prometheus/storage/remote/queue_manager.go

// Copyright 2013 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 remote

import (
	"context"
	"errors"
	"math"
	"strconv"
	"sync"
	"time"

	"github.com/go-kit/log"
	"github.com/go-kit/log/level"
	"github.com/gogo/protobuf/proto"
	"github.com/golang/snappy"
	"github.com/opentracing/opentracing-go"
	"github.com/opentracing/opentracing-go/ext"
	"github.com/prometheus/client_golang/prometheus"
	"github.com/prometheus/common/model"
	"go.uber.org/atomic"

	"github.com/prometheus/prometheus/config"
	"github.com/prometheus/prometheus/model/labels"
	"github.com/prometheus/prometheus/model/relabel"
	"github.com/prometheus/prometheus/prompb"
	"github.com/prometheus/prometheus/scrape"
	"github.com/prometheus/prometheus/tsdb/chunks"
	"github.com/prometheus/prometheus/tsdb/record"
	"github.com/prometheus/prometheus/tsdb/wal"
)

const (
	// We track samples in/out and how long pushes take using an Exponentially
	// Weighted Moving Average.
	ewmaWeight          = 0.2
	shardUpdateDuration = 10 * time.Second

	// Allow 30% too many shards before scaling down.
	shardToleranceFraction = 0.3
)

type queueManagerMetrics struct {
	reg prometheus.Registerer

	samplesTotal          prometheus.Counter
	exemplarsTotal        prometheus.Counter
	metadataTotal         prometheus.Counter
	failedSamplesTotal    prometheus.Counter
	failedExemplarsTotal  prometheus.Counter
	failedMetadataTotal   prometheus.Counter
	retriedSamplesTotal   prometheus.Counter
	retriedExemplarsTotal prometheus.Counter
	retriedMetadataTotal  prometheus.Counter
	droppedSamplesTotal   prometheus.Counter
	droppedExemplarsTotal prometheus.Counter
	enqueueRetriesTotal   prometheus.Counter
	sentBatchDuration     prometheus.Histogram
	highestSentTimestamp  *maxTimestamp
	pendingSamples        prometheus.Gauge
	pendingExemplars      prometheus.Gauge
	shardCapacity         prometheus.Gauge
	numShards             prometheus.Gauge
	maxNumShards          prometheus.Gauge
	minNumShards          prometheus.Gauge
	desiredNumShards      prometheus.Gauge
	sentBytesTotal        prometheus.Counter
	metadataBytesTotal    prometheus.Counter
	maxSamplesPerSend     prometheus.Gauge
}

func newQueueManagerMetrics(r prometheus.Registerer, rn, e string) *queueManagerMetrics {
	m := &queueManagerMetrics{
		reg: r,
	}
	constLabels := prometheus.Labels{
		remoteName: rn,
		endpoint:   e,
	}

	m.samplesTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "samples_total",
		Help:        "Total number of samples sent to remote storage.",
		ConstLabels: constLabels,
	})
	m.exemplarsTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "exemplars_total",
		Help:        "Total number of exemplars sent to remote storage.",
		ConstLabels: constLabels,
	})
	m.metadataTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "metadata_total",
		Help:        "Total number of metadata entries sent to remote storage.",
		ConstLabels: constLabels,
	})
	m.failedSamplesTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "samples_failed_total",
		Help:        "Total number of samples which failed on send to remote storage, non-recoverable errors.",
		ConstLabels: constLabels,
	})
	m.failedExemplarsTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "exemplars_failed_total",
		Help:        "Total number of exemplars which failed on send to remote storage, non-recoverable errors.",
		ConstLabels: constLabels,
	})
	m.failedMetadataTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "metadata_failed_total",
		Help:        "Total number of metadata entries which failed on send to remote storage, non-recoverable errors.",
		ConstLabels: constLabels,
	})
	m.retriedSamplesTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "samples_retried_total",
		Help:        "Total number of samples which failed on send to remote storage but were retried because the send error was recoverable.",
		ConstLabels: constLabels,
	})
	m.retriedExemplarsTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "exemplars_retried_total",
		Help:        "Total number of exemplars which failed on send to remote storage but were retried because the send error was recoverable.",
		ConstLabels: constLabels,
	})
	m.retriedMetadataTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "metadata_retried_total",
		Help:        "Total number of metadata entries which failed on send to remote storage but were retried because the send error was recoverable.",
		ConstLabels: constLabels,
	})
	m.droppedSamplesTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "samples_dropped_total",
		Help:        "Total number of samples which were dropped after being read from the WAL before being sent via remote write, either via relabelling or unintentionally because of an unknown reference ID.",
		ConstLabels: constLabels,
	})
	m.droppedExemplarsTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "exemplars_dropped_total",
		Help:        "Total number of exemplars which were dropped after being read from the WAL before being sent via remote write, either via relabelling or unintentionally because of an unknown reference ID.",
		ConstLabels: constLabels,
	})
	m.enqueueRetriesTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "enqueue_retries_total",
		Help:        "Total number of times enqueue has failed because a shards queue was full.",
		ConstLabels: constLabels,
	})
	m.sentBatchDuration = prometheus.NewHistogram(prometheus.HistogramOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "sent_batch_duration_seconds",
		Help:        "Duration of send calls to the remote storage.",
		Buckets:     append(prometheus.DefBuckets, 25, 60, 120, 300),
		ConstLabels: constLabels,
	})
	m.highestSentTimestamp = &maxTimestamp{
		Gauge: prometheus.NewGauge(prometheus.GaugeOpts{
			Namespace:   namespace,
			Subsystem:   subsystem,
			Name:        "queue_highest_sent_timestamp_seconds",
			Help:        "Timestamp from a WAL sample, the highest timestamp successfully sent by this queue, in seconds since epoch.",
			ConstLabels: constLabels,
		}),
	}
	m.pendingSamples = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "samples_pending",
		Help:        "The number of samples pending in the queues shards to be sent to the remote storage.",
		ConstLabels: constLabels,
	})
	m.pendingExemplars = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "exemplars_pending",
		Help:        "The number of exemplars pending in the queues shards to be sent to the remote storage.",
		ConstLabels: constLabels,
	})
	m.shardCapacity = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "shard_capacity",
		Help:        "The capacity of each shard of the queue used for parallel sending to the remote storage.",
		ConstLabels: constLabels,
	})
	m.numShards = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "shards",
		Help:        "The number of shards used for parallel sending to the remote storage.",
		ConstLabels: constLabels,
	})
	m.maxNumShards = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "shards_max",
		Help:        "The maximum number of shards that the queue is allowed to run.",
		ConstLabels: constLabels,
	})
	m.minNumShards = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "shards_min",
		Help:        "The minimum number of shards that the queue is allowed to run.",
		ConstLabels: constLabels,
	})
	m.desiredNumShards = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "shards_desired",
		Help:        "The number of shards that the queues shard calculation wants to run based on the rate of samples in vs. samples out.",
		ConstLabels: constLabels,
	})
	m.sentBytesTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "bytes_total",
		Help:        "The total number of bytes of data (not metadata) sent by the queue after compression. Note that when exemplars over remote write is enabled the exemplars included in a remote write request count towards this metric.",
		ConstLabels: constLabels,
	})
	m.metadataBytesTotal = prometheus.NewCounter(prometheus.CounterOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "metadata_bytes_total",
		Help:        "The total number of bytes of metadata sent by the queue after compression.",
		ConstLabels: constLabels,
	})
	m.maxSamplesPerSend = prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace:   namespace,
		Subsystem:   subsystem,
		Name:        "max_samples_per_send",
		Help:        "The maximum number of samples to be sent, in a single request, to the remote storage. Note that, when sending of exemplars over remote write is enabled, exemplars count towards this limt.",
		ConstLabels: constLabels,
	})

