github.com/m3db/m3@v1.5.1-0.20231129193456-75a402aa583b/src/query/storage/m3/storage.go

// Copyright (c) 2018 Uber Technologies, Inc.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.

package m3

import (
	"bytes"
	"context"
	goerrors "errors"
	"fmt"
	"sync"
	"time"

	coordmodel "github.com/m3db/m3/src/cmd/services/m3coordinator/model"
	"github.com/m3db/m3/src/dbnode/client"
	"github.com/m3db/m3/src/dbnode/storage/index"
	"github.com/m3db/m3/src/query/block"
	"github.com/m3db/m3/src/query/errors"
	"github.com/m3db/m3/src/query/generated/proto/prompb"
	"github.com/m3db/m3/src/query/models"
	"github.com/m3db/m3/src/query/storage"
	"github.com/m3db/m3/src/query/storage/m3/consolidators"
	"github.com/m3db/m3/src/query/storage/m3/storagemetadata"
	"github.com/m3db/m3/src/query/tracepoint"
	"github.com/m3db/m3/src/query/ts"
	xcontext "github.com/m3db/m3/src/x/context"
	xerrors "github.com/m3db/m3/src/x/errors"
	"github.com/m3db/m3/src/x/ident"
	"github.com/m3db/m3/src/x/instrument"
	xtime "github.com/m3db/m3/src/x/time"

	"github.com/opentracing/opentracing-go/log"
	"github.com/prometheus/common/model"
	"go.uber.org/zap"
	"go.uber.org/zap/zapcore"
)

const (
	minWriteWaitTimeout = time.Second
)

var (
	// The default name for the name tag in Prometheus metrics.
	promDefaultName = []byte(model.MetricNameLabel)
	// The prefix for reserved labels, e.g. __name__
	reservedLabelPrefix = []byte(model.ReservedLabelPrefix)
	// The name for the rollup tag defined by the coordinator model.
	rollupTagName = []byte(coordmodel.RollupTagName)
	// The value for the rollup tag defined by the coordinator model.
	rollupTagValue = []byte(coordmodel.RollupTagValue)

	errUnaggregatedAndAggregatedDisabled = goerrors.New("fetch options has both" +
		" aggregated and unaggregated namespace lookup disabled")
	errNoNamespacesConfigured             = goerrors.New("no namespaces configured")
	errUnaggregatedNamespaceUninitialized = goerrors.New(
		"unaggregated namespace is not yet initialized")
)

type m3storage struct {
	clusters Clusters
	opts     Options
	nowFn    func() time.Time
	logger   *zap.Logger
}

// NewStorage creates a new local m3storage instance.
func NewStorage(
	clusters Clusters,
	opts Options,
	instrumentOpts instrument.Options,
) (Storage, error) {
	if err := opts.Validate(); err != nil {
		return nil, err
	}

	return &m3storage{
		clusters: clusters,
		opts:     opts,
		nowFn:    time.Now,
		logger:   instrumentOpts.Logger(),
	}, nil
}

func (s *m3storage) QueryStorageMetadataAttributes(
	_ context.Context,
	queryStart, queryEnd time.Time,
	opts *storage.FetchOptions,
) ([]storagemetadata.Attributes, error) {
	now := xtime.ToUnixNano(s.nowFn())
	_, namespaces, err := resolveClusterNamespacesForQuery(now,
		xtime.ToUnixNano(queryStart),
		xtime.ToUnixNano(queryEnd),
		s.clusters,
		opts.FanoutOptions,
		opts.RestrictQueryOptions,
		opts.RelatedQueryOptions)
	if err != nil {
		return nil, err
	}

	results := make([]storagemetadata.Attributes, 0, len(namespaces))
	for _, ns := range namespaces {
		results = append(results, ns.Options().Attributes())
	}
	return results, nil
}

func (s *m3storage) ErrorBehavior() storage.ErrorBehavior {
	return storage.BehaviorFail
}

