prometheus/head.go

// Copyright 2017 The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

package tsdb

import (
	"math"
	"runtime"
	"sort"
	"strings"
	"sync"
	"sync/atomic"
	"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/tsdb/chunkenc"
	"github.com/prometheus/tsdb/chunks"
	"github.com/prometheus/tsdb/index"
	"github.com/prometheus/tsdb/labels"
	"github.com/prometheus/tsdb/wal"
)

var (
	// ErrNotFound is returned if a looked up resource was not found.
	ErrNotFound = errors.Errorf("not found")

	// ErrOutOfOrderSample is returned if an appended sample has a
	// timestamp larger than the most recent sample.
	ErrOutOfOrderSample = errors.New("out of order sample")

	// ErrAmendSample is returned if an appended sample has the same timestamp
	// as the most recent sample but a different value.
	ErrAmendSample = errors.New("amending sample")

	// ErrOutOfBounds is returned if an appended sample is out of the
	// writable time range.
	ErrOutOfBounds = errors.New("out of bounds")
)

// Head handles reads and writes of time series data within a time window.
type Head struct {
	chunkRange int64
	metrics    *headMetrics
	wal        *wal.WAL
	logger     log.Logger
	appendPool sync.Pool
	bytesPool  sync.Pool

	minTime, maxTime int64
	lastSeriesID     uint64

	// All series addressable by their ID or hash.
	series *stripeSeries

	symMtx  sync.RWMutex
	symbols map[string]struct{}
	values  map[string]stringset // label names to possible values

	postings *index.MemPostings // postings lists for terms

	tombstones *memTombstones
}

type headMetrics struct {
	activeAppenders         prometheus.Gauge
	series                  prometheus.Gauge
	seriesCreated           prometheus.Counter
	seriesRemoved           prometheus.Counter
	seriesNotFound          prometheus.Counter
	chunks                  prometheus.Gauge
	chunksCreated           prometheus.Counter
	chunksRemoved           prometheus.Counter
	gcDuration              prometheus.Summary
	minTime                 prometheus.GaugeFunc
	maxTime                 prometheus.GaugeFunc
	samplesAppended         prometheus.Counter
	walTruncateDuration     prometheus.Summary
	headTruncateFail        prometheus.Counter
	headTruncateTotal       prometheus.Counter
	checkpointDeleteFail    prometheus.Counter
	checkpointDeleteTotal   prometheus.Counter
	checkpointCreationFail  prometheus.Counter
	checkpointCreationTotal prometheus.Counter
}

func newHeadMetrics(h *Head, r prometheus.Registerer) *headMetrics {
	m := &headMetrics{}

	m.activeAppenders = prometheus.NewGauge(prometheus.GaugeOpts{
		Name: "prometheus_tsdb_head_active_appenders",
		Help: "Number of currently active appender transactions",
	})
	m.series = prometheus.NewGauge(prometheus.GaugeOpts{
		Name: "prometheus_tsdb_head_series",
		Help: "Total number of series in the head block.",
	})
	m.seriesCreated = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_series_created_total",
		Help: "Total number of series created in the head",
	})
	m.seriesRemoved = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_series_removed_total",
		Help: "Total number of series removed in the head",
	})
	m.seriesNotFound = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_series_not_found_total",
		Help: "Total number of requests for series that were not found.",
	})
	m.chunks = prometheus.NewGauge(prometheus.GaugeOpts{
		Name: "prometheus_tsdb_head_chunks",
		Help: "Total number of chunks in the head block.",
	})
	m.chunksCreated = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_chunks_created_total",
		Help: "Total number of chunks created in the head",
	})
	m.chunksRemoved = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_chunks_removed_total",
		Help: "Total number of chunks removed in the head",
	})
	m.gcDuration = prometheus.NewSummary(prometheus.SummaryOpts{
		Name: "prometheus_tsdb_head_gc_duration_seconds",
		Help: "Runtime of garbage collection in the head block.",
	})
	m.maxTime = prometheus.NewGaugeFunc(prometheus.GaugeOpts{
		Name: "prometheus_tsdb_head_max_time",
		Help: "Maximum timestamp of the head block.",
	}, func() float64 {
		return float64(h.MaxTime())
	})
	m.minTime = prometheus.NewGaugeFunc(prometheus.GaugeOpts{
		Name: "prometheus_tsdb_head_min_time",
		Help: "Minimum time bound of the head block.",
	}, func() float64 {
		return float64(h.MinTime())
	})
	m.walTruncateDuration = prometheus.NewSummary(prometheus.SummaryOpts{
		Name: "prometheus_tsdb_wal_truncate_duration_seconds",
		Help: "Duration of WAL truncation.",
	})
	m.samplesAppended = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_samples_appended_total",
		Help: "Total number of appended samples.",
	})
	m.headTruncateFail = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_truncations_failed_total",
		Help: "Total number of head truncations that failed.",
	})
	m.headTruncateTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_head_truncations_total",
		Help: "Total number of head truncations attempted.",
	})
	m.checkpointDeleteFail = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_checkpoint_deletions_failed_total",
		Help: "Total number of checkpoint deletions that failed.",
	})
	m.checkpointDeleteTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_checkpoint_deletions_total",
		Help: "Total number of checkpoint deletions attempted.",
	})
	m.checkpointCreationFail = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_checkpoint_creations_failed_total",
		Help: "Total number of checkpoint creations that failed.",
	})
	m.checkpointCreationTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Name: "prometheus_tsdb_checkpoint_creations_total",
		Help: "Total number of checkpoint creations attempted.",
	})

	if r != nil {
		r.MustRegister(
			m.activeAppenders,
			m.chunks,
			m.chunksCreated,
			m.chunksRemoved,
			m.series,
			m.seriesCreated,
			m.seriesRemoved,
			m.seriesNotFound,
			m.minTime,
			m.maxTime,
			m.gcDuration,
			m.walTruncateDuration,
			m.samplesAppended,
			m.headTruncateFail,
			m.headTruncateTotal,
			m.checkpointDeleteFail,
			m.checkpointDeleteTotal,
			m.checkpointCreationFail,
			m.checkpointCreationTotal,
		)
	}
	return m
}

