2019-01-24 20:15:10 +00:00
|
|
|
// Copyright (C) 2019 Storj Labs, Inc.
|
2018-10-02 20:46:29 +01:00
|
|
|
// See LICENSE for copying information.
|
|
|
|
|
|
|
|
package checker
|
2018-10-03 19:35:56 +01:00
|
|
|
|
|
|
|
import (
|
|
|
|
"context"
|
2018-10-30 19:16:40 +00:00
|
|
|
"time"
|
2018-10-04 22:40:34 +01:00
|
|
|
|
2018-11-20 18:29:07 +00:00
|
|
|
"github.com/gogo/protobuf/proto"
|
2019-01-23 19:58:44 +00:00
|
|
|
"github.com/zeebo/errs"
|
2018-10-04 22:40:34 +01:00
|
|
|
"go.uber.org/zap"
|
2019-01-23 19:58:44 +00:00
|
|
|
monkit "gopkg.in/spacemonkeygo/monkit.v2"
|
2018-10-04 22:40:34 +01:00
|
|
|
|
2019-02-11 21:06:39 +00:00
|
|
|
"storj.io/storj/internal/sync2"
|
2018-12-10 19:08:45 +00:00
|
|
|
"storj.io/storj/pkg/datarepair/irreparable"
|
2018-10-09 17:09:33 +01:00
|
|
|
"storj.io/storj/pkg/datarepair/queue"
|
2019-03-23 08:06:11 +00:00
|
|
|
"storj.io/storj/pkg/overlay"
|
2018-10-09 17:09:33 +01:00
|
|
|
"storj.io/storj/pkg/pb"
|
2018-11-29 18:39:27 +00:00
|
|
|
"storj.io/storj/pkg/storj"
|
2019-04-25 09:46:32 +01:00
|
|
|
"storj.io/storj/satellite/metainfo"
|
2018-10-09 17:09:33 +01:00
|
|
|
"storj.io/storj/storage"
|
2018-10-03 19:35:56 +01:00
|
|
|
)
|
|
|
|
|
2019-01-23 19:58:44 +00:00
|
|
|
// Error is a standard error class for this package.
|
|
|
|
var (
|
|
|
|
Error = errs.Class("checker error")
|
|
|
|
mon = monkit.Package()
|
|
|
|
)
|
|
|
|
|
|
|
|
// Config contains configurable values for checker
|
|
|
|
type Config struct {
|
|
|
|
Interval time.Duration `help:"how frequently checker should audit segments" default:"30s"`
|
|
|
|
}
|
|
|
|
|
2018-10-09 17:09:33 +01:00
|
|
|
// Checker contains the information needed to do checks for missing pieces
|
2019-02-11 21:06:39 +00:00
|
|
|
type Checker struct {
|
2019-04-25 09:46:32 +01:00
|
|
|
metainfo *metainfo.Service
|
2019-05-08 18:59:50 +01:00
|
|
|
lastChecked string
|
2018-12-21 15:11:19 +00:00
|
|
|
repairQueue queue.RepairQueue
|
2019-03-23 08:06:11 +00:00
|
|
|
overlay *overlay.Cache
|
2018-12-10 19:08:45 +00:00
|
|
|
irrdb irreparable.DB
|
2018-10-09 17:09:33 +01:00
|
|
|
logger *zap.Logger
|
2019-02-14 12:33:41 +00:00
|
|
|
Loop sync2.Cycle
|
2018-10-09 17:09:33 +01:00
|
|
|
}
|
|
|
|
|
2019-01-18 13:54:08 +00:00
|
|
|
// NewChecker creates a new instance of checker
|
2019-04-25 09:46:32 +01:00
|
|
|
func NewChecker(metainfo *metainfo.Service, repairQueue queue.RepairQueue, overlay *overlay.Cache, irrdb irreparable.DB, limit int, logger *zap.Logger, interval time.Duration) *Checker {
|
2019-01-18 13:54:08 +00:00
|
|
|
// TODO: reorder arguments
|
2019-02-11 21:06:39 +00:00
|
|
|
checker := &Checker{
|
2019-04-25 09:46:32 +01:00
|
|
|
metainfo: metainfo,
|
2019-05-08 18:59:50 +01:00
|
|
|
lastChecked: "",
|
2018-10-09 17:09:33 +01:00
|
|
|
repairQueue: repairQueue,
|
|
|
|
overlay: overlay,
|
2018-12-04 16:26:30 +00:00
|
|
|
irrdb: irrdb,
|
2018-10-09 17:09:33 +01:00
|
|
|
logger: logger,
|
2019-02-14 12:33:41 +00:00
|
|
|
Loop: *sync2.NewCycle(interval),
|
2018-10-09 17:09:33 +01:00
|
|
|
}
|
2019-02-11 21:06:39 +00:00
|
|
|
return checker
|
2018-10-09 17:09:33 +01:00
|
|
|
}
|
|
|
|
|
2018-11-01 14:03:45 +00:00
|
|
|
// Run the checker loop
|
2019-02-11 21:06:39 +00:00
|
|
|
func (checker *Checker) Run(ctx context.Context) (err error) {
|
2018-11-01 14:03:45 +00:00
|
|
|
defer mon.Task()(&ctx)(&err)
|
|
|
|
|
2019-02-14 12:33:41 +00:00
|
|
|
return checker.Loop.Run(ctx, func(ctx context.Context) error {
|
2019-02-11 21:06:39 +00:00
|
|
|
err := checker.IdentifyInjuredSegments(ctx)