func (s *m3storage) Name() string {
	return "local_store"
}

// Find a reserved label target (one that begins with the reservedLabelPrefix)
// from an array of sorted labels.
func findReservedLabel(labels []prompb.Label, target []byte) []byte {
	// The target should always contain the reservedLabelPrefix.
	// If it doesn't, then we won't be able to find it within
	// the reserved labels by definition.
	if !bytes.HasPrefix(target, reservedLabelPrefix) {
		return nil
	}

	foundReservedLabels := false
	for idx := 0; idx < len(labels); idx++ {
		label := labels[idx]
		if !bytes.HasPrefix(label.Name, reservedLabelPrefix) {
			if foundReservedLabels {
				// We previously found reserved labels, and now that we've iterated
				// past the end of the section that contains them, we know the target
				// doesn't exist.
				return nil
			}
			// We haven't found reserve labels yet, so keep going.
			continue
		}

		// At this point we know that the current label contains the reservedLabelPrefix
		foundReservedLabels = true
		if bytes.Equal(label.Name, target) {
			return label.Value
		}
	}

	return nil
}

func calculateMetadataByName(result *prompb.QueryResult, metadata *block.ResultMetadata) {
	for _, series := range result.Timeseries {
		if series == nil {
			continue
		}

		name := findReservedLabel(series.Labels, promDefaultName)
		rollup := findReservedLabel(series.Labels, rollupTagName)
		if bytes.Equal(rollup, rollupTagValue) {
			metadata.ByName(name).Aggregated++
		} else {
			metadata.ByName(name).Unaggregated++
		}
	}
}

func (s *m3storage) FetchProm(
	ctx context.Context,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
) (storage.PromResult, error) {
	queryOptions, err := storage.FetchOptionsToM3Options(options, query)
	if err != nil {
		return storage.PromResult{}, err
	}

	accumulator, _, err := s.fetchCompressed(ctx, query, options, queryOptions)
	if err != nil {
		return storage.PromResult{}, err
	}

	defer accumulator.Close()
	result, attrs, err := accumulator.FinalResultWithAttrs()
	if err != nil {
		return storage.PromResult{}, err
	}

	resolutions := make([]time.Duration, 0, len(attrs))
	for _, attr := range attrs {
		resolutions = append(resolutions, attr.Resolution)
	}

	result.Metadata.Resolutions = resolutions
	fetchResult, err := storage.SeriesIteratorsToPromResult(
		ctx,
		result,
		s.opts.ReadWorkerPool(),
		s.opts.TagOptions(),
		s.opts.PromConvertOptions(),
		options,
	)
	if err != nil {
		return storage.PromResult{}, err
	}

	if options != nil && options.MaxMetricMetadataStats > 0 {
		calculateMetadataByName(fetchResult.PromResult, &fetchResult.Metadata)
	}

	return fetchResult, nil
}

// FetchResultToBlockResult converts an encoded SeriesIterator fetch result
// into blocks.
func FetchResultToBlockResult(
	result consolidators.SeriesFetchResult,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
	opts Options,
) (block.Result, error) {
	// If using multiblock, update options to reflect this.
	if options.BlockType == models.TypeMultiBlock {
		opts = opts.
			SetSplitSeriesByBlock(true)
	}

	start := query.Start
	bounds := models.Bounds{
		Start:    xtime.ToUnixNano(start),
		Duration: query.End.Sub(start),
		StepSize: query.Interval,
	}

	blocks, err := ConvertM3DBSeriesIterators(
		result,
		bounds,
		opts,
	)
	if err != nil {
		return block.Result{
			Metadata: block.NewResultMetadata(),
		}, err
	}

	return block.Result{
		Blocks:   blocks,
		Metadata: result.Metadata,
	}, nil
}

func (s *m3storage) FetchBlocks(
	ctx context.Context,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
) (block.Result, error) {
	// Override options with whatever is the current specified lookback duration.
	opts := s.opts.SetLookbackDuration(
		options.LookbackDurationOrDefault(s.opts.LookbackDuration()))