	return m
}

func (m *queueManagerMetrics) register() {
	if m.reg != nil {
		m.reg.MustRegister(
			m.samplesTotal,
			m.exemplarsTotal,
			m.metadataTotal,
			m.failedSamplesTotal,
			m.failedExemplarsTotal,
			m.failedMetadataTotal,
			m.retriedSamplesTotal,
			m.retriedExemplarsTotal,
			m.retriedMetadataTotal,
			m.droppedSamplesTotal,
			m.droppedExemplarsTotal,
			m.enqueueRetriesTotal,
			m.sentBatchDuration,
			m.highestSentTimestamp,
			m.pendingSamples,
			m.pendingExemplars,
			m.shardCapacity,
			m.numShards,
			m.maxNumShards,
			m.minNumShards,
			m.desiredNumShards,
			m.sentBytesTotal,
			m.metadataBytesTotal,
			m.maxSamplesPerSend,
		)
	}
}

func (m *queueManagerMetrics) unregister() {
	if m.reg != nil {
		m.reg.Unregister(m.samplesTotal)
		m.reg.Unregister(m.exemplarsTotal)
		m.reg.Unregister(m.metadataTotal)
		m.reg.Unregister(m.failedSamplesTotal)
		m.reg.Unregister(m.failedExemplarsTotal)
		m.reg.Unregister(m.failedMetadataTotal)
		m.reg.Unregister(m.retriedSamplesTotal)
		m.reg.Unregister(m.retriedExemplarsTotal)
		m.reg.Unregister(m.retriedMetadataTotal)
		m.reg.Unregister(m.droppedSamplesTotal)
		m.reg.Unregister(m.droppedExemplarsTotal)
		m.reg.Unregister(m.enqueueRetriesTotal)
		m.reg.Unregister(m.sentBatchDuration)
		m.reg.Unregister(m.highestSentTimestamp)
		m.reg.Unregister(m.pendingSamples)
		m.reg.Unregister(m.pendingExemplars)
		m.reg.Unregister(m.shardCapacity)
		m.reg.Unregister(m.numShards)
		m.reg.Unregister(m.maxNumShards)
		m.reg.Unregister(m.minNumShards)
		m.reg.Unregister(m.desiredNumShards)
		m.reg.Unregister(m.sentBytesTotal)
		m.reg.Unregister(m.metadataBytesTotal)
		m.reg.Unregister(m.maxSamplesPerSend)
	}
}

// WriteClient defines an interface for sending a batch of samples to an
// external timeseries database.
type WriteClient interface {
	// Store stores the given samples in the remote storage.
	Store(context.Context, []byte) error
	// Name uniquely identifies the remote storage.
	Name() string
	// Endpoint is the remote read or write endpoint for the storage client.
	Endpoint() string
}

// QueueManager manages a queue of samples to be sent to the Storage
// indicated by the provided WriteClient. Implements writeTo interface
// used by WAL Watcher.
type QueueManager struct {
	lastSendTimestamp atomic.Int64

	logger          log.Logger
	flushDeadline   time.Duration
	cfg             config.QueueConfig
	mcfg            config.MetadataConfig
	externalLabels  labels.Labels
	relabelConfigs  []*relabel.Config
	sendExemplars   bool
	watcher         *wal.Watcher
	metadataWatcher *MetadataWatcher

	clientMtx   sync.RWMutex
	storeClient WriteClient

	seriesMtx     sync.Mutex // Covers seriesLabels and droppedSeries.
	seriesLabels  map[chunks.HeadSeriesRef]labels.Labels
	droppedSeries map[chunks.HeadSeriesRef]struct{}

	seriesSegmentMtx     sync.Mutex // Covers seriesSegmentIndexes - if you also lock seriesMtx, take seriesMtx first.
	seriesSegmentIndexes map[chunks.HeadSeriesRef]int

	shards      *shards
	numShards   int
	reshardChan chan int
	quit        chan struct{}
	wg          sync.WaitGroup

	dataIn, dataDropped, dataOut, dataOutDuration *ewmaRate

	metrics              *queueManagerMetrics
	interner             *pool
	highestRecvTimestamp *maxTimestamp
}