// NewHead opens the head block in dir.
func NewHead(r prometheus.Registerer, l log.Logger, wal *wal.WAL, chunkRange int64) (*Head, error) {
	if l == nil {
		l = log.NewNopLogger()
	}
	if chunkRange < 1 {
		return nil, errors.Errorf("invalid chunk range %d", chunkRange)
	}
	h := &Head{
		wal:        wal,
		logger:     l,
		chunkRange: chunkRange,
		minTime:    math.MaxInt64,
		maxTime:    math.MinInt64,
		series:     newStripeSeries(),
		values:     map[string]stringset{},
		symbols:    map[string]struct{}{},
		postings:   index.NewUnorderedMemPostings(),
		tombstones: newMemTombstones(),
	}
	h.metrics = newHeadMetrics(h, r)

	return h, nil
}

// processWALSamples adds a partition of samples it receives to the head and passes
// them on to other workers.
// Samples before the mint timestamp are discarded.
func (h *Head) processWALSamples(
	minValidTime int64,
	input <-chan []RefSample, output chan<- []RefSample,
) (unknownRefs uint64) {
	defer close(output)

	// Mitigate lock contention in getByID.
	refSeries := map[uint64]*memSeries{}

	mint, maxt := int64(math.MaxInt64), int64(math.MinInt64)

	for samples := range input {
		for _, s := range samples {
			if s.T < minValidTime {
				continue
			}
			ms := refSeries[s.Ref]
			if ms == nil {
				ms = h.series.getByID(s.Ref)
				if ms == nil {
					unknownRefs++
					continue
				}
				refSeries[s.Ref] = ms
			}
			_, chunkCreated := ms.append(s.T, s.V)
			if chunkCreated {
				h.metrics.chunksCreated.Inc()
				h.metrics.chunks.Inc()
			}
			if s.T > maxt {
				maxt = s.T
			}
			if s.T < mint {
				mint = s.T
			}
		}
		output <- samples
	}
	h.updateMinMaxTime(mint, maxt)

	return unknownRefs
}

func (h *Head) updateMinMaxTime(mint, maxt int64) {
	for {
		lt := h.MinTime()
		if mint >= lt {
			break
		}
		if atomic.CompareAndSwapInt64(&h.minTime, lt, mint) {
			break
		}
	}
	for {
		ht := h.MaxTime()
		if maxt <= ht {
			break
		}
		if atomic.CompareAndSwapInt64(&h.maxTime, ht, maxt) {
			break
		}
	}
}

func (h *Head) loadWAL(r *wal.Reader) error {
	minValidTime := h.MinTime()
	// If the min time is still uninitialized (no persisted blocks yet),
	// we accept all sample timestamps from the WAL.
	if minValidTime == math.MaxInt64 {
		minValidTime = math.MinInt64
	}

	// Track number of samples that referenced a series we don't know about
	// for error reporting.
	var unknownRefs uint64

	// Start workers that each process samples for a partition of the series ID space.
	// They are connected through a ring of channels which ensures that all sample batches
	// read from the WAL are processed in order.
	var (
		wg      sync.WaitGroup
		n       = runtime.GOMAXPROCS(0)
		inputs  = make([]chan []RefSample, n)
		outputs = make([]chan []RefSample, n)
	)
	wg.Add(n)

	for i := 0; i < n; i++ {
		outputs[i] = make(chan []RefSample, 300)
		inputs[i] = make(chan []RefSample, 300)

		go func(input <-chan []RefSample, output chan<- []RefSample) {
			unknown := h.processWALSamples(minValidTime, input, output)
			atomic.AddUint64(&unknownRefs, unknown)
			wg.Done()
		}(inputs[i], outputs[i])
	}

	var (
		dec     RecordDecoder
		series  []RefSeries
		samples []RefSample
		tstones []Stone
		err     error
	)
	for r.Next() {
		series, samples, tstones = series[:0], samples[:0], tstones[:0]
		rec := r.Record()

		switch dec.Type(rec) {
		case RecordSeries:
			series, err = dec.Series(rec, series)
			if err != nil {
				return &wal.CorruptionErr{
					Err:     errors.Wrap(err, "decode series"),
					Segment: r.Segment(),
					Offset:  r.Offset(),
				}
			}
			for _, s := range series {
				h.getOrCreateWithID(s.Ref, s.Labels.Hash(), s.Labels)