	result, _, err := s.FetchCompressedResult(ctx, query, options)
	if err != nil {
		return block.Result{
			Metadata: block.NewResultMetadata(),
		}, err
	}

	return FetchResultToBlockResult(result, query, options, opts)
}

func (s *m3storage) FetchCompressed(
	ctx context.Context,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
) (consolidators.MultiFetchResult, error) {
	queryOptions, _ := storage.FetchOptionsToM3Options(options, query)
	accumulator, _, err := s.fetchCompressed(ctx, query, options, queryOptions)
	return accumulator, err
}

func (s *m3storage) FetchCompressedResult(
	ctx context.Context,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
) (consolidators.SeriesFetchResult, Cleanup, error) {
	queryOptions, err := storage.FetchOptionsToM3Options(options, query)
	if err != nil {
		return consolidators.SeriesFetchResult{
			Metadata: block.NewResultMetadata(),
		}, noop, err
	}

	accumulator, m3query, err := s.fetchCompressed(ctx, query, options, queryOptions)
	if err != nil {
		return consolidators.SeriesFetchResult{
			Metadata: block.NewResultMetadata(),
		}, noop, err
	}

	result, attrs, err := accumulator.FinalResultWithAttrs()
	if err != nil {
		accumulator.Close()
		return result, noop, err
	}

	if processor := s.opts.SeriesIteratorProcessor(); processor != nil {
		_, span, sampled := xcontext.StartSampledTraceSpan(ctx,
			tracepoint.FetchCompressedInspectSeries)
		iters := result.SeriesIterators()
		if err := processor.InspectSeries(ctx, m3query, queryOptions, iters); err != nil {
			s.logger.Error("error inspecting series", zap.Error(err))
		}
		if sampled {
			span.LogFields(
				log.String("query", query.Raw),
				log.String("start", query.Start.String()),
				log.String("end", query.End.String()),
				log.String("interval", query.Interval.String()),
			)
		}
		span.Finish()
	}

	resolutions := make([]time.Duration, 0, len(attrs))
	for _, attr := range attrs {
		resolutions = append(resolutions, attr.Resolution)
	}

	result.Metadata.Resolutions = resolutions
	return result, accumulator.Close, nil
}

// fetches compressed series, returning a MultiFetchResult accumulator
func (s *m3storage) fetchCompressed(
	ctx context.Context,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
	queryOptions index.QueryOptions,
) (consolidators.MultiFetchResult, index.Query, error) {
	if err := options.BlockType.Validate(); err != nil {
		// This is an invariant error; should not be able to get to here.
		return nil, index.Query{}, instrument.InvariantErrorf("invalid block type on "+
			"fetch, got: %v with error %v", options.BlockType, err)
	}

	// Check if the query was interrupted.
	select {
	case <-ctx.Done():
		return nil, index.Query{}, ctx.Err()
	default:
	}

	m3query, err := storage.FetchQueryToM3Query(query, options)
	if err != nil {
		return nil, index.Query{}, err
	}

	var (
		queryStart = queryOptions.StartInclusive
		queryEnd   = queryOptions.EndExclusive
	)

	// NB(r): Since we don't use a single index we fan out to each
	// cluster that can completely fulfill this range and then prefer the
	// highest resolution (most fine grained) results.
	// This needs to be optimized, however this is a start.
	fanout, namespaces, err := resolveClusterNamespacesForQuery(
		xtime.ToUnixNano(s.nowFn()),
		queryStart,
		queryEnd,
		s.clusters,
		options.FanoutOptions,
		options.RestrictQueryOptions,
		options.RelatedQueryOptions,
	)
	if err != nil {
		return nil, index.Query{}, err
	}

	if s.logger.Core().Enabled(zapcore.DebugLevel) {
		for _, n := range namespaces {
			// NB(r): Need to perform log on inner loop, cannot reuse a
			// checked entry returned from logger.Check(...).
			// Will see: "Unsafe CheckedEntry re-use near Entry ..." otherwise.
			debugLog := s.logger.Check(zapcore.DebugLevel,
				"query resolved cluster namespace, will use most granular per result")
			if debugLog == nil {
				continue
			}