// NewQueueManager builds a new QueueManager.
func NewQueueManager(
	metrics *queueManagerMetrics,
	watcherMetrics *wal.WatcherMetrics,
	readerMetrics *wal.LiveReaderMetrics,
	logger log.Logger,
	walDir string,
	samplesIn *ewmaRate,
	cfg config.QueueConfig,
	mCfg config.MetadataConfig,
	externalLabels labels.Labels,
	relabelConfigs []*relabel.Config,
	client WriteClient,
	flushDeadline time.Duration,
	interner *pool,
	highestRecvTimestamp *maxTimestamp,
	sm ReadyScrapeManager,
	enableExemplarRemoteWrite bool,
) *QueueManager {
	if logger == nil {
		logger = log.NewNopLogger()
	}

	logger = log.With(logger, remoteName, client.Name(), endpoint, client.Endpoint())
	t := &QueueManager{
		logger:         logger,
		flushDeadline:  flushDeadline,
		cfg:            cfg,
		mcfg:           mCfg,
		externalLabels: externalLabels,
		relabelConfigs: relabelConfigs,
		storeClient:    client,
		sendExemplars:  enableExemplarRemoteWrite,

		seriesLabels:         make(map[chunks.HeadSeriesRef]labels.Labels),
		seriesSegmentIndexes: make(map[chunks.HeadSeriesRef]int),
		droppedSeries:        make(map[chunks.HeadSeriesRef]struct{}),

		numShards:   cfg.MinShards,
		reshardChan: make(chan int),
		quit:        make(chan struct{}),

		dataIn:          samplesIn,
		dataDropped:     newEWMARate(ewmaWeight, shardUpdateDuration),
		dataOut:         newEWMARate(ewmaWeight, shardUpdateDuration),
		dataOutDuration: newEWMARate(ewmaWeight, shardUpdateDuration),

		metrics:              metrics,
		interner:             interner,
		highestRecvTimestamp: highestRecvTimestamp,
	}

	t.watcher = wal.NewWatcher(watcherMetrics, readerMetrics, logger, client.Name(), t, walDir, enableExemplarRemoteWrite)
	if t.mcfg.Send {
		t.metadataWatcher = NewMetadataWatcher(logger, sm, client.Name(), t, t.mcfg.SendInterval, flushDeadline)
	}
	t.shards = t.newShards()

	return t
}

// AppendMetadata sends metadata the remote storage. Metadata is sent in batches, but is not parallelized.
func (t *QueueManager) AppendMetadata(ctx context.Context, metadata []scrape.MetricMetadata) {
	mm := make([]prompb.MetricMetadata, 0, len(metadata))
	for _, entry := range metadata {
		mm = append(mm, prompb.MetricMetadata{
			MetricFamilyName: entry.Metric,
			Help:             entry.Help,
			Type:             metricTypeToMetricTypeProto(entry.Type),
			Unit:             entry.Unit,
		})
	}

	pBuf := proto.NewBuffer(nil)
	numSends := int(math.Ceil(float64(len(metadata)) / float64(t.mcfg.MaxSamplesPerSend)))
	for i := 0; i < numSends; i++ {
		last := (i + 1) * t.mcfg.MaxSamplesPerSend
		if last > len(metadata) {
			last = len(metadata)
		}
		err := t.sendMetadataWithBackoff(ctx, mm[i*t.mcfg.MaxSamplesPerSend:last], pBuf)
		if err != nil {
			t.metrics.failedMetadataTotal.Add(float64(last - (i * t.mcfg.MaxSamplesPerSend)))
			level.Error(t.logger).Log("msg", "non-recoverable error while sending metadata", "count", last-(i*t.mcfg.MaxSamplesPerSend), "err", err)
		}
	}
}

func (t *QueueManager) sendMetadataWithBackoff(ctx context.Context, metadata []prompb.MetricMetadata, pBuf *proto.Buffer) error {
	// Build the WriteRequest with no samples.
	req, _, err := buildWriteRequest(nil, metadata, pBuf, nil)
	if err != nil {
		return err
	}

	metadataCount := len(metadata)

	attemptStore := func(try int) error {
		span, ctx := opentracing.StartSpanFromContext(ctx, "Remote Metadata Send Batch")
		defer span.Finish()

		span.SetTag("metadata", metadataCount)
		span.SetTag("try", try)
		span.SetTag("remote_name", t.storeClient.Name())
		span.SetTag("remote_url", t.storeClient.Endpoint())

		begin := time.Now()
		err := t.storeClient.Store(ctx, req)
		t.metrics.sentBatchDuration.Observe(time.Since(begin).Seconds())

		if err != nil {
			span.LogKV("error", err)
			ext.Error.Set(span, true)
			return err
		}

		return nil
	}

	retry := func() {
		t.metrics.retriedMetadataTotal.Add(float64(len(metadata)))
	}
	err = sendWriteRequestWithBackoff(ctx, t.cfg, t.logger, attemptStore, retry)
	if err != nil {
		return err
	}
	t.metrics.metadataTotal.Add(float64(len(metadata)))
	t.metrics.metadataBytesTotal.Add(float64(len(req)))
	return nil
}

// Append queues a sample to be sent to the remote storage. Blocks until all samples are
// enqueued on their shards or a shutdown signal is received.
func (t *QueueManager) Append(samples []record.RefSample) bool {
outer:
	for _, s := range samples {
		t.seriesMtx.Lock()
		lbls, ok := t.seriesLabels[s.Ref]
		if !ok {
			t.metrics.droppedSamplesTotal.Inc()
			t.dataDropped.incr(1)
			if _, ok := t.droppedSeries[s.Ref]; !ok {
				level.Info(t.logger).Log("msg", "Dropped sample for series that was not explicitly dropped via relabelling", "ref", s.Ref)
			}
			t.seriesMtx.Unlock()
			continue
		}
		t.seriesMtx.Unlock()
		// This will only loop if the queues are being resharded.
		backoff := t.cfg.MinBackoff
		for {
			select {
			case <-t.quit:
				return false
			default:
			}
			if t.shards.enqueue(s.Ref, sampleOrExemplar{
				seriesLabels: lbls,
				timestamp:    s.T,
				value:        s.V,
				isSample:     true,
			}) {
				continue outer
			}

			t.metrics.enqueueRetriesTotal.Inc()
			time.Sleep(time.Duration(backoff))
			backoff = backoff * 2
			if backoff > t.cfg.MaxBackoff {
				backoff = t.cfg.MaxBackoff
			}
		}
	}
	return true
}

