storj/satellite/gracefulexit/pathcollector.go
Kaloyan Raev 92a2be2abd satellite/metainfo: get away from using pb.Pointer in Metainfo Loop
As part of the Metainfo Refactoring, we need to make the Metainfo Loop
working with both the current PointerDB and the new Metabase. Thus, the
Metainfo Loop should pass to the Observer interface more specific Object
and Segment types instead of pb.Pointer.

After this change, there are still a couple of use cases that require
access to the pb.Pointer (hence we have it as a field in the
metainfo.Segment type):
1. Expired Deletion Service
2. Repair Service

It would require additional refactoring in these two services before we
are able to clean this.

Change-Id: Ib3eb6b7507ed89d5ba745ffbb6b37524ef10ed9f
2020-10-27 13:06:47 +00:00

120 lines
3.4 KiB
Go

// Copyright (C) 2019 Storj Labs, Inc.
// See LICENSE for copying information.
package gracefulexit
import (
"context"
"sync"
"github.com/zeebo/errs"
"go.uber.org/zap"
"storj.io/common/storj"
"storj.io/storj/satellite/metainfo"
"storj.io/uplink/private/eestream"
)
var _ metainfo.Observer = (*PathCollector)(nil)
// PathCollector uses the metainfo loop to add paths to node reservoirs
//
// architecture: Observer
type PathCollector struct {
db DB
nodeIDMutex sync.Mutex
nodeIDStorage map[storj.NodeID]int64
buffer []TransferQueueItem
log *zap.Logger
batchSize int
}
// NewPathCollector instantiates a path collector.
func NewPathCollector(db DB, nodeIDs storj.NodeIDList, log *zap.Logger, batchSize int) *PathCollector {
buffer := make([]TransferQueueItem, 0, batchSize)
collector := &PathCollector{
db: db,
log: log,
buffer: buffer,
batchSize: batchSize,
}
if len(nodeIDs) > 0 {
collector.nodeIDStorage = make(map[storj.NodeID]int64, len(nodeIDs))
for _, nodeID := range nodeIDs {
collector.nodeIDStorage[nodeID] = 0
}
}
return collector
}
// Flush persists the current buffer items to the database.
func (collector *PathCollector) Flush(ctx context.Context) (err error) {
return collector.flush(ctx, 1)
}
// RemoteSegment takes a remote segment found in metainfo and creates a graceful exit transfer queue item if it doesn't exist already.
func (collector *PathCollector) RemoteSegment(ctx context.Context, segment *metainfo.Segment) (err error) {
if len(collector.nodeIDStorage) == 0 {
return nil
}
collector.nodeIDMutex.Lock()
defer collector.nodeIDMutex.Unlock()
numPieces := len(segment.Pieces)
key := segment.Location.Encode()
for _, piece := range segment.Pieces {
if _, ok := collector.nodeIDStorage[piece.StorageNode]; !ok {
continue
}
redundancy, err := eestream.NewRedundancyStrategyFromStorj(segment.Redundancy)
if err != nil {
return err
}
pieceSize := eestream.CalcPieceSize(int64(segment.DataSize), redundancy)
collector.nodeIDStorage[piece.StorageNode] += pieceSize
item := TransferQueueItem{
NodeID: piece.StorageNode,
Key: key,
PieceNum: int32(piece.Number),
RootPieceID: segment.RootPieceID,
DurabilityRatio: float64(numPieces) / float64(segment.Redundancy.TotalShares),
}
collector.log.Debug("adding piece to transfer queue.", zap.Stringer("Node ID", piece.StorageNode),
zap.ByteString("key", key), zap.Uint16("piece num", piece.Number),
zap.Int("num pieces", numPieces), zap.Int16("total possible pieces", segment.Redundancy.TotalShares))
collector.buffer = append(collector.buffer, item)
err = collector.flush(ctx, collector.batchSize)
if err != nil {
return err
}
}
return nil
}
// Object returns nil because the audit service does not interact with objects.
func (collector *PathCollector) Object(ctx context.Context, object *metainfo.Object) (err error) {
return nil
}
// InlineSegment returns nil because we're only auditing for storage nodes for now.
func (collector *PathCollector) InlineSegment(ctx context.Context, segment *metainfo.Segment) (err error) {
return nil
}
func (collector *PathCollector) flush(ctx context.Context, limit int) (err error) {
if len(collector.buffer) >= limit {
err = collector.db.Enqueue(ctx, collector.buffer)
collector.buffer = collector.buffer[:0]
return errs.Wrap(err)
}
return nil
}