				if h.lastSeriesID < s.Ref {
					h.lastSeriesID = s.Ref
				}
			}
		case RecordSamples:
			samples, err = dec.Samples(rec, samples)
			s := samples
			if err != nil {
				return &wal.CorruptionErr{
					Err:     errors.Wrap(err, "decode samples"),
					Segment: r.Segment(),
					Offset:  r.Offset(),
				}
			}
			// We split up the samples into chunks of 5000 samples or less.
			// With O(300 * #cores) in-flight sample batches, large scrapes could otherwise
			// cause thousands of very large in flight buffers occupying large amounts
			// of unused memory.
			for len(samples) > 0 {
				m := 5000
				if len(samples) < m {
					m = len(samples)
				}
				shards := make([][]RefSample, n)
				for i := 0; i < n; i++ {
					var buf []RefSample
					select {
					case buf = <-outputs[i]:
					default:
					}
					shards[i] = buf[:0]
				}
				for _, sam := range samples[:m] {
					mod := sam.Ref % uint64(n)
					shards[mod] = append(shards[mod], sam)
				}
				for i := 0; i < n; i++ {
					inputs[i] <- shards[i]
				}
				samples = samples[m:]
			}
			samples = s // Keep whole slice for reuse.
		case RecordTombstones:
			tstones, err = dec.Tombstones(rec, tstones)
			if err != nil {
				return &wal.CorruptionErr{
					Err:     errors.Wrap(err, "decode tombstones"),
					Segment: r.Segment(),
					Offset:  r.Offset(),
				}
			}
			for _, s := range tstones {
				for _, itv := range s.intervals {
					if itv.Maxt < minValidTime {
						continue
					}
					h.tombstones.addInterval(s.ref, itv)
				}
			}
		default:
			return &wal.CorruptionErr{
				Err:     errors.Errorf("invalid record type %v", dec.Type(rec)),
				Segment: r.Segment(),
				Offset:  r.Offset(),
			}
		}
	}
	if r.Err() != nil {
		return errors.Wrap(r.Err(), "read records")
	}

	// Signal termination to each worker and wait for it to close its output channel.
	for i := 0; i < n; i++ {
		close(inputs[i])
		for range outputs[i] {
		}
	}
	wg.Wait()

	if unknownRefs > 0 {
		level.Warn(h.logger).Log("msg", "unknown series references", "count", unknownRefs)
	}
	return nil
}

// Init loads data from the write ahead log and prepares the head for writes.
func (h *Head) Init() error {
	defer h.postings.EnsureOrder()

	if h.wal == nil {
		return nil
	}

	// Backfill the checkpoint first if it exists.
	dir, startFrom, err := LastCheckpoint(h.wal.Dir())
	if err != nil && err != ErrNotFound {
		return errors.Wrap(err, "find last checkpoint")
	}
	if err == nil {
		sr, err := wal.NewSegmentsReader(dir)
		if err != nil {
			return errors.Wrap(err, "open checkpoint")
		}
		defer sr.Close()

		// A corrupted checkpoint is a hard error for now and requires user
		// intervention. There's likely little data that can be recovered anyway.
		if err := h.loadWAL(wal.NewReader(sr)); err != nil {
			return errors.Wrap(err, "backfill checkpoint")
		}
		startFrom++
	}

	// Backfill segments from the last checkpoint onwards
	sr, err := wal.NewSegmentsRangeReader(wal.SegmentRange{Dir: h.wal.Dir(), First: startFrom, Last: -1})
	if err != nil {
		return errors.Wrap(err, "open WAL segments")
	}
	defer sr.Close()

	err = h.loadWAL(wal.NewReader(sr))
	if err == nil {
		return nil
	}
	level.Warn(h.logger).Log("msg", "encountered WAL error, attempting repair", "err", err)

	if err := h.wal.Repair(err); err != nil {
		return errors.Wrap(err, "repair corrupted WAL")
	}
	return nil
}

// Truncate removes old data before mint from the head.
func (h *Head) Truncate(mint int64) (err error) {
	defer func() {
		if err != nil {
			h.metrics.headTruncateFail.Inc()
		}
	}()
	initialize := h.MinTime() == math.MaxInt64

	if h.MinTime() >= mint && !initialize {
		return nil
	}
	atomic.StoreInt64(&h.minTime, mint)

	// Ensure that max time is at least as high as min time.
	for h.MaxTime() < mint {
		atomic.CompareAndSwapInt64(&h.maxTime, h.MaxTime(), mint)
	}

	// This was an initial call to Truncate after loading blocks on startup.
	// We haven't read back the WAL yet, so do not attempt to truncate it.
	if initialize {
		return nil
	}

	h.metrics.headTruncateTotal.Inc()
	start := time.Now()

	h.gc()
	level.Info(h.logger).Log("msg", "head GC completed", "duration", time.Since(start))
	h.metrics.gcDuration.Observe(time.Since(start).Seconds())

	if h.wal == nil {
		return nil
	}
	start = time.Now()

	first, last, err := h.wal.Segments()
	if err != nil {
		return errors.Wrap(err, "get segment range")
	}
	last-- // Never consider last segment for checkpoint.
	if last < 0 {
		return nil // no segments yet.
	}
	// The lower third of segments should contain mostly obsolete samples.
	// If we have less than three segments, it's not worth checkpointing yet.
	last = first + (last-first)/3
	if last <= first {
		return nil
	}

	keep := func(id uint64) bool {
		return h.series.getByID(id) != nil
	}
	h.metrics.checkpointCreationTotal.Inc()
	if _, err = Checkpoint(h.wal, first, last, keep, mint); err != nil {
		h.metrics.checkpointCreationFail.Inc()
		return errors.Wrap(err, "create checkpoint")
	}
	if err := h.wal.Truncate(last + 1); err != nil {
		// If truncating fails, we'll just try again at the next checkpoint.
		// Leftover segments will just be ignored in the future if there's a checkpoint
		// that supersedes them.
		level.Error(h.logger).Log("msg", "truncating segments failed", "err", err)
	}
	h.metrics.checkpointDeleteTotal.Inc()
	if err := DeleteCheckpoints(h.wal.Dir(), last); err != nil {
		// Leftover old checkpoints do not cause problems down the line beyond
		// occupying disk space.
		// They will just be ignored since a higher checkpoint exists.
		level.Error(h.logger).Log("msg", "delete old checkpoints", "err", err)
		h.metrics.checkpointDeleteFail.Inc()
	}
	h.metrics.walTruncateDuration.Observe(time.Since(start).Seconds())

	level.Info(h.logger).Log("msg", "WAL checkpoint complete",
		"first", first, "last", last, "duration", time.Since(start))