func (t *QueueManager) AppendExemplars(exemplars []record.RefExemplar) bool {
	if !t.sendExemplars {
		return true
	}

outer:
	for _, e := range exemplars {
		t.seriesMtx.Lock()
		lbls, ok := t.seriesLabels[e.Ref]
		if !ok {
			t.metrics.droppedExemplarsTotal.Inc()
			// Track dropped exemplars in the same EWMA for sharding calc.
			t.dataDropped.incr(1)
			if _, ok := t.droppedSeries[e.Ref]; !ok {
				level.Info(t.logger).Log("msg", "Dropped exemplar for series that was not explicitly dropped via relabelling", "ref", e.Ref)
			}
			t.seriesMtx.Unlock()
			continue
		}
		t.seriesMtx.Unlock()
		// This will only loop if the queues are being resharded.
		backoff := t.cfg.MinBackoff
		for {
			select {
			case <-t.quit:
				return false
			default:
			}
			if t.shards.enqueue(e.Ref, sampleOrExemplar{
				seriesLabels:   lbls,
				timestamp:      e.T,
				value:          e.V,
				exemplarLabels: e.Labels,
			}) {
				continue outer
			}

			t.metrics.enqueueRetriesTotal.Inc()
			time.Sleep(time.Duration(backoff))
			backoff = backoff * 2
			if backoff > t.cfg.MaxBackoff {
				backoff = t.cfg.MaxBackoff
			}
		}
	}
	return true
}

// Start the queue manager sending samples to the remote storage.
// Does not block.
func (t *QueueManager) Start() {
	// Register and initialise some metrics.
	t.metrics.register()
	t.metrics.shardCapacity.Set(float64(t.cfg.Capacity))
	t.metrics.maxNumShards.Set(float64(t.cfg.MaxShards))
	t.metrics.minNumShards.Set(float64(t.cfg.MinShards))
	t.metrics.desiredNumShards.Set(float64(t.cfg.MinShards))
	t.metrics.maxSamplesPerSend.Set(float64(t.cfg.MaxSamplesPerSend))

	t.shards.start(t.numShards)
	t.watcher.Start()
	if t.mcfg.Send {
		t.metadataWatcher.Start()
	}

	t.wg.Add(2)
	go t.updateShardsLoop()
	go t.reshardLoop()
}

// Stop stops sending samples to the remote storage and waits for pending
// sends to complete.
func (t *QueueManager) Stop() {
	level.Info(t.logger).Log("msg", "Stopping remote storage...")
	defer level.Info(t.logger).Log("msg", "Remote storage stopped.")

	close(t.quit)
	t.wg.Wait()
	// Wait for all QueueManager routines to end before stopping shards, metadata watcher, and WAL watcher. This
	// is to ensure we don't end up executing a reshard and shards.stop() at the same time, which
	// causes a closed channel panic.
	t.shards.stop()
	t.watcher.Stop()
	if t.mcfg.Send {
		t.metadataWatcher.Stop()
	}

	// On shutdown, release the strings in the labels from the intern pool.
	t.seriesMtx.Lock()
	for _, labels := range t.seriesLabels {
		t.releaseLabels(labels)
	}
	t.seriesMtx.Unlock()
	t.metrics.unregister()
}

// StoreSeries keeps track of which series we know about for lookups when sending samples to remote.
func (t *QueueManager) StoreSeries(series []record.RefSeries, index int) {
	t.seriesMtx.Lock()
	defer t.seriesMtx.Unlock()
	t.seriesSegmentMtx.Lock()
	defer t.seriesSegmentMtx.Unlock()
	for _, s := range series {
		// Just make sure all the Refs of Series will insert into seriesSegmentIndexes map for tracking.
		t.seriesSegmentIndexes[s.Ref] = index

		ls := processExternalLabels(s.Labels, t.externalLabels)
		lbls := relabel.Process(ls, t.relabelConfigs...)
		if len(lbls) == 0 {
			t.droppedSeries[s.Ref] = struct{}{}
			continue
		}
		t.internLabels(lbls)

		// We should not ever be replacing a series labels in the map, but just
		// in case we do we need to ensure we do not leak the replaced interned
		// strings.
		if orig, ok := t.seriesLabels[s.Ref]; ok {
			t.releaseLabels(orig)
		}
		t.seriesLabels[s.Ref] = lbls
	}
}

// UpdateSeriesSegment updates the segment number held against the series,
// so we can trim older ones in SeriesReset.
func (t *QueueManager) UpdateSeriesSegment(series []record.RefSeries, index int) {
	t.seriesSegmentMtx.Lock()
	defer t.seriesSegmentMtx.Unlock()
	for _, s := range series {
		t.seriesSegmentIndexes[s.Ref] = index
	}
}

// SeriesReset is used when reading a checkpoint. WAL Watcher should have
// stored series records with the checkpoints index number, so we can now
// delete any ref ID's lower than that # from the two maps.
func (t *QueueManager) SeriesReset(index int) {
	t.seriesMtx.Lock()
	defer t.seriesMtx.Unlock()
	t.seriesSegmentMtx.Lock()
	defer t.seriesSegmentMtx.Unlock()
	// Check for series that are in segments older than the checkpoint
	// that were not also present in the checkpoint.
	for k, v := range t.seriesSegmentIndexes {
		if v < index {
			delete(t.seriesSegmentIndexes, k)
			t.releaseLabels(t.seriesLabels[k])
			delete(t.seriesLabels, k)
			delete(t.droppedSeries, k)
		}
	}
}

// SetClient updates the client used by a queue. Used when only client specific
// fields are updated to avoid restarting the queue.
func (t *QueueManager) SetClient(c WriteClient) {
	t.clientMtx.Lock()
	t.storeClient = c
	t.clientMtx.Unlock()
}

func (t *QueueManager) client() WriteClient {
	t.clientMtx.RLock()
	defer t.clientMtx.RUnlock()
	return t.storeClient
}

func (t *QueueManager) internLabels(lbls labels.Labels) {
	for i, l := range lbls {
		lbls[i].Name = t.interner.intern(l.Name)
		lbls[i].Value = t.interner.intern(l.Value)
	}
}

func (t *QueueManager) releaseLabels(ls labels.Labels) {
	for _, l := range ls {
		t.interner.release(l.Name)
		t.interner.release(l.Value)
	}
}

