Skip to content

Commit

Permalink
receive: Fixed small options race; Removed unused StartTime feature.
Browse files Browse the repository at this point in the history 
startTimeMargin and StartTime is used only by Prometheus remote read, Thanos does not use it.
Fixed following race:

```
=== RUN   TestMultiTSDB/run_on_existing_storage
==================
WARNING: DATA RACE
Read at 0x00c00073ae80 by goroutine 69:
  github.com/prometheus/prometheus/tsdb.validateOpts()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/github.com/prometheus/[email protected]/tsdb/db.go:510 +0x55
  github.com/prometheus/prometheus/tsdb.Open()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/github.com/prometheus/[email protected]/tsdb/db.go:502 +0x61
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).getOrLoadTenant.func1()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:268 +0x56b
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).getOrLoadTenant()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:302 +0x4ef
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).Open.func1()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:142 +0x66
  golang.org/x/sync/errgroup.(*Group).Go.func1()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/golang.org/x/[email protected]/errgroup/errgroup.go:57 +0x85

Previous write at 0x00c00073ae80 by goroutine 57:
  github.com/prometheus/prometheus/tsdb.validateOpts()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/github.com/prometheus/[email protected]/tsdb/db.go:511 +0x1f2
  github.com/prometheus/prometheus/tsdb.Open()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/github.com/prometheus/[email protected]/tsdb/db.go:502 +0x61
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).getOrLoadTenant.func1()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:268 +0x56b
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).getOrLoadTenant()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:302 +0x4ef
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).Open.func1()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:142 +0x66
  golang.org/x/sync/errgroup.(*Group).Go.func1()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/golang.org/x/[email protected]/errgroup/errgroup.go:57 +0x85

Goroutine 69 (running) created at:
  golang.org/x/sync/errgroup.(*Group).Go()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/golang.org/x/[email protected]/errgroup/errgroup.go:54 +0x73
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).Open()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:141 +0x2af
  github.com/thanos-io/thanos/pkg/receive.TestMultiTSDB.func3()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb_test.go:118 +0x6d3
  testing.tRunner()
      /home/bwplotka/.gvm/gos/go1.14.2/src/testing/testing.go:991 +0x1eb

Goroutine 57 (running) created at:
  golang.org/x/sync/errgroup.(*Group).Go()
      /home/bwplotka/Repos/thanosgopath/pkg/mod/golang.org/x/[email protected]/errgroup/errgroup.go:54 +0x73
  github.com/thanos-io/thanos/pkg/receive.(*MultiTSDB).Open()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb.go:141 +0x2af
  github.com/thanos-io/thanos/pkg/receive.TestMultiTSDB.func3()
      /home/bwplotka/Repos/thanos/pkg/receive/multitsdb_test.go:118 +0x6d3
  testing.tRunner()
      /home/bwplotka/.gvm/gos/go1.14.2/src/testing/testing.go:991 +0x1eb
==================
```

Signed-off-by: Bartlomiej Plotka <[email protected]>
  • Loading branch information
@bwplotka
bwplotka committed Jun 29, 2020
1 parent dd2e003 commit ec9652e
Showing 1 changed file with 58 additions and 81 deletions.
139 changes: 58 additions & 81 deletions pkg/receive/multitsdb.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,21 +9,18 @@ import (
"os"
"path"
"sync"
"time"

"github.com/go-kit/kit/log"
"github.com/go-kit/kit/log/level"
"github.com/pkg/errors"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/common/model"
"github.com/prometheus/prometheus/pkg/labels"
"github.com/prometheus/prometheus/storage"
"github.com/prometheus/prometheus/tsdb"
terrors "github.com/prometheus/prometheus/tsdb/errors"
"github.com/thanos-io/thanos/pkg/block/metadata"
"github.com/thanos-io/thanos/pkg/component"
"github.com/thanos-io/thanos/pkg/objstore"
"github.com/thanos-io/thanos/pkg/runutil"
"github.com/thanos-io/thanos/pkg/shipper"
"github.com/thanos-io/thanos/pkg/store"
"golang.org/x/sync/errgroup"
Expand Down Expand Up @@ -72,7 +69,7 @@ func NewMultiTSDB(
}

type tenant struct {
tsdbOpts *tsdb.Options
startTimeMargin int64

readyS *ReadyStorage
tsdb *tsdb.DB
Expand All @@ -82,11 +79,10 @@ type tenant struct {
mtx *sync.RWMutex
}

func newTenant(tsdbOpts *tsdb.Options) *tenant {
func newTenant() *tenant {
return &tenant{
tsdbOpts: tsdbOpts,
readyS: &ReadyStorage{},
mtx: &sync.RWMutex{},
readyS: &ReadyStorage{},
mtx: &sync.RWMutex{},
}
}