	return nil
}

// initTime initializes a head with the first timestamp. This only needs to be called
// for a compltely fresh head with an empty WAL.
// Returns true if the initialization took an effect.
func (h *Head) initTime(t int64) (initialized bool) {
	if !atomic.CompareAndSwapInt64(&h.minTime, math.MaxInt64, t) {
		return false
	}
	// Ensure that max time is initialized to at least the min time we just set.
	// Concurrent appenders may already have set it to a higher value.
	atomic.CompareAndSwapInt64(&h.maxTime, math.MinInt64, t)

	return true
}

type rangeHead struct {
	head       *Head
	mint, maxt int64
}

func (h *rangeHead) Index() (IndexReader, error) {
	return h.head.indexRange(h.mint, h.maxt), nil
}

func (h *rangeHead) Chunks() (ChunkReader, error) {
	return h.head.chunksRange(h.mint, h.maxt), nil
}

func (h *rangeHead) Tombstones() (TombstoneReader, error) {
	return h.head.tombstones, nil
}

// initAppender is a helper to initialize the time bounds of the head
// upon the first sample it receives.
type initAppender struct {
	app  Appender
	head *Head
}

func (a *initAppender) Add(lset labels.Labels, t int64, v float64) (uint64, error) {
	if a.app != nil {
		return a.app.Add(lset, t, v)
	}
	a.head.initTime(t)
	a.app = a.head.appender()

	return a.app.Add(lset, t, v)
}

func (a *initAppender) AddFast(ref uint64, t int64, v float64) error {
	if a.app == nil {
		return ErrNotFound
	}
	return a.app.AddFast(ref, t, v)
}

func (a *initAppender) Commit() error {
	if a.app == nil {
		return nil
	}
	return a.app.Commit()
}

func (a *initAppender) Rollback() error {
	if a.app == nil {
		return nil
	}
	return a.app.Rollback()
}

// Appender returns a new Appender on the database.
func (h *Head) Appender() Appender {
	h.metrics.activeAppenders.Inc()

	// The head cache might not have a starting point yet. The init appender
	// picks up the first appended timestamp as the base.
	if h.MinTime() == math.MaxInt64 {
		return &initAppender{head: h}
	}
	return h.appender()
}

func (h *Head) appender() *headAppender {
	return &headAppender{
		head:         h,
		minValidTime: h.MaxTime() - h.chunkRange/2,
		mint:         math.MaxInt64,
		maxt:         math.MinInt64,
		samples:      h.getAppendBuffer(),
	}
}

func (h *Head) getAppendBuffer() []RefSample {
	b := h.appendPool.Get()
	if b == nil {
		return make([]RefSample, 0, 512)
	}
	return b.([]RefSample)
}

func (h *Head) putAppendBuffer(b []RefSample) {
	h.appendPool.Put(b[:0])
}

func (h *Head) getBytesBuffer() []byte {
	b := h.bytesPool.Get()
	if b == nil {
		return make([]byte, 0, 1024)
	}
	return b.([]byte)
}

func (h *Head) putBytesBuffer(b []byte) {
	h.bytesPool.Put(b[:0])
}

type headAppender struct {
	head         *Head
	minValidTime int64 // No samples below this timestamp are allowed.
	mint, maxt   int64

	series  []RefSeries
	samples []RefSample
}

func (a *headAppender) Add(lset labels.Labels, t int64, v float64) (uint64, error) {
	if t < a.minValidTime {
		return 0, ErrOutOfBounds
	}

	s, created := a.head.getOrCreate(lset.Hash(), lset)
	if created {
		a.series = append(a.series, RefSeries{
			Ref:    s.ref,
			Labels: lset,
		})
	}
	return s.ref, a.AddFast(s.ref, t, v)
}

func (a *headAppender) AddFast(ref uint64, t int64, v float64) error {
	if t < a.minValidTime {
		return ErrOutOfBounds
	}

	s := a.head.series.getByID(ref)
	if s == nil {
		return errors.Wrap(ErrNotFound, "unknown series")
	}
	s.Lock()
	if err := s.appendable(t, v); err != nil {
		s.Unlock()
		return err
	}
	s.pendingCommit = true
	s.Unlock()

	if t < a.mint {
		a.mint = t
	}
	if t > a.maxt {
		a.maxt = t
	}

	a.samples = append(a.samples, RefSample{
		Ref:    ref,
		T:      t,
		V:      v,
		series: s,
	})
	return nil
}

func (a *headAppender) log() error {
	if a.head.wal == nil {
		return nil
	}

	buf := a.head.getBytesBuffer()
	defer func() { a.head.putBytesBuffer(buf) }()

	var rec []byte
	var enc RecordEncoder

	if len(a.series) > 0 {
		rec = enc.Series(a.series, buf)
		buf = rec[:0]

		if err := a.head.wal.Log(rec); err != nil {
			return errors.Wrap(err, "log series")
		}
	}
	if len(a.samples) > 0 {
		rec = enc.Samples(a.samples, buf)
		buf = rec[:0]

		if err := a.head.wal.Log(rec); err != nil {
			return errors.Wrap(err, "log samples")
		}
	}
	return nil
}

func (a *headAppender) Commit() error {
	defer a.head.metrics.activeAppenders.Dec()
	defer a.head.putAppendBuffer(a.samples)

	if err := a.log(); err != nil {
		return errors.Wrap(err, "write to WAL")
	}

	total := len(a.samples)

	for _, s := range a.samples {
		s.series.Lock()
		ok, chunkCreated := s.series.append(s.T, s.V)
		s.series.pendingCommit = false
		s.series.Unlock()

		if !ok {
			total--
		}
		if chunkCreated {
			a.head.metrics.chunks.Inc()
			a.head.metrics.chunksCreated.Inc()
		}
	}

	a.head.metrics.samplesAppended.Add(float64(total))
	a.head.updateMinMaxTime(a.mint, a.maxt)

	return nil
}

func (a *headAppender) Rollback() error {
	a.head.metrics.activeAppenders.Dec()
	for _, s := range a.samples {
		s.series.Lock()
		s.series.pendingCommit = false
		s.series.Unlock()
	}
	a.head.putAppendBuffer(a.samples)