// processExternalLabels merges externalLabels into ls. If ls contains
// a label in externalLabels, the value in ls wins.
func processExternalLabels(ls, externalLabels labels.Labels) labels.Labels {
	i, j, result := 0, 0, make(labels.Labels, 0, len(ls)+len(externalLabels))
	for i < len(ls) && j < len(externalLabels) {
		if ls[i].Name < externalLabels[j].Name {
			result = append(result, labels.Label{
				Name:  ls[i].Name,
				Value: ls[i].Value,
			})
			i++
		} else if ls[i].Name > externalLabels[j].Name {
			result = append(result, externalLabels[j])
			j++
		} else {
			result = append(result, labels.Label{
				Name:  ls[i].Name,
				Value: ls[i].Value,
			})
			i++
			j++
		}
	}

	return append(append(result, ls[i:]...), externalLabels[j:]...)
}

func (t *QueueManager) updateShardsLoop() {
	defer t.wg.Done()

	ticker := time.NewTicker(shardUpdateDuration)
	defer ticker.Stop()
	for {
		select {
		case <-ticker.C:
			desiredShards := t.calculateDesiredShards()
			if !t.shouldReshard(desiredShards) {
				continue
			}
			// Resharding can take some time, and we want this loop
			// to stay close to shardUpdateDuration.
			select {
			case t.reshardChan <- desiredShards:
				level.Info(t.logger).Log("msg", "Remote storage resharding", "from", t.numShards, "to", desiredShards)
				t.numShards = desiredShards
			default:
				level.Info(t.logger).Log("msg", "Currently resharding, skipping.")
			}
		case <-t.quit:
			return
		}
	}
}

// shouldReshard returns if resharding should occur
func (t *QueueManager) shouldReshard(desiredShards int) bool {
	if desiredShards == t.numShards {
		return false
	}
	// We shouldn't reshard if Prometheus hasn't been able to send to the
	// remote endpoint successfully within some period of time.
	minSendTimestamp := time.Now().Add(-2 * time.Duration(t.cfg.BatchSendDeadline)).Unix()
	lsts := t.lastSendTimestamp.Load()
	if lsts < minSendTimestamp {
		level.Warn(t.logger).Log("msg", "Skipping resharding, last successful send was beyond threshold", "lastSendTimestamp", lsts, "minSendTimestamp", minSendTimestamp)
		return false
	}
	return true
}

// calculateDesiredShards returns the number of desired shards, which will be
// the current QueueManager.numShards if resharding should not occur for reasons
// outlined in this functions implementation. It is up to the caller to reshard, or not,
// based on the return value.
func (t *QueueManager) calculateDesiredShards() int {
	t.dataOut.tick()
	t.dataDropped.tick()
	t.dataOutDuration.tick()

	// We use the number of incoming samples as a prediction of how much work we
	// will need to do next iteration.  We add to this any pending samples
	// (received - send) so we can catch up with any backlog. We use the average
	// outgoing batch latency to work out how many shards we need.
	var (
		dataInRate      = t.dataIn.rate()
		dataOutRate     = t.dataOut.rate()
		dataKeptRatio   = dataOutRate / (t.dataDropped.rate() + dataOutRate)
		dataOutDuration = t.dataOutDuration.rate() / float64(time.Second)
		dataPendingRate = dataInRate*dataKeptRatio - dataOutRate
		highestSent     = t.metrics.highestSentTimestamp.Get()
		highestRecv     = t.highestRecvTimestamp.Get()
		delay           = highestRecv - highestSent
		dataPending     = delay * dataInRate * dataKeptRatio
	)

	if dataOutRate <= 0 {
		return t.numShards
	}

	// When behind we will try to catch up on a proporation of samples per tick.
	// This works similarly to an integral accumulator in that pending samples
	// is the result of the error integral.
	const integralGain = 0.1 / float64(shardUpdateDuration/time.Second)

	var (
		timePerSample = dataOutDuration / dataOutRate
		desiredShards = timePerSample * (dataInRate*dataKeptRatio + integralGain*dataPending)
	)
	t.metrics.desiredNumShards.Set(desiredShards)
	level.Debug(t.logger).Log("msg", "QueueManager.calculateDesiredShards",
		"dataInRate", dataInRate,
		"dataOutRate", dataOutRate,
		"dataKeptRatio", dataKeptRatio,
		"dataPendingRate", dataPendingRate,
		"dataPending", dataPending,
		"dataOutDuration", dataOutDuration,
		"timePerSample", timePerSample,
		"desiredShards", desiredShards,
		"highestSent", highestSent,
		"highestRecv", highestRecv,
	)

	// Changes in the number of shards must be greater than shardToleranceFraction.
	var (
		lowerBound = float64(t.numShards) * (1. - shardToleranceFraction)
		upperBound = float64(t.numShards) * (1. + shardToleranceFraction)
	)
	level.Debug(t.logger).Log("msg", "QueueManager.updateShardsLoop",
		"lowerBound", lowerBound, "desiredShards", desiredShards, "upperBound", upperBound)
	if lowerBound <= desiredShards && desiredShards <= upperBound {
		return t.numShards
	}

	numShards := int(math.Ceil(desiredShards))
	// Do not downshard if we are more than ten seconds back.
	if numShards < t.numShards && delay > 10.0 {
		level.Debug(t.logger).Log("msg", "Not downsharding due to being too far behind")
		return t.numShards
	}

	if numShards > t.cfg.MaxShards {
		numShards = t.cfg.MaxShards
	} else if numShards < t.cfg.MinShards {
		numShards = t.cfg.MinShards
	}
	return numShards
}

func (t *QueueManager) reshardLoop() {
	defer t.wg.Done()

	for {
		select {
		case numShards := <-t.reshardChan:
			// We start the newShards after we have stopped (the therefore completely
			// flushed) the oldShards, to guarantee we only every deliver samples in
			// order.
			t.shards.stop()
			t.shards.start(numShards)
		case <-t.quit:
			return
		}
	}
}

func (t *QueueManager) newShards() *shards {
	s := &shards{
		qm:   t,
		done: make(chan struct{}),
	}
	return s
}

type shards struct {
	mtx sync.RWMutex // With the WAL, this is never actually contended.