			debugLog.Write(
				zap.String("query", query.Raw),
				zap.String("m3query", m3query.String()),
				zap.Time("start", queryStart.ToTime()),
				zap.Time("narrowing.start", n.narrowing.start.ToTime()),
				zap.Time("end", queryEnd.ToTime()),
				zap.Time("narrowing.end", n.narrowing.end.ToTime()),
				zap.String("fanoutType", fanout.String()),
				zap.String("namespace", n.NamespaceID().String()),
				zap.String("type", n.Options().Attributes().MetricsType.String()),
				zap.String("retention", n.Options().Attributes().Retention.String()),
				zap.String("resolution", n.Options().Attributes().Resolution.String()),
				zap.Bool("remote", options.Remote))
		}
	}

	var wg sync.WaitGroup
	if len(namespaces) == 0 {
		return nil, index.Query{}, errNoNamespacesConfigured
	}

	matchOpts := s.opts.SeriesConsolidationMatchOptions()
	tagOpts := s.opts.TagOptions()
	limitOpts := consolidators.LimitOptions{
		Limit: options.SeriesLimit,
		// Piggy back on the new InstanceMultiple option to enable checking require exhaustive. This preserves the
		// existing buggy behavior of the coordinators not requiring exhaustive. Once InstanceMultiple is enabled by
		// default, this can be removed.
		RequireExhaustive: queryOptions.InstanceMultiple > 0 && options.RequireExhaustive,
	}
	result := consolidators.NewMultiFetchResult(fanout, matchOpts, tagOpts, limitOpts)
	for _, namespace := range namespaces {
		namespace := namespace // Capture var

		wg.Add(1)
		go func() {
			defer wg.Done()
			_, span, sampled := xcontext.StartSampledTraceSpan(ctx,
				tracepoint.FetchCompressedFetchTagged)
			defer span.Finish()

			session := namespace.Session()
			namespaceID := namespace.NamespaceID()
			narrowedQueryOpts := narrowQueryOpts(queryOptions, namespace)
			iters, metadata, err := session.FetchTagged(ctx, namespaceID, m3query, narrowedQueryOpts)
			if err == nil && sampled {
				span.LogFields(
					log.String("namespace", namespaceID.String()),
					log.Int("series", iters.Len()),
					log.Bool("exhaustive", metadata.Exhaustive),
					log.Int("responses", metadata.Responses),
					log.Int("estimateTotalBytes", metadata.EstimateTotalBytes),
				)
			}

			blockMeta := block.NewResultMetadata()
			blockMeta.AddNamespace(namespaceID.String())
			blockMeta.FetchedResponses = metadata.Responses
			blockMeta.FetchedBytesEstimate = metadata.EstimateTotalBytes
			blockMeta.Exhaustive = metadata.Exhaustive
			blockMeta.WaitedIndex = metadata.WaitedIndex
			blockMeta.WaitedSeriesRead = metadata.WaitedSeriesRead
			// Ignore error from getting iterator pools, since operation
			// will not be dramatically impacted if pools is nil
			result.Add(consolidators.MultiFetchResults{
				SeriesIterators: iters,
				Metadata:        blockMeta,
				Attrs:           namespace.Options().Attributes(),
				Err:             err,
			})
		}()
	}

	wg.Wait()

	// Check if the query was interrupted.
	select {
	case <-ctx.Done():
		return nil, index.Query{}, ctx.Err()
	default:
	}

	return result, m3query, err
}

func (s *m3storage) SearchSeries(
	ctx context.Context,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
) (*storage.SearchResults, error) {
	tagResult, cleanup, err := s.SearchCompressed(ctx, query, options)
	defer cleanup()
	if err != nil {
		return nil, err
	}

	metrics := make(models.Metrics, 0, len(tagResult.Tags))
	for _, result := range tagResult.Tags {
		m, err := storage.FromM3IdentToMetric(result.ID,
			result.Iter, s.opts.TagOptions())
		if err != nil {
			return nil, err
		}

		metrics = append(metrics, m)
	}

	return &storage.SearchResults{
		Metrics:  metrics,
		Metadata: tagResult.Metadata,
	}, nil
}