	// Series are created in the head memory regardless of rollback. Thus we have
	// to log them to the WAL in any case.
	a.samples = nil
	return a.log()
}

// Delete all samples in the range of [mint, maxt] for series that satisfy the given
// label matchers.
func (h *Head) Delete(mint, maxt int64, ms ...labels.Matcher) error {
	// Do not delete anything beyond the currently valid range.
	mint, maxt = clampInterval(mint, maxt, h.MinTime(), h.MaxTime())

	ir := h.indexRange(mint, maxt)

	p, err := PostingsForMatchers(ir, ms...)
	if err != nil {
		return errors.Wrap(err, "select series")
	}

	var stones []Stone

	for p.Next() {
		series := h.series.getByID(p.At())

		t0, t1 := series.minTime(), series.maxTime()
		if t0 == math.MinInt64 || t1 == math.MinInt64 {
			continue
		}
		// Delete only until the current values and not beyond.
		t0, t1 = clampInterval(mint, maxt, t0, t1)
		stones = append(stones, Stone{p.At(), Intervals{{t0, t1}}})
	}

	if p.Err() != nil {
		return p.Err()
	}
	var enc RecordEncoder

	if h.wal != nil {
		if err := h.wal.Log(enc.Tombstones(stones, nil)); err != nil {
			return err
		}
	}
	for _, s := range stones {
		h.tombstones.addInterval(s.ref, s.intervals[0])
	}
	return nil
}

// gc removes data before the minimum timestamp from the head.
func (h *Head) gc() {
	// Only data strictly lower than this timestamp must be deleted.
	mint := h.MinTime()

	// Drop old chunks and remember series IDs and hashes if they can be
	// deleted entirely.
	deleted, chunksRemoved := h.series.gc(mint)
	seriesRemoved := len(deleted)

	h.metrics.seriesRemoved.Add(float64(seriesRemoved))
	h.metrics.series.Sub(float64(seriesRemoved))
	h.metrics.chunksRemoved.Add(float64(chunksRemoved))
	h.metrics.chunks.Sub(float64(chunksRemoved))

	// Remove deleted series IDs from the postings lists.
	h.postings.Delete(deleted)

	// Rebuild symbols and label value indices from what is left in the postings terms.
	symbols := make(map[string]struct{})
	values := make(map[string]stringset, len(h.values))

	if err := h.postings.Iter(func(t labels.Label, _ index.Postings) error {
		symbols[t.Name] = struct{}{}
		symbols[t.Value] = struct{}{}

		ss, ok := values[t.Name]
		if !ok {
			ss = stringset{}
			values[t.Name] = ss
		}
		ss.set(t.Value)
		return nil
	}); err != nil {
		// This should never happen, as the iteration function only returns nil.
		panic(err)
	}

	h.symMtx.Lock()

	h.symbols = symbols
	h.values = values

	h.symMtx.Unlock()
}

// Tombstones returns a new reader over the head's tombstones
func (h *Head) Tombstones() (TombstoneReader, error) {
	return h.tombstones, nil
}

// Index returns an IndexReader against the block.
func (h *Head) Index() (IndexReader, error) {
	return h.indexRange(math.MinInt64, math.MaxInt64), nil
}

func (h *Head) indexRange(mint, maxt int64) *headIndexReader {
	if hmin := h.MinTime(); hmin > mint {
		mint = hmin
	}
	return &headIndexReader{head: h, mint: mint, maxt: maxt}
}

// Chunks returns a ChunkReader against the block.
func (h *Head) Chunks() (ChunkReader, error) {
	return h.chunksRange(math.MinInt64, math.MaxInt64), nil
}

func (h *Head) chunksRange(mint, maxt int64) *headChunkReader {
	if hmin := h.MinTime(); hmin > mint {
		mint = hmin
	}
	return &headChunkReader{head: h, mint: mint, maxt: maxt}
}

// MinTime returns the lowest time bound on visible data in the head.
func (h *Head) MinTime() int64 {
	return atomic.LoadInt64(&h.minTime)
}

// MaxTime returns the highest timestamp seen in data of the head.
func (h *Head) MaxTime() int64 {
	return atomic.LoadInt64(&h.maxTime)
}

// Close flushes the WAL and closes the head.
func (h *Head) Close() error {
	if h.wal == nil {
		return nil
	}
	return h.wal.Close()
}

type headChunkReader struct {
	head       *Head
	mint, maxt int64
}

func (h *headChunkReader) Close() error {
	return nil
}

// packChunkID packs a seriesID and a chunkID within it into a global 8 byte ID.
// It panicks if the seriesID exceeds 5 bytes or the chunk ID 3 bytes.
func packChunkID(seriesID, chunkID uint64) uint64 {
	if seriesID > (1<<40)-1 {
		panic("series ID exceeds 5 bytes")
	}
	if chunkID > (1<<24)-1 {
		panic("chunk ID exceeds 3 bytes")
	}
	return (seriesID << 24) | chunkID
}

func unpackChunkID(id uint64) (seriesID, chunkID uint64) {
	return id >> 24, (id << 40) >> 40
}