	qm     *QueueManager
	queues []*queue
	// So we can accurately track how many of each are lost during shard shutdowns.
	enqueuedSamples   atomic.Int64
	enqueuedExemplars atomic.Int64

	// Emulate a wait group with a channel and an atomic int, as you
	// cannot select on a wait group.
	done    chan struct{}
	running atomic.Int32

	// Soft shutdown context will prevent new enqueues and deadlocks.
	softShutdown chan struct{}

	// Hard shutdown context is used to terminate outgoing HTTP connections
	// after giving them a chance to terminate.
	hardShutdown                   context.CancelFunc
	samplesDroppedOnHardShutdown   atomic.Uint32
	exemplarsDroppedOnHardShutdown atomic.Uint32
}

// start the shards; must be called before any call to enqueue.
func (s *shards) start(n int) {
	s.mtx.Lock()
	defer s.mtx.Unlock()

	s.qm.metrics.pendingSamples.Set(0)
	s.qm.metrics.numShards.Set(float64(n))

	newQueues := make([]*queue, n)
	for i := 0; i < n; i++ {
		newQueues[i] = newQueue(s.qm.cfg.MaxSamplesPerSend, s.qm.cfg.Capacity)
	}

	s.queues = newQueues

	var hardShutdownCtx context.Context
	hardShutdownCtx, s.hardShutdown = context.WithCancel(context.Background())
	s.softShutdown = make(chan struct{})
	s.running.Store(int32(n))
	s.done = make(chan struct{})
	s.samplesDroppedOnHardShutdown.Store(0)
	s.exemplarsDroppedOnHardShutdown.Store(0)
	for i := 0; i < n; i++ {
		go s.runShard(hardShutdownCtx, i, newQueues[i])
	}
}

// stop the shards; subsequent call to enqueue will return false.
func (s *shards) stop() {
	// Attempt a clean shutdown, but only wait flushDeadline for all the shards
	// to cleanly exit.  As we're doing RPCs, enqueue can block indefinitely.
	// We must be able so call stop concurrently, hence we can only take the
	// RLock here.
	s.mtx.RLock()
	close(s.softShutdown)
	s.mtx.RUnlock()

	// Enqueue should now be unblocked, so we can take the write lock.  This
	// also ensures we don't race with writes to the queues, and get a panic:
	// send on closed channel.
	s.mtx.Lock()
	defer s.mtx.Unlock()
	for _, queue := range s.queues {
		go queue.FlushAndShutdown(s.done)
	}
	select {
	case <-s.done:
		return
	case <-time.After(s.qm.flushDeadline):
	}

	// Force an unclean shutdown.
	s.hardShutdown()
	<-s.done
	if dropped := s.samplesDroppedOnHardShutdown.Load(); dropped > 0 {
		level.Error(s.qm.logger).Log("msg", "Failed to flush all samples on shutdown", "count", dropped)
	}
	if dropped := s.exemplarsDroppedOnHardShutdown.Load(); dropped > 0 {
		level.Error(s.qm.logger).Log("msg", "Failed to flush all exemplars on shutdown", "count", dropped)
	}
}

// enqueue data (sample or exemplar).  If we are currently in the process of shutting down or resharding,
// will return false; in this case, you should back off and retry.
func (s *shards) enqueue(ref chunks.HeadSeriesRef, data sampleOrExemplar) bool {
	s.mtx.RLock()
	defer s.mtx.RUnlock()

	select {
	case <-s.softShutdown:
		return false
	default:
	}

	shard := uint64(ref) % uint64(len(s.queues))
	select {
	case <-s.softShutdown:
		return false
	default:
		appended := s.queues[shard].Append(data, s.softShutdown)
		if !appended {
			return false
		}
		if data.isSample {
			s.qm.metrics.pendingSamples.Inc()
			s.enqueuedSamples.Inc()
		} else {
			s.qm.metrics.pendingExemplars.Inc()
			s.enqueuedExemplars.Inc()
		}
		return true
	}
}

type queue struct {
	batch      []sampleOrExemplar
	batchQueue chan []sampleOrExemplar

	// Since we know there are a limited number of batches out, using a stack
	// is easy and safe so a sync.Pool is not necessary.
	batchPool [][]sampleOrExemplar
	// This mutex covers adding and removing batches from the batchPool.
	poolMux sync.Mutex
}

type sampleOrExemplar struct {
	seriesLabels   labels.Labels
	value          float64
	timestamp      int64
	exemplarLabels labels.Labels
	isSample       bool
}

func newQueue(batchSize, capacity int) *queue {
	batches := capacity / batchSize
	return &queue{
		batch:      make([]sampleOrExemplar, 0, batchSize),
		batchQueue: make(chan []sampleOrExemplar, batches),
		// batchPool should have capacity for everything in the channel + 1 for
		// the batch being processed.
		batchPool: make([][]sampleOrExemplar, 0, batches+1),
	}
}

func (q *queue) Append(datum sampleOrExemplar, stop <-chan struct{}) bool {
	q.batch = append(q.batch, datum)
	if len(q.batch) == cap(q.batch) {
		select {
		case q.batchQueue <- q.batch:
			q.batch = q.newBatch(cap(q.batch))
			return true
		case <-stop:
			// Remove the sample we just appended. It will get retried.
			q.batch = q.batch[:len(q.batch)-1]
			return false
		}
	}
	return true
}

func (q *queue) Chan() <-chan []sampleOrExemplar {
	return q.batchQueue
}

// Batch returns the current batch and allocates a new batch. Must not be
// called concurrently with Append.
func (q *queue) Batch() []sampleOrExemplar {
	batch := q.batch
	q.batch = q.newBatch(cap(batch))
	return batch
}

// ReturnForReuse adds the batch buffer back to the internal pool.
func (q *queue) ReturnForReuse(batch []sampleOrExemplar) {
	q.poolMux.Lock()
	defer q.poolMux.Unlock()
	if len(q.batchPool) < cap(q.batchPool) {
		q.batchPool = append(q.batchPool, batch[:0])
	}
}