// CompleteTagsCompressed has the same behavior as CompleteTags.
func (s *m3storage) CompleteTagsCompressed(
	ctx context.Context,
	query *storage.CompleteTagsQuery,
	options *storage.FetchOptions,
) (*consolidators.CompleteTagsResult, error) {
	return s.CompleteTags(ctx, query, options)
}

func (s *m3storage) CompleteTags(
	ctx context.Context,
	query *storage.CompleteTagsQuery,
	options *storage.FetchOptions,
) (*consolidators.CompleteTagsResult, error) {
	// Check if the query was interrupted.
	select {
	case <-ctx.Done():
		return nil, ctx.Err()
	default:
	}

	fetchQuery := &storage.FetchQuery{
		TagMatchers: query.TagMatchers,
	}

	m3query, err := storage.FetchQueryToM3Query(fetchQuery, options)
	if err != nil {
		return nil, err
	}

	aggOpts, err := storage.FetchOptionsToAggregateOptions(options, query)
	if err != nil {
		return nil, err
	}

	var (
		queryStart      = aggOpts.StartInclusive
		queryEnd        = aggOpts.EndExclusive
		nameOnly        = query.CompleteNameOnly
		tagOpts         = s.opts.TagOptions()
		accumulatedTags = consolidators.NewCompleteTagsResultBuilder(nameOnly, tagOpts)
		multiErr        syncMultiErrs
		wg              sync.WaitGroup
	)

	debugLog := s.logger.Check(zapcore.DebugLevel,
		"completing tags")
	if debugLog != nil {
		filters := make([]string, len(query.FilterNameTags))
		for i, t := range query.FilterNameTags {
			filters[i] = string(t)
		}

		debugLog.Write(zap.Bool("nameOnly", nameOnly),
			zap.Strings("filterNames", filters),
			zap.String("matchers", query.TagMatchers.String()),
			zap.String("m3query", m3query.String()),
			zap.Time("start", queryStart.ToTime()),
			zap.Time("end", queryEnd.ToTime()),
			zap.Bool("remote", options.Remote),
		)
	}

	// NB(r): Since we don't use a single index we fan out to each
	// cluster that can completely fulfill this range and then prefer the
	// highest resolution (most fine-grained) results.
	// This needs to be optimized, however this is a start.
	_, namespaces, err := resolveClusterNamespacesForQuery(xtime.ToUnixNano(s.nowFn()),
		queryStart,
		queryEnd,
		s.clusters,
		options.FanoutOptions,
		options.RestrictQueryOptions,
		nil)
	if err != nil {
		return nil, err
	}

	var mu sync.Mutex
	aggIterators := make([]client.AggregatedTagsIterator, 0, len(namespaces))
	defer func() {
		mu.Lock()
		for _, it := range aggIterators {
			it.Finalize()
		}

		mu.Unlock()
	}()

	wg.Add(len(namespaces))
	for _, namespace := range namespaces {
		namespace := namespace // Capture var
		go func() {
			_, span, sampled := xcontext.StartSampledTraceSpan(ctx, tracepoint.CompleteTagsAggregate)
			defer func() {
				span.Finish()
				wg.Done()
			}()

			session := namespace.Session()
			namespaceID := namespace.NamespaceID()
			narrowedAggOpts := narrowAggOpts(aggOpts, namespace)
			aggTagIter, metadata, err := session.Aggregate(ctx, namespaceID, m3query, narrowedAggOpts)
			if err != nil {
				multiErr.add(err)
				return
			}