// Chunk returns the chunk for the reference number.
func (h *headChunkReader) Chunk(ref uint64) (chunkenc.Chunk, error) {
	sid, cid := unpackChunkID(ref)

	s := h.head.series.getByID(sid)
	// This means that the series has been garbage collected.
	if s == nil {
		return nil, ErrNotFound
	}

	s.Lock()
	c := s.chunk(int(cid))

	// This means that the chunk has been garbage collected or is outside
	// the specified range.
	if c == nil || !c.OverlapsClosedInterval(h.mint, h.maxt) {
		s.Unlock()
		return nil, ErrNotFound
	}
	s.Unlock()

	return &safeChunk{
		Chunk: c.chunk,
		s:     s,
		cid:   int(cid),
	}, nil
}

type safeChunk struct {
	chunkenc.Chunk
	s   *memSeries
	cid int
}

func (c *safeChunk) Iterator() chunkenc.Iterator {
	c.s.Lock()
	it := c.s.iterator(c.cid)
	c.s.Unlock()
	return it
}

type headIndexReader struct {
	head       *Head
	mint, maxt int64
}

func (h *headIndexReader) Close() error {
	return nil
}

func (h *headIndexReader) Symbols() (map[string]struct{}, error) {
	h.head.symMtx.RLock()
	defer h.head.symMtx.RUnlock()

	res := make(map[string]struct{}, len(h.head.symbols))

	for s := range h.head.symbols {
		res[s] = struct{}{}
	}
	return res, nil
}

// LabelValues returns the possible label values
func (h *headIndexReader) LabelValues(names ...string) (index.StringTuples, error) {
	if len(names) != 1 {
		return nil, errInvalidSize
	}

	h.head.symMtx.RLock()
	sl := make([]string, 0, len(h.head.values[names[0]]))
	for s := range h.head.values[names[0]] {
		sl = append(sl, s)
	}
	h.head.symMtx.RUnlock()
	sort.Strings(sl)

	return index.NewStringTuples(sl, len(names))
}

// LabelNames returns all the unique label names present in the head.
func (h *headIndexReader) LabelNames() ([]string, error) {
	h.head.symMtx.RLock()
	defer h.head.symMtx.RUnlock()
	labelNames := make([]string, 0, len(h.head.values))
	for name := range h.head.values {
		if name == "" {
			continue
		}
		labelNames = append(labelNames, name)
	}
	sort.Strings(labelNames)
	return labelNames, nil
}

// Postings returns the postings list iterator for the label pair.
func (h *headIndexReader) Postings(name, value string) (index.Postings, error) {
	return h.head.postings.Get(name, value), nil
}

func (h *headIndexReader) SortedPostings(p index.Postings) index.Postings {
	ep := make([]uint64, 0, 128)

	for p.Next() {
		ep = append(ep, p.At())
	}
	if err := p.Err(); err != nil {
		return index.ErrPostings(errors.Wrap(err, "expand postings"))
	}

	sort.Slice(ep, func(i, j int) bool {
		a := h.head.series.getByID(ep[i])
		b := h.head.series.getByID(ep[j])

		if a == nil || b == nil {
			level.Debug(h.head.logger).Log("msg", "looked up series not found")
			return false
		}
		return labels.Compare(a.lset, b.lset) < 0
	})
	return index.NewListPostings(ep)
}

// Series returns the series for the given reference.
func (h *headIndexReader) Series(ref uint64, lbls *labels.Labels, chks *[]chunks.Meta) error {
	s := h.head.series.getByID(ref)

	if s == nil {
		h.head.metrics.seriesNotFound.Inc()
		return ErrNotFound
	}
	*lbls = append((*lbls)[:0], s.lset...)

	s.Lock()
	defer s.Unlock()

	*chks = (*chks)[:0]

	for i, c := range s.chunks {
		// Do not expose chunks that are outside of the specified range.
		if !c.OverlapsClosedInterval(h.mint, h.maxt) {
			continue
		}
		*chks = append(*chks, chunks.Meta{
			MinTime: c.minTime,
			MaxTime: c.maxTime,
			Ref:     packChunkID(s.ref, uint64(s.chunkID(i))),
		})
	}

	return nil
}

func (h *headIndexReader) LabelIndices() ([][]string, error) {
	h.head.symMtx.RLock()
	defer h.head.symMtx.RUnlock()
	res := [][]string{}
	for s := range h.head.values {
		res = append(res, []string{s})
	}
	return res, nil
}

func (h *Head) getOrCreate(hash uint64, lset labels.Labels) (*memSeries, bool) {
	// Just using `getOrSet` below would be semantically sufficient, but we'd create
	// a new series on every sample inserted via Add(), which causes allocations
	// and makes our series IDs rather random and harder to compress in postings.
	s := h.series.getByHash(hash, lset)
	if s != nil {
		return s, false
	}

	// Optimistically assume that we are the first one to create the series.
	id := atomic.AddUint64(&h.lastSeriesID, 1)

	return h.getOrCreateWithID(id, hash, lset)
}

func (h *Head) getOrCreateWithID(id, hash uint64, lset labels.Labels) (*memSeries, bool) {
	s := newMemSeries(lset, id, h.chunkRange)

	s, created := h.series.getOrSet(hash, s)
	if !created {
		return s, false
	}

	h.metrics.series.Inc()
	h.metrics.seriesCreated.Inc()

	h.postings.Add(id, lset)

	h.symMtx.Lock()
	defer h.symMtx.Unlock()

	for _, l := range lset {
		valset, ok := h.values[l.Name]
		if !ok {
			valset = stringset{}
			h.values[l.Name] = valset
		}
		valset.set(l.Value)

		h.symbols[l.Name] = struct{}{}
		h.symbols[l.Value] = struct{}{}
	}

	return s, true
}

// seriesHashmap is a simple hashmap for memSeries by their label set. It is built
// on top of a regular hashmap and holds a slice of series to resolve hash collisions.
// Its methods require the hash to be submitted with it to avoid re-computations throughout
// the code.
type seriesHashmap map[uint64][]*memSeries