Expand All @@ -113,7 +109,7 @@ func (t *tenant) db() *tsdb.DB {
}

func (t *tenant) set(storeTSDB *store.TSDBStore, tenantTSDB *tsdb.DB, ship *shipper.Shipper) {
t.readyS.Set(tenantTSDB, int64(2*time.Duration(t.tsdbOpts.MinBlockDuration).Seconds()*1000))
t.readyS.Set(tenantTSDB)
t.mtx.Lock()
t.tsdb = tenantTSDB
t.storeTSDB = storeTSDB
Expand Down Expand Up @@ -220,6 +216,47 @@ func (t *MultiTSDB) TSDBStores() map[string]*store.TSDBStore {
return res
}

func (t *MultiTSDB) startTSDB(logger log.Logger, tenantID string, tenant *tenant) error {
reg := prometheus.WrapRegistererWith(prometheus.Labels{"tenant": tenantID}, t.reg)
lbls := append(t.labels, labels.Label{Name: t.tenantLabelName, Value: tenantID})
dataDir := path.Join(t.dataDir, tenantID)

opts := *t.tsdbOpts
s, err := tsdb.Open(
dataDir,
logger,
&UnRegisterer{Registerer: reg},
&opts,
)
if err != nil {
t.mtx.Lock()
delete(t.tenants, tenantID)
t.mtx.Unlock()
return err
}

var ship *shipper.Shipper
if t.bucket != nil {
ship = shipper.New(
logger,
reg,
dataDir,
t.bucket,
func() labels.Labels { return lbls },
metadata.ReceiveSource,
t.allowOutOfOrderUpload,
)
}
tenant.set(store.NewTSDBStore(
logger,
reg,
s,
component.Receive,
lbls,
), s, ship)

return nil
}
func (t *MultiTSDB) getOrLoadTenant(tenantID string, blockingStart bool) (*tenant, error) {
// Fast path, as creating tenants is a very rare operation.
t.mtx.RLock()
Expand All @@ -239,68 +276,20 @@ func (t *MultiTSDB) getOrLoadTenant(tenantID string, blockingStart bool) (*tenan
return tenant, nil
}

tenant = newTenant(t.tsdbOpts)
tenant = newTenant()
t.tenants[tenantID] = tenant
t.mtx.Unlock()

var err error
startTSDB := func() {
reg := prometheus.WrapRegistererWith(prometheus.Labels{
"tenant": tenantID,
}, t.reg)
logger := log.With(t.logger, "tenant", tenantID)
lbls := append(t.labels, labels.Label{Name: t.tenantLabelName, Value: tenantID})
dataDir := path.Join(t.dataDir, tenantID)

var ship *shipper.Shipper
if t.bucket != nil {
ship = shipper.New(
logger,
reg,
dataDir,
t.bucket,
func() labels.Labels { return lbls },
metadata.ReceiveSource,
t.allowOutOfOrderUpload,
)
}

s, err := tsdb.Open(
dataDir,
logger,
&UnRegisterer{Registerer: reg},
t.tsdbOpts,
)

// Assign to outer error to report in blocking case.
if err != nil {
level.Error(logger).Log("msg", "failed to open tsdb", "err", err)
t.mtx.Lock()
delete(t.tenants, tenantID)
t.mtx.Unlock()
runutil.CloseWithLogOnErr(logger, s, "failed to close tsdb")
return
}

tenant.set(
store.NewTSDBStore(
logger,
reg,
s,
component.Receive,
lbls,
),
s,
ship,
)
}
logger := log.With(t.logger, "tenant", tenantID)
if !blockingStart {
go startTSDB()
go func() {
if err := t.startTSDB(logger, tenantID, tenant); err != nil {
level.Error(logger).Log("msg", "failed to start tsdb asynchronously", "err", err)
}
}()
return tenant, nil
}

startTSDB()
return tenant, err
return tenant, t.startTSDB(logger, tenantID, tenant)
}

func (t *MultiTSDB) TenantAppendable(tenantID string) (Appendable, error) {
Expand All @@ -323,11 +312,11 @@ type ReadyStorage struct {
}

// Set the storage.
func (s *ReadyStorage) Set(db *tsdb.DB, startTimeMargin int64) {
func (s *ReadyStorage) Set(db *tsdb.DB) {
s.mtx.Lock()
defer s.mtx.Unlock()

s.a = &adapter{db: db, startTimeMargin: startTimeMargin}
s.a = &adapter{db: db}
}

// Get the storage.
Expand All @@ -347,10 +336,7 @@ func (s *ReadyStorage) get() *adapter {

// StartTime implements the Storage interface.
func (s *ReadyStorage) StartTime() (int64, error) {
if x := s.get(); x != nil {
return x.StartTime()
}
return int64(model.Latest), ErrNotReady
return 0, errors.New("not implemented")
}

// Querier implements the Storage interface.
Expand Down Expand Up @@ -385,16 +371,7 @@ type adapter struct {

// StartTime implements the Storage interface.
func (a adapter) StartTime() (int64, error) {
var startTime int64

if len(a.db.Blocks()) > 0 {
startTime = a.db.Blocks()[0].Meta().MinTime
} else {
startTime = time.Now().Unix() * 1000
}

// Add a safety margin as it may take a few minutes for everything to spin up.
return startTime + a.startTimeMargin, nil
return 0, errors.New("not implemented")
}

func (a adapter) Querier(ctx context.Context, mint, maxt int64) (storage.Querier, error) {
Expand Down

0 comments on commit ec9652e

Please sign in to comment.