			if sampled {
				span.LogFields(
					log.String("namespace", namespaceID.String()),
					log.Int("results", aggTagIter.Remaining()),
					log.Bool("exhaustive", metadata.Exhaustive),
					log.Int("responses", metadata.Responses),
					log.Int("estimateTotalBytes", metadata.EstimateTotalBytes),
				)
			}

			mu.Lock()
			aggIterators = append(aggIterators, aggTagIter)
			mu.Unlock()

			completedTags := make([]consolidators.CompletedTag, 0, aggTagIter.Remaining())
			for aggTagIter.Next() {
				name, values := aggTagIter.Current()
				tagValues := make([][]byte, 0, values.Remaining())
				for values.Next() {
					tagValues = append(tagValues, values.Current().Bytes())
				}

				if err := values.Err(); err != nil {
					multiErr.add(err)
					return
				}

				completedTags = append(completedTags, consolidators.CompletedTag{
					Name:   name.Bytes(),
					Values: tagValues,
				})
			}

			if err := aggTagIter.Err(); err != nil {
				multiErr.add(err)
				return
			}

			blockMeta := block.NewResultMetadata()
			blockMeta.AddNamespace(namespaceID.String())
			blockMeta.FetchedResponses = metadata.Responses
			blockMeta.FetchedBytesEstimate = metadata.EstimateTotalBytes
			blockMeta.Exhaustive = metadata.Exhaustive
			blockMeta.WaitedIndex = metadata.WaitedIndex
			blockMeta.WaitedSeriesRead = metadata.WaitedSeriesRead
			result := &consolidators.CompleteTagsResult{
				CompleteNameOnly: query.CompleteNameOnly,
				CompletedTags:    completedTags,
				Metadata:         blockMeta,
			}

			if err := accumulatedTags.Add(result); err != nil {
				multiErr.add(err)
			}
		}()
	}

	wg.Wait()
	if err := multiErr.lastError(); err != nil {
		return nil, err
	}

	built := accumulatedTags.Build()
	return &built, nil
}

func (s *m3storage) SearchCompressed(
	ctx context.Context,
	query *storage.FetchQuery,
	options *storage.FetchOptions,
) (consolidators.TagResult, Cleanup, error) {
	// Check if the query was interrupted.
	tagResult := consolidators.TagResult{
		Metadata: block.NewResultMetadata(),
	}

	select {
	case <-ctx.Done():
		return tagResult, noop, ctx.Err()
	default:
	}

	m3query, err := storage.FetchQueryToM3Query(query, options)
	if err != nil {
		return tagResult, noop, err
	}

	m3opts, err := storage.FetchOptionsToM3Options(options, query)
	if err != nil {
		return tagResult, noop, err
	}

	var (
		queryStart = m3opts.StartInclusive
		queryEnd   = m3opts.EndExclusive
		result     = consolidators.NewMultiFetchTagsResult(s.opts.TagOptions())
		wg         sync.WaitGroup
	)

	// NB(r): Since we don't use a single index we fan out to each
	// cluster that can completely fulfill this range and then prefer the
	// highest resolution (most fine grained) results.
	// This needs to be optimized, however this is a start.
	_, namespaces, err := resolveClusterNamespacesForQuery(xtime.ToUnixNano(s.nowFn()),
		queryStart,
		queryEnd,
		s.clusters,
		options.FanoutOptions,
		options.RestrictQueryOptions,
		nil)
	if err != nil {
		return tagResult, noop, err
	}

	debugLog := s.logger.Check(zapcore.DebugLevel,
		"searching")
	if debugLog != nil {
		debugLog.Write(zap.String("query", query.Raw),
			zap.String("m3_query", m3query.String()),
			zap.Time("start", queryStart.ToTime()),
			zap.Time("end", queryEnd.ToTime()),
			zap.Bool("remote", options.Remote),
		)
	}