func (m seriesHashmap) get(hash uint64, lset labels.Labels) *memSeries {
	for _, s := range m[hash] {
		if s.lset.Equals(lset) {
			return s
		}
	}
	return nil
}

func (m seriesHashmap) set(hash uint64, s *memSeries) {
	l := m[hash]
	for i, prev := range l {
		if prev.lset.Equals(s.lset) {
			l[i] = s
			return
		}
	}
	m[hash] = append(l, s)
}

func (m seriesHashmap) del(hash uint64, lset labels.Labels) {
	var rem []*memSeries
	for _, s := range m[hash] {
		if !s.lset.Equals(lset) {
			rem = append(rem, s)
		}
	}
	if len(rem) == 0 {
		delete(m, hash)
	} else {
		m[hash] = rem
	}
}

// stripeSeries locks modulo ranges of IDs and hashes to reduce lock contention.
// The locks are padded to not be on the same cache line. Filling the padded space
// with the maps was profiled to be slower – likely due to the additional pointer
// dereferences.
type stripeSeries struct {
	series [stripeSize]map[uint64]*memSeries
	hashes [stripeSize]seriesHashmap
	locks  [stripeSize]stripeLock
}

const (
	stripeSize = 1 << 14
	stripeMask = stripeSize - 1
)

type stripeLock struct {
	sync.RWMutex
	// Padding to avoid multiple locks being on the same cache line.
	_ [40]byte
}

func newStripeSeries() *stripeSeries {
	s := &stripeSeries{}

	for i := range s.series {
		s.series[i] = map[uint64]*memSeries{}
	}
	for i := range s.hashes {
		s.hashes[i] = seriesHashmap{}
	}
	return s
}

// gc garbage collects old chunks that are strictly before mint and removes
// series entirely that have no chunks left.
func (s *stripeSeries) gc(mint int64) (map[uint64]struct{}, int) {
	var (
		deleted  = map[uint64]struct{}{}
		rmChunks = 0
	)
	// Run through all series and truncate old chunks. Mark those with no
	// chunks left as deleted and store their ID.
	for i := 0; i < stripeSize; i++ {
		s.locks[i].Lock()

		for hash, all := range s.hashes[i] {
			for _, series := range all {
				series.Lock()
				rmChunks += series.truncateChunksBefore(mint)

				if len(series.chunks) > 0 || series.pendingCommit {
					series.Unlock()
					continue
				}

				// The series is gone entirely. We need to keep the series lock
				// and make sure we have acquired the stripe locks for hash and ID of the
				// series alike.
				// If we don't hold them all, there's a very small chance that a series receives
				// samples again while we are half-way into deleting it.
				j := int(series.ref & stripeMask)

				if i != j {
					s.locks[j].Lock()
				}

				deleted[series.ref] = struct{}{}
				s.hashes[i].del(hash, series.lset)
				delete(s.series[j], series.ref)

				if i != j {
					s.locks[j].Unlock()
				}

				series.Unlock()
			}
		}

		s.locks[i].Unlock()
	}

	return deleted, rmChunks
}

func (s *stripeSeries) getByID(id uint64) *memSeries {
	i := id & stripeMask

	s.locks[i].RLock()
	series := s.series[i][id]
	s.locks[i].RUnlock()

	return series
}

func (s *stripeSeries) getByHash(hash uint64, lset labels.Labels) *memSeries {
	i := hash & stripeMask

	s.locks[i].RLock()
	series := s.hashes[i].get(hash, lset)
	s.locks[i].RUnlock()

	return series
}

func (s *stripeSeries) getOrSet(hash uint64, series *memSeries) (*memSeries, bool) {
	i := hash & stripeMask

	s.locks[i].Lock()

	if prev := s.hashes[i].get(hash, series.lset); prev != nil {
		s.locks[i].Unlock()
		return prev, false
	}
	s.hashes[i].set(hash, series)
	s.locks[i].Unlock()

	i = series.ref & stripeMask

	s.locks[i].Lock()
	s.series[i][series.ref] = series
	s.locks[i].Unlock()

	return series, true
}

type sample struct {
	t int64
	v float64
}

func (s sample) T() int64 {
	return s.t
}

func (s sample) V() float64 {
	return s.v
}

// memSeries is the in-memory representation of a series. None of its methods
// are goroutine safe and it is the caller's responsibility to lock it.
type memSeries struct {
	sync.Mutex

	ref          uint64
	lset         labels.Labels
	chunks       []*memChunk
	headChunk    *memChunk
	chunkRange   int64
	firstChunkID int

	nextAt        int64 // Timestamp at which to cut the next chunk.
	sampleBuf     [4]sample
	pendingCommit bool // Whether there are samples waiting to be committed to this series.

	app chunkenc.Appender // Current appender for the chunk.
}

func (s *memSeries) minTime() int64 {
	if len(s.chunks) == 0 {
		return math.MinInt64
	}
	return s.chunks[0].minTime
}

func (s *memSeries) maxTime() int64 {
	c := s.head()
	if c == nil {
		return math.MinInt64
	}
	return c.maxTime
}

func (s *memSeries) cut(mint int64) *memChunk {
	c := &memChunk{
		chunk:   chunkenc.NewXORChunk(),
		minTime: mint,
		maxTime: math.MinInt64,
	}
	s.chunks = append(s.chunks, c)
	s.headChunk = c