|
2018-11-01 14:03:45 +00:00
|
|
|
if err != nil {
|
2019-02-11 21:06:39 +00:00
|
|
|
checker.logger.Error("error with injured segments identification: ", zap.Error(err))
|
2018-11-01 14:03:45 +00:00
|
|
|
}
|
2019-04-01 10:16:17 +01:00
|
|
|
return nil
|
2019-02-11 21:06:39 +00:00
|
|
|
})
|
2018-11-01 14:03:45 +00:00
|
|
|
}
|
|
|
|
|
2019-02-14 12:33:41 +00:00
|
|
|
// Close halts the Checker loop
|
|
|
|
func (checker *Checker) Close() error {
|
|
|
|
checker.Loop.Close()
|
|
|
|
return nil
|
|
|
|
}
|
2019-01-18 13:54:08 +00:00
|
|
|
|
2019-04-25 09:46:32 +01:00
|
|
|
// IdentifyInjuredSegments checks for missing pieces off of the metainfo and overlay cache
|
2019-02-11 21:06:39 +00:00
|
|
|
func (checker *Checker) IdentifyInjuredSegments(ctx context.Context) (err error) {
|
2018-10-09 17:09:33 +01:00
|
|
|
defer mon.Task()(&ctx)(&err)
|
|
|
|
|
2019-02-26 15:17:51 +00:00
|
|
|
var remoteSegmentsChecked int64
|
|
|
|
var remoteSegmentsNeedingRepair int64
|
|
|
|
var remoteSegmentsLost int64
|
|
|
|
var remoteSegmentInfo []string
|
|
|
|
|
2019-05-08 18:59:50 +01:00
|
|
|
err = checker.metainfo.Iterate("", checker.lastChecked, true, false,
|
2018-10-09 17:09:33 +01:00
|
|
|
func(it storage.Iterator) error {
|
|
|
|
var item storage.ListItem
|
2019-05-08 18:59:50 +01:00
|
|
|
|
|
|
|
defer func() {
|
|
|
|
var nextItem storage.ListItem
|
|
|
|
it.Next(&nextItem)
|
|
|
|
// start at the next item in the next call
|
|
|
|
checker.lastChecked = nextItem.Key.String()
|
|
|
|
// if keys are equal, start from the beginning in the next call
|
|
|
|
if nextItem.Key.String() == item.Key.String() {
|
|
|
|
checker.lastChecked = ""
|
|
|
|
}
|
|
|
|
}()
|
|
|
|
|
2019-02-14 12:33:41 +00:00
|
|
|
for it.Next(&item) {
|
2018-10-09 17:09:33 +01:00
|
|
|
pointer := &pb.Pointer{}
|
2018-12-12 15:39:16 +00:00
|
|
|
|
2018-10-09 17:09:33 +01:00
|
|
|
err = proto.Unmarshal(item.Value, pointer)
|
|
|
|
if err != nil {
|
|
|
|
return Error.New("error unmarshalling pointer %s", err)
|
|
|
|
}
|
2018-12-12 15:39:16 +00:00
|
|
|
|
2018-11-20 15:54:22 +00:00
|
|
|
remote := pointer.GetRemote()
|
|
|
|
if remote == nil {
|
|
|
|
continue
|
|
|
|
}
|
2018-12-12 15:39:16 +00:00
|
|
|
|
2018-11-20 15:54:22 +00:00
|
|
|
pieces := remote.GetRemotePieces()
|
|
|
|
if pieces == nil {
|
2019-02-11 21:06:39 +00:00
|
|
|
checker.logger.Debug("no pieces on remote segment")
|
2018-11-20 15:54:22 +00:00
|
|
|
continue
|
|
|
|
}
|
2018-12-12 15:39:16 +00:00
|
|
|
|
2019-05-16 14:49:10 +01:00
|
|
|
missingPieces, err := checker.overlay.GetMissingPieces(ctx, pieces)
|
2018-12-06 18:51:23 +00:00
|
|
|
if err != nil {
|
2019-04-23 21:54:39 +01:00
|
|
|
return Error.New("error getting missing pieces %s", err)
|
2019-04-08 20:46:23 +01:00
|
|
|
}
|
2018-12-06 18:51:23 +00:00
|
|
|
|
2019-02-26 15:17:51 +00:00
|
|
|
remoteSegmentsChecked++
|
2019-05-17 20:02:40 +01:00
|
|
|
numHealthy := int32(len(pieces) - len(missingPieces))
|
|
|
|
redundancy := pointer.Remote.Redundancy
|
|
|
|
// we repair when the number of healthy files is less than or equal to the repair threshold
|
|
|
|
// except for the case when the repair and success thresholds are the same (a case usually seen during testing)
|
|
|
|
if numHealthy >= redundancy.MinReq && numHealthy <= redundancy.RepairThreshold && redundancy.RepairThreshold != redundancy.SuccessThreshold {
|
|
|
|
if len(missingPieces) == 0 {
|
|
|
|
checker.logger.Warn("Missing pieces is zero in checker, but this should be impossible -- bad redundancy scheme.")