// FlushAndShutdown stops the queue and flushes any samples. No appends can be
// made after this is called.
func (q *queue) FlushAndShutdown(done <-chan struct{}) {
	if len(q.batch) > 0 {
		select {
		case q.batchQueue <- q.batch:
		case <-done:
			// The shard has been hard shut down, so no more samples can be
			// sent. Drop everything left in the queue.
		}
	}
	q.batch = nil
	close(q.batchQueue)
}

func (q *queue) newBatch(capacity int) []sampleOrExemplar {
	q.poolMux.Lock()
	defer q.poolMux.Unlock()
	batches := len(q.batchPool)
	if batches > 0 {
		batch := q.batchPool[batches-1]
		q.batchPool = q.batchPool[:batches-1]
		return batch
	}
	return make([]sampleOrExemplar, 0, capacity)
}

func (s *shards) runShard(ctx context.Context, shardID int, queue *queue) {
	defer func() {
		if s.running.Dec() == 0 {
			close(s.done)
		}
	}()

	shardNum := strconv.Itoa(shardID)

	// Send batches of at most MaxSamplesPerSend samples to the remote storage.
	// If we have fewer samples than that, flush them out after a deadline anyways.
	var (
		max = s.qm.cfg.MaxSamplesPerSend

		pBuf = proto.NewBuffer(nil)
		buf  []byte
	)
	if s.qm.sendExemplars {
		max += int(float64(max) * 0.1)
	}

	batchQueue := queue.Chan()
	pendingData := make([]prompb.TimeSeries, max)
	for i := range pendingData {
		pendingData[i].Samples = []prompb.Sample{{}}
		if s.qm.sendExemplars {
			pendingData[i].Exemplars = []prompb.Exemplar{{}}
		}
	}

	timer := time.NewTimer(time.Duration(s.qm.cfg.BatchSendDeadline))
	stop := func() {
		if !timer.Stop() {
			select {
			case <-timer.C:
			default:
			}
		}
	}
	defer stop()

	for {
		select {
		case <-ctx.Done():
			// In this case we drop all samples in the buffer and the queue.
			// Remove them from pending and mark them as failed.
			droppedSamples := int(s.enqueuedSamples.Load())
			droppedExemplars := int(s.enqueuedExemplars.Load())
			s.qm.metrics.pendingSamples.Sub(float64(droppedSamples))
			s.qm.metrics.pendingExemplars.Sub(float64(droppedExemplars))
			s.qm.metrics.failedSamplesTotal.Add(float64(droppedSamples))
			s.qm.metrics.failedExemplarsTotal.Add(float64(droppedExemplars))
			s.samplesDroppedOnHardShutdown.Add(uint32(droppedSamples))
			s.exemplarsDroppedOnHardShutdown.Add(uint32(droppedExemplars))
			return

		case batch, ok := <-batchQueue:
			if !ok {
				return
			}
			nPendingSamples, nPendingExemplars := s.populateTimeSeries(batch, pendingData)
			queue.ReturnForReuse(batch)
			n := nPendingSamples + nPendingExemplars
			s.sendSamples(ctx, pendingData[:n], nPendingSamples, nPendingExemplars, pBuf, &buf)

			stop()
			timer.Reset(time.Duration(s.qm.cfg.BatchSendDeadline))

		case <-timer.C:
			// We need to take the lock when getting a batch to avoid
			// concurrent Appends. Generally this will only happen on low
			// traffic instances.
			s.mtx.Lock()
			// First, we need to see if we can happen to get a batch from the queue if it filled while acquiring the lock.
			var batch []sampleOrExemplar
			select {
			case batch = <-batchQueue:
			default:
				batch = queue.Batch()
			}
			s.mtx.Unlock()
			if len(batch) > 0 {
				nPendingSamples, nPendingExemplars := s.populateTimeSeries(batch, pendingData)
				n := nPendingSamples + nPendingExemplars
				level.Debug(s.qm.logger).Log("msg", "runShard timer ticked, sending buffered data", "samples", nPendingSamples, "exemplars", nPendingExemplars, "shard", shardNum)
				s.sendSamples(ctx, pendingData[:n], nPendingSamples, nPendingExemplars, pBuf, &buf)
			}
			queue.ReturnForReuse(batch)
			timer.Reset(time.Duration(s.qm.cfg.BatchSendDeadline))
		}
	}
}

func (s *shards) populateTimeSeries(batch []sampleOrExemplar, pendingData []prompb.TimeSeries) (int, int) {
	var nPendingSamples, nPendingExemplars int
	for nPending, d := range batch {
		pendingData[nPending].Samples = pendingData[nPending].Samples[:0]
		if s.qm.sendExemplars {
			pendingData[nPending].Exemplars = pendingData[nPending].Exemplars[:0]
		}
		// Number of pending samples is limited by the fact that sendSamples (via sendSamplesWithBackoff)
		// retries endlessly, so once we reach max samples, if we can never send to the endpoint we'll
		// stop reading from the queue. This makes it safe to reference pendingSamples by index.
		if d.isSample {
			pendingData[nPending].Labels = labelsToLabelsProto(d.seriesLabels, pendingData[nPending].Labels)
			pendingData[nPending].Samples = append(pendingData[nPending].Samples, prompb.Sample{
				Value:     d.value,
				Timestamp: d.timestamp,
			})
			nPendingSamples++
		} else {
			pendingData[nPending].Labels = labelsToLabelsProto(d.seriesLabels, pendingData[nPending].Labels)
			pendingData[nPending].Exemplars = append(pendingData[nPending].Exemplars, prompb.Exemplar{
				Labels:    labelsToLabelsProto(d.exemplarLabels, nil),
				Value:     d.value,
				Timestamp: d.timestamp,
			})
			nPendingExemplars++
		}
	}
	return nPendingSamples, nPendingExemplars
}

func (s *shards) sendSamples(ctx context.Context, samples []prompb.TimeSeries, sampleCount, exemplarCount int, pBuf *proto.Buffer, buf *[]byte) {
	begin := time.Now()
	err := s.sendSamplesWithBackoff(ctx, samples, sampleCount, exemplarCount, pBuf, buf)
	if err != nil {
		level.Error(s.qm.logger).Log("msg", "non-recoverable error", "count", sampleCount, "exemplarCount", exemplarCount, "err", err)
		s.qm.metrics.failedSamplesTotal.Add(float64(sampleCount))
		s.qm.metrics.failedExemplarsTotal.Add(float64(exemplarCount))
	}

