storj/satellite/satellitedb/irreparabledb.go
paul cannon 79553059cb satellite/repair: put irreparable segments in irreparableDB
Previously, we were simply discarding rows from the repair queue when
they couldn't be repaired (either because the overlay said too many
nodes were down, or because we failed to download enough pieces).

Now, such segments will be put into the irreparableDB for further
and (hopefully) more focused attention.

This change also better differentiates some error cases from Repair()
for monitoring purposes.

Change-Id: I82a52a6da50c948ddd651048e2a39cb4b1e6df5c
2020-03-09 21:45:16 +00:00

124 lines
3.9 KiB
Go

// Copyright (C) 2019 Storj Labs, Inc.
// See LICENSE for copying information.
package satellitedb
import (
"context"
"database/sql"
"github.com/golang/protobuf/proto"
"storj.io/common/pb"
"storj.io/storj/satellite/satellitedb/dbx"
)
type irreparableDB struct {
db *satelliteDB
}
// IncrementRepairAttempts a db entry for to increment the repair attempts field
func (db *irreparableDB) IncrementRepairAttempts(ctx context.Context, segmentInfo *pb.IrreparableSegment) (err error) {
defer mon.Task()(&ctx)(&err)
err = db.db.WithTx(ctx, func(ctx context.Context, tx *dbx.Tx) (err error) {
bytes, err := proto.Marshal(segmentInfo.SegmentDetail)
if err != nil {
return err
}
dbxInfo, err := tx.Get_Irreparabledb_By_Segmentpath(ctx, dbx.Irreparabledb_Segmentpath(segmentInfo.Path))
if err != nil {
if err == sql.ErrNoRows {
// no rows err, so create/insert an entry
return tx.CreateNoReturn_Irreparabledb(
ctx,
dbx.Irreparabledb_Segmentpath(segmentInfo.Path),
dbx.Irreparabledb_Segmentdetail(bytes),
dbx.Irreparabledb_PiecesLostCount(int64(segmentInfo.LostPieces)),
dbx.Irreparabledb_SegDamagedUnixSec(segmentInfo.LastRepairAttempt),
dbx.Irreparabledb_RepairAttemptCount(segmentInfo.RepairAttemptCount),
)
}
return err
}
// row exits increment the attempt counter
dbxInfo.RepairAttemptCount++
updateFields := dbx.Irreparabledb_Update_Fields{}
updateFields.RepairAttemptCount = dbx.Irreparabledb_RepairAttemptCount(dbxInfo.RepairAttemptCount)
updateFields.SegDamagedUnixSec = dbx.Irreparabledb_SegDamagedUnixSec(segmentInfo.LastRepairAttempt)
err = tx.UpdateNoReturn_Irreparabledb_By_Segmentpath(
ctx,
dbx.Irreparabledb_Segmentpath(dbxInfo.Segmentpath),
updateFields,
)
return err
})
return Error.Wrap(err)
}
// Get a irreparable's segment info from the db
func (db *irreparableDB) Get(ctx context.Context, segmentPath []byte) (resp *pb.IrreparableSegment, err error) {
defer mon.Task()(&ctx)(&err)
dbxInfo, err := db.db.Get_Irreparabledb_By_Segmentpath(ctx, dbx.Irreparabledb_Segmentpath(segmentPath))
if err != nil {
return nil, Error.Wrap(err)
}
p := &pb.Pointer{}
err = proto.Unmarshal(dbxInfo.Segmentdetail, p)
if err != nil {
return nil, Error.Wrap(err)
}
return &pb.IrreparableSegment{
Path: dbxInfo.Segmentpath,
SegmentDetail: p,
LostPieces: int32(dbxInfo.PiecesLostCount),
LastRepairAttempt: dbxInfo.SegDamagedUnixSec,
RepairAttemptCount: dbxInfo.RepairAttemptCount,
}, nil
}
// GetLimited returns a list of irreparable segment info starting after the last segment info we retrieved
func (db *irreparableDB) GetLimited(ctx context.Context, limit int, lastSeenSegmentPath []byte) (resp []*pb.IrreparableSegment, err error) {
defer mon.Task()(&ctx)(&err)
// the offset is hardcoded to 0 since we are using the lastSeenSegmentPath to
// indicate the item we last listed instead. In a perfect world this db query would
// not take an offset as an argument, but currently dbx only supports `limitoffset`
const offset = 0
rows, err := db.db.Limited_Irreparabledb_By_Segmentpath_Greater_OrderBy_Asc_Segmentpath(ctx,
dbx.Irreparabledb_Segmentpath(lastSeenSegmentPath),
limit, offset,
)
if err != nil {
return nil, err
}
for _, row := range rows {
p := &pb.Pointer{}
err = proto.Unmarshal(row.Segmentdetail, p)
if err != nil {
return nil, err
}
segment := &pb.IrreparableSegment{
Path: row.Segmentpath,
SegmentDetail: p,
LostPieces: int32(row.PiecesLostCount),
LastRepairAttempt: row.SegDamagedUnixSec,
RepairAttemptCount: row.RepairAttemptCount,
}
resp = append(resp, segment)
}
return resp, err
}
// Delete a irreparable's segment info from the db
func (db *irreparableDB) Delete(ctx context.Context, segmentPath []byte) (err error) {
defer mon.Task()(&ctx)(&err)
_, err = db.db.Delete_Irreparabledb_By_Segmentpath(ctx, dbx.Irreparabledb_Segmentpath(segmentPath))
return Error.Wrap(err)
}