|
|
|
|
continue
|
|
|
|
}
|
2019-02-26 15:17:51 +00:00
|
|
|
remoteSegmentsNeedingRepair++
|
2019-04-16 19:14:09 +01:00
|
|
|
err = checker.repairQueue.Insert(ctx, &pb.InjuredSegment{
|
2018-10-09 17:09:33 +01:00
|
|
|
Path: string(item.Key),
|
|
|
|
LostPieces: missingPieces,
|
|
|
|
})
|
|
|
|
if err != nil {
|
|
|
|
return Error.New("error adding injured segment to queue %s", err)
|
|
|
|
}
|
2019-05-17 20:02:40 +01:00
|
|
|
} else if numHealthy < redundancy.MinReq {
|
2019-02-26 15:17:51 +00:00
|
|
|
pathElements := storj.SplitPath(storj.Path(item.Key))
|
|
|
|
// check to make sure there are at least *4* path elements. the first three
|
|
|
|
// are project, segment, and bucket name, but we want to make sure we're talking
|
|
|
|
// about an actual object, and that there's an object name specified
|
|
|
|
if len(pathElements) >= 4 {
|
|
|
|
project, bucketName, segmentpath := pathElements[0], pathElements[2], pathElements[3]
|
|
|
|
lostSegInfo := storj.JoinPaths(project, bucketName, segmentpath)
|
|
|
|
if contains(remoteSegmentInfo, lostSegInfo) == false {
|
|
|
|
remoteSegmentInfo = append(remoteSegmentInfo, lostSegInfo)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-03-23 08:06:11 +00:00
|
|
|
// TODO: irreparable segment should be using storj.NodeID or something, since at the point of repair
|
|
|
|
// it may have been already repaired once.
|
|
|
|
|
2019-02-26 15:17:51 +00:00
|
|
|
remoteSegmentsLost++
|
2018-12-04 16:26:30 +00:00
|
|
|
// make an entry in to the irreparable table
|
2019-03-15 20:21:52 +00:00
|
|
|
segmentInfo := &pb.IrreparableSegment{
|
|
|
|
Path: item.Key,
|
|
|
|
SegmentDetail: pointer,
|
|
|
|
LostPieces: int32(len(missingPieces)),
|
|
|
|
LastRepairAttempt: time.Now().Unix(),
|
|
|
|
RepairAttemptCount: int64(1),
|
2018-12-04 16:26:30 +00:00
|
|
|
}
|
|
|
|
|
2019-05-08 18:59:50 +01:00
|
|
|
// add the entry if new or update attempt count if already exists
|
2019-02-11 21:06:39 +00:00
|
|
|
err := checker.irrdb.IncrementRepairAttempts(ctx, segmentInfo)
|
2018-12-04 16:26:30 +00:00
|
|
|
if err != nil {
|
|
|
|
return Error.New("error handling irreparable segment to queue %s", err)
|
|
|
|
}
|
2018-10-09 17:09:33 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
return nil
|
|
|
|
},
|
|
|
|
)
|
2019-02-26 15:17:51 +00:00
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
mon.IntVal("remote_segments_checked").Observe(remoteSegmentsChecked)
|
|
|
|
mon.IntVal("remote_segments_needing_repair").Observe(remoteSegmentsNeedingRepair)
|
|
|
|
mon.IntVal("remote_segments_lost").Observe(remoteSegmentsLost)
|
|
|
|
mon.IntVal("remote_files_lost").Observe(int64(len(remoteSegmentInfo)))
|
|
|
|
|
|
|
|
return nil
|
2018-10-09 17:09:33 +01:00
|
|
|
}
|
|
|
|
|
2019-02-26 15:17:51 +00:00
|
|
|
// checks for a string in slice
|
|
|
|
func contains(a []string, x string) bool {
|
|
|
|
for _, n := range a {
|
|
|
|
if x == n {
|
|
|
|
return true
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return false
|
|
|
|
}
|