	// These counters are used to calculate the dynamic sharding, and as such
	// should be maintained irrespective of success or failure.
	s.qm.dataOut.incr(int64(len(samples)))
	s.qm.dataOutDuration.incr(int64(time.Since(begin)))
	s.qm.lastSendTimestamp.Store(time.Now().Unix())
	// Pending samples/exemplars also should be subtracted as an error means
	// they will not be retried.
	s.qm.metrics.pendingSamples.Sub(float64(sampleCount))
	s.qm.metrics.pendingExemplars.Sub(float64(exemplarCount))
	s.enqueuedSamples.Sub(int64(sampleCount))
	s.enqueuedExemplars.Sub(int64(exemplarCount))
}

// sendSamples to the remote storage with backoff for recoverable errors.
func (s *shards) sendSamplesWithBackoff(ctx context.Context, samples []prompb.TimeSeries, sampleCount, exemplarCount int, pBuf *proto.Buffer, buf *[]byte) error {
	// Build the WriteRequest with no metadata.
	req, highest, err := buildWriteRequest(samples, nil, pBuf, *buf)
	if err != nil {
		// Failing to build the write request is non-recoverable, since it will
		// only error if marshaling the proto to bytes fails.
		return err
	}

	reqSize := len(req)
	*buf = req

	// An anonymous function allows us to defer the completion of our per-try spans
	// without causing a memory leak, and it has the nice effect of not propagating any
	// parameters for sendSamplesWithBackoff/3.
	attemptStore := func(try int) error {
		span, ctx := opentracing.StartSpanFromContext(ctx, "Remote Send Batch")
		defer span.Finish()

		span.SetTag("samples", sampleCount)
		if exemplarCount > 0 {
			span.SetTag("exemplars", exemplarCount)
		}
		span.SetTag("request_size", reqSize)
		span.SetTag("try", try)
		span.SetTag("remote_name", s.qm.storeClient.Name())
		span.SetTag("remote_url", s.qm.storeClient.Endpoint())

		begin := time.Now()
		s.qm.metrics.samplesTotal.Add(float64(sampleCount))
		s.qm.metrics.exemplarsTotal.Add(float64(exemplarCount))
		err := s.qm.client().Store(ctx, *buf)
		s.qm.metrics.sentBatchDuration.Observe(time.Since(begin).Seconds())

		if err != nil {
			span.LogKV("error", err)
			ext.Error.Set(span, true)
			return err
		}

		return nil
	}

	onRetry := func() {
		s.qm.metrics.retriedSamplesTotal.Add(float64(sampleCount))
		s.qm.metrics.retriedExemplarsTotal.Add(float64(exemplarCount))
	}

	err = sendWriteRequestWithBackoff(ctx, s.qm.cfg, s.qm.logger, attemptStore, onRetry)
	if errors.Is(err, context.Canceled) {
		// When there is resharding, we cancel the context for this queue, which means the data is not sent.
		// So we exit early to not update the metrics.
		return err
	}

	s.qm.metrics.sentBytesTotal.Add(float64(reqSize))
	s.qm.metrics.highestSentTimestamp.Set(float64(highest / 1000))

	return err
}

func sendWriteRequestWithBackoff(ctx context.Context, cfg config.QueueConfig, l log.Logger, attempt func(int) error, onRetry func()) error {
	backoff := cfg.MinBackoff
	sleepDuration := model.Duration(0)
	try := 0

	for {
		select {
		case <-ctx.Done():
			return ctx.Err()
		default:
		}

		err := attempt(try)

		if err == nil {
			return nil
		}

		// If the error is unrecoverable, we should not retry.
		backoffErr, ok := err.(RecoverableError)
		if !ok {
			return err
		}

		sleepDuration = backoff
		if backoffErr.retryAfter > 0 {
			sleepDuration = backoffErr.retryAfter
			level.Info(l).Log("msg", "Retrying after duration specified by Retry-After header", "duration", sleepDuration)
		} else if backoffErr.retryAfter < 0 {
			level.Debug(l).Log("msg", "retry-after cannot be in past, retrying using default backoff mechanism")
		}

		select {
		case <-ctx.Done():
		case <-time.After(time.Duration(sleepDuration)):
		}

		// If we make it this far, we've encountered a recoverable error and will retry.
		onRetry()
		level.Warn(l).Log("msg", "Failed to send batch, retrying", "err", err)

		backoff = sleepDuration * 2

		if backoff > cfg.MaxBackoff {
			backoff = cfg.MaxBackoff
		}

		try++
	}
}

func buildWriteRequest(samples []prompb.TimeSeries, metadata []prompb.MetricMetadata, pBuf *proto.Buffer, buf []byte) ([]byte, int64, error) {
	var highest int64
	for _, ts := range samples {
		// At the moment we only ever append a TimeSeries with a single sample or exemplar in it.
		if len(ts.Samples) > 0 && ts.Samples[0].Timestamp > highest {
			highest = ts.Samples[0].Timestamp
		}
		if len(ts.Exemplars) > 0 && ts.Exemplars[0].Timestamp > highest {
			highest = ts.Exemplars[0].Timestamp
		}
	}

	req := &prompb.WriteRequest{
		Timeseries: samples,
		Metadata:   metadata,
	}

	if pBuf == nil {
		pBuf = proto.NewBuffer(nil) // For convenience in tests. Not efficient.
	} else {
		pBuf.Reset()
	}
	err := pBuf.Marshal(req)
	if err != nil {
		return nil, highest, err
	}

	// snappy uses len() to see if it needs to allocate a new slice. Make the
	// buffer as long as possible.
	if buf != nil {
		buf = buf[0:cap(buf)]
	}
	compressed := snappy.Encode(buf, pBuf.Bytes())
	return compressed, highest, nil
}