	// Set upper bound on when the next chunk must be started. An earlier timestamp
	// may be chosen dynamically at a later point.
	_, s.nextAt = rangeForTimestamp(mint, s.chunkRange)

	app, err := c.chunk.Appender()
	if err != nil {
		panic(err)
	}
	s.app = app
	return c
}

func newMemSeries(lset labels.Labels, id uint64, chunkRange int64) *memSeries {
	s := &memSeries{
		lset:       lset,
		ref:        id,
		chunkRange: chunkRange,
		nextAt:     math.MinInt64,
	}
	return s
}

// appendable checks whether the given sample is valid for appending to the series.
func (s *memSeries) appendable(t int64, v float64) error {
	c := s.head()
	if c == nil {
		return nil
	}

	if t > c.maxTime {
		return nil
	}
	if t < c.maxTime {
		return ErrOutOfOrderSample
	}
	// We are allowing exact duplicates as we can encounter them in valid cases
	// like federation and erroring out at that time would be extremely noisy.
	if math.Float64bits(s.sampleBuf[3].v) != math.Float64bits(v) {
		return ErrAmendSample
	}
	return nil
}

func (s *memSeries) chunk(id int) *memChunk {
	ix := id - s.firstChunkID
	if ix < 0 || ix >= len(s.chunks) {
		return nil
	}
	return s.chunks[ix]
}

func (s *memSeries) chunkID(pos int) int {
	return pos + s.firstChunkID
}

// truncateChunksBefore removes all chunks from the series that have not timestamp
// at or after mint. Chunk IDs remain unchanged.
func (s *memSeries) truncateChunksBefore(mint int64) (removed int) {
	var k int
	for i, c := range s.chunks {
		if c.maxTime >= mint {
			break
		}
		k = i + 1
	}
	s.chunks = append(s.chunks[:0], s.chunks[k:]...)
	s.firstChunkID += k
	if len(s.chunks) == 0 {
		s.headChunk = nil
	} else {
		s.headChunk = s.chunks[len(s.chunks)-1]
	}

	return k
}

// append adds the sample (t, v) to the series.
func (s *memSeries) append(t int64, v float64) (success, chunkCreated bool) {
	// Based on Gorilla white papers this offers near-optimal compression ratio
	// so anything bigger that this has diminishing returns and increases
	// the time range within which we have to decompress all samples.
	const samplesPerChunk = 120

	c := s.head()

	if c == nil {
		c = s.cut(t)
		chunkCreated = true
	}
	numSamples := c.chunk.NumSamples()

	// Out of order sample.
	if c.maxTime >= t {
		return false, chunkCreated
	}
	// If we reach 25% of a chunk's desired sample count, set a definitive time
	// at which to start the next chunk.
	// At latest it must happen at the timestamp set when the chunk was cut.
	if numSamples == samplesPerChunk/4 {
		s.nextAt = computeChunkEndTime(c.minTime, c.maxTime, s.nextAt)
	}
	if t >= s.nextAt {
		c = s.cut(t)
		chunkCreated = true
	}
	s.app.Append(t, v)

	c.maxTime = t

	s.sampleBuf[0] = s.sampleBuf[1]
	s.sampleBuf[1] = s.sampleBuf[2]
	s.sampleBuf[2] = s.sampleBuf[3]
	s.sampleBuf[3] = sample{t: t, v: v}

	return true, chunkCreated
}

// computeChunkEndTime estimates the end timestamp based the beginning of a chunk,
// its current timestamp and the upper bound up to which we insert data.
// It assumes that the time range is 1/4 full.
func computeChunkEndTime(start, cur, max int64) int64 {
	a := (max - start) / ((cur - start + 1) * 4)
	if a == 0 {
		return max
	}
	return start + (max-start)/a
}

func (s *memSeries) iterator(id int) chunkenc.Iterator {
	c := s.chunk(id)
	// TODO(fabxc): Work around! A querier may have retrieved a pointer to a series' chunk,
	// which got then garbage collected before it got accessed.
	// We must ensure to not garbage collect as long as any readers still hold a reference.
	if c == nil {
		return chunkenc.NewNopIterator()
	}

	if id-s.firstChunkID < len(s.chunks)-1 {
		return c.chunk.Iterator()
	}
	// Serve the last 4 samples for the last chunk from the sample buffer
	// as their compressed bytes may be mutated by added samples.
	it := &memSafeIterator{
		Iterator: c.chunk.Iterator(),
		i:        -1,
		total:    c.chunk.NumSamples(),
		buf:      s.sampleBuf,
	}
	return it
}

func (s *memSeries) head() *memChunk {
	return s.headChunk
}

type memChunk struct {
	chunk            chunkenc.Chunk
	minTime, maxTime int64
}

// Returns true if the chunk overlaps [mint, maxt].
func (mc *memChunk) OverlapsClosedInterval(mint, maxt int64) bool {
	return mc.minTime <= maxt && mint <= mc.maxTime
}

type memSafeIterator struct {
	chunkenc.Iterator

	i     int
	total int
	buf   [4]sample
}

func (it *memSafeIterator) Next() bool {
	if it.i+1 >= it.total {
		return false
	}
	it.i++
	if it.total-it.i > 4 {
		return it.Iterator.Next()
	}
	return true
}

func (it *memSafeIterator) At() (int64, float64) {
	if it.total-it.i > 4 {
		return it.Iterator.At()
	}
	s := it.buf[4-(it.total-it.i)]
	return s.t, s.v
}

type stringset map[string]struct{}

func (ss stringset) set(s string) {
	ss[s] = struct{}{}
}

func (ss stringset) has(s string) bool {
	_, ok := ss[s]
	return ok
}

func (ss stringset) String() string {
	return strings.Join(ss.slice(), ",")
}

func (ss stringset) slice() []string {
	slice := make([]string, 0, len(ss))
	for k := range ss {
		slice = append(slice, k)
	}
	sort.Strings(slice)
	return slice
}