLocation, pointer *pb.Pointer, getOrderLimits []*pb.AddressedOrderLimit, newNodes []*overlay.SelectedNode, optimalThresholdMultiplier float64) (_ []*pb.AddressedOrderLimit, _ storj.PiecePrivateKey, err error) {
defer mon.Task()(&ctx)(&err)
2019-03-28 20:09:23 +00:00
// Create the order limits for being used to upload the repaired pieces
redundancy, err := eestream.NewRedundancyStrategyFromProto(pointer.GetRemote().GetRedundancy())
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
pieceSize := eestream.CalcPieceSize(pointer.GetSegmentSize(), redundancy)
totalPieces := redundancy.TotalCount()
totalPiecesAfterRepair := int(math.Ceil(float64(redundancy.OptimalThreshold()) * optimalThresholdMultiplier))
if totalPiecesAfterRepair > totalPieces {
totalPiecesAfterRepair = totalPieces
}
var numCurrentPieces int
for _, o := range getOrderLimits {
if o != nil {
numCurrentPieces++
}
}
totalPiecesToRepair := totalPiecesAfterRepair - numCurrentPieces
limits := make([]*pb.AddressedOrderLimit, totalPieces)
signer, err := NewSignerRepairPut(service, pointer.GetRemote().RootPieceId, pointer.ExpirationDate, time.Now(), pieceSize)
2019-03-28 20:09:23 +00:00
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
var pieceNum int32
for _, node := range newNodes {
for int(pieceNum) < totalPieces && getOrderLimits[pieceNum] != nil {
pieceNum++
2019-03-28 20:09:23 +00:00
}
if int(pieceNum) >= totalPieces { // should not happen
return nil, storj.PiecePrivateKey{}, Error.New("piece num greater than total pieces: %d >= %d", pieceNum, totalPieces)
2019-03-28 20:09:23 +00:00
}
limit, err := signer.Sign(ctx, storj.NodeURL{
ID: node.ID,
Address: node.Address.Address,
}, pieceNum)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
2019-03-28 20:09:23 +00:00
}
limits[pieceNum] = limit
pieceNum++
totalPiecesToRepair--
[V3-1927] Repairer uploads to max threshold instead of success… (#2423) * pkg/datarepair: Add test to check num upload pieces Add a new test for ensuring the number of pieces that the repair process upload when a segment is injured. * satellite/orders: Don't create "put order limits" over total Repair must not create "put order limits" more than the total count. * pkg/datarepair: Update upload repair pieces test Update the test which checks the number of pieces which are uploaded during a repair for using the same excess over the success threshold value than the implementation. * satellites/orders: Limit repair put order for not being total Limit the number of put orders to be used by repair for only uploading pieces to a % excess over the successful threshold. * pkg/datarepair: Change DataRepair test to pass again Make some changes in the DataRepair test to make pass again after the repair upload repaired pieces only until a % excess over success threshold. Also update the steps description of the DataRepair test after it has been changed, to match on what's now, besides to leave it more generic for avoiding having to update it on minimal future refactorings. * satellite: Make repair excess optimal threshold configurable Add a new configuration parameter to the satellite for being able to configure the percentage excess over the optimal threshold, used for determining how many pieces should be repaired/uploaded, rather than having the value hard coded. * repairer: Add configurable param to segments/repairer Add a new parameters to the segment/repairer to calculate the maximum number of excess nodes, based on the optimal threshold, that repaired pieces can be uploaded. This new parameter has been added for not returning more nodes than the number of upload orders for data repair satellite service calculate for repairing pieces. * pkg/storage/ec: Update log message in clien.Repair * satellite: Update configuration lock file
2019-07-11 23:44:47 +01:00
if totalPiecesToRepair == 0 {
break
2019-03-28 20:09:23 +00:00
}
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if err := service.updateBandwidth(ctx, bucket, limits...); err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return limits, signer.PrivateKey, nil
2019-03-28 20:09:23 +00:00
}
2019-04-05 08:42:56 +01:00
// CreateGracefulExitPutOrderLimit creates an order limit for graceful exit put transfers.
func (service *Service) CreateGracefulExitPutOrderLimit(ctx context.Context, bucket metabase.BucketLocation, nodeID storj.NodeID, pieceNum int32, rootPieceID storj.PieceID, shareSize int32) (limit *pb.AddressedOrderLimit, _ storj.PiecePrivateKey, err error) {
defer mon.Task()(&ctx)(&err)
// should this use KnownReliable or similar?
node, err := service.overlay.Get(ctx, nodeID)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if node.Disqualified != nil {
return nil, storj.PiecePrivateKey{}, overlay.ErrNodeDisqualified.New("%v", nodeID)
}
if !service.overlay.IsOnline(node) {
return nil, storj.PiecePrivateKey{}, overlay.ErrNodeOffline.New("%v", nodeID)
}
signer, err := NewSignerGracefulExit(service, rootPieceID, time.Now(), shareSize)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
nodeURL := storj.NodeURL{ID: nodeID, Address: node.Address.Address}
limit, err = signer.Sign(ctx, nodeURL, pieceNum)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
err = service.saveSerial(ctx, signer.Serial, bucket, signer.OrderExpiration)
if err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
if err := service.updateBandwidth(ctx, bucket, limit); err != nil {
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
}
return limit, signer.PrivateKey, nil
}
// UpdateGetInlineOrder updates amount of inline GET bandwidth for given bucket.
func (service *Service) UpdateGetInlineOrder(ctx context.Context, bucket metabase.BucketLocation, amount int64) (err error) {
defer mon.Task()(&ctx)(&err)
now := time.Now().UTC()
2019-04-05 08:42:56 +01:00
intervalStart := time.Date(now.Year(), now.Month(), now.Day(), now.Hour(), 0, 0, 0, now.Location())
return service.orders.UpdateBucketBandwidthInline(ctx, bucket.ProjectID, []byte(bucket.BucketName), pb.PieceAction_GET, amount, intervalStart)
2019-04-05 08:42:56 +01:00
}
// UpdatePutInlineOrder updates amount of inline PUT bandwidth for given bucket.
func (service *Service) UpdatePutInlineOrder(ctx context.Context, bucket metabase.BucketLocation, amount int64) (err error) {
defer mon.Task()(&ctx)(&err)
now := time.Now().UTC()
2019-04-05 08:42:56 +01:00
intervalStart := time.Date(now.Year(), now.Month(), now.Day(), now.Hour(), 0, 0, 0, now.Location())
return service.orders.UpdateBucketBandwidthInline(ctx, bucket.ProjectID, []byte(bucket.BucketName), pb.PieceAction_PUT, amount, intervalStart)
2019-04-05 08:42:56 +01:00
}