	wg.Add(len(namespaces))
	for _, namespace := range namespaces {
		namespace := namespace // Capture var
		go func() {
			_, span, sampled := xcontext.StartSampledTraceSpan(ctx,
				tracepoint.SearchCompressedFetchTaggedIDs)
			defer span.Finish()

			session := namespace.Session()
			namespaceID := namespace.NamespaceID()
			narrowedM3Opts := narrowQueryOpts(m3opts, namespace)
			iter, metadata, err := session.FetchTaggedIDs(ctx, namespaceID, m3query, narrowedM3Opts)
			if err == nil && sampled {
				span.LogFields(
					log.String("namespace", namespaceID.String()),
					log.Int("series", iter.Remaining()),
					log.Bool("exhaustive", metadata.Exhaustive),
					log.Int("responses", metadata.Responses),
					log.Int("estimateTotalBytes", metadata.EstimateTotalBytes),
				)
			}

			blockMeta := block.NewResultMetadata()
			blockMeta.AddNamespace(namespaceID.String())
			blockMeta.FetchedResponses = metadata.Responses
			blockMeta.FetchedBytesEstimate = metadata.EstimateTotalBytes
			blockMeta.Exhaustive = metadata.Exhaustive
			blockMeta.WaitedIndex = metadata.WaitedIndex
			blockMeta.WaitedSeriesRead = metadata.WaitedSeriesRead
			result.Add(iter, blockMeta, err)
			wg.Done()
		}()
	}

	wg.Wait()

	tagResult, err = result.FinalResult()
	return tagResult, result.Close, err
}

func (s *m3storage) Write(
	ctx context.Context,
	query *storage.WriteQuery,
) error {
	if query == nil {
		return errors.ErrNilWriteQuery
	}

	var (
		// TODO: Pool this once an ident pool is setup. We will have
		// to stop calling NoFinalize() below if we do that.
		tags       = query.Tags()
		datapoints = query.Datapoints()
		idBuf      = tags.ID()
		id         = ident.BytesID(idBuf)
		err        error
		namespace  ClusterNamespace
		exists     bool
	)

	attributes := query.Attributes()
	switch attributes.MetricsType {
	case storagemetadata.UnaggregatedMetricsType:
		namespace, exists = s.clusters.UnaggregatedClusterNamespace()
		if !exists {
			err = errUnaggregatedNamespaceUninitialized
		}
	case storagemetadata.AggregatedMetricsType:
		attrs := RetentionResolution{
			Retention:  attributes.Retention,
			Resolution: attributes.Resolution,
		}
		namespace, exists = s.clusters.AggregatedClusterNamespace(attrs)
		if !exists {
			err = fmt.Errorf("no configured cluster namespace for: retention=%s,"+
				" resolution=%s", attrs.Retention.String(), attrs.Resolution.String())
			break
		}
		if namespace.Options().ReadOnly() {
			err = fmt.Errorf(
				"cannot write to read only namespace %s (%s:%s)",
				namespace.NamespaceID(), attrs.Resolution.String(), attrs.Retention.String())
		}
	default:
		metricsType := attributes.MetricsType
		err = fmt.Errorf("invalid write request metrics type: %s (%d)",
			metricsType.String(), uint(metricsType))
	}
	if err != nil {
		return err
	}

	// Set id to NoFinalize to avoid cloning it in write operations
	id.NoFinalize()

	if s.opts.RateLimiter().Limit(ctx, namespace, datapoints, tags.Tags) {
		return xerrors.NewResourceExhaustedError(goerrors.New("rate limit exceeded"))
	}

	tags.Tags, err = s.opts.TagsTransform()(ctx, namespace, tags.Tags)
	if err != nil {
		return err
	}
	tagIterator := storage.TagsToIdentTagIterator(tags)

	if len(datapoints) == 1 {
		// Special case single datapoint because it is common and we
		// can avoid the overhead of a waitgroup, goroutine, multierr,
		// iterator duplication etc.
		return s.writeSingle(query, datapoints[0], id, tagIterator, namespace)
	}

	var (
		wg       sync.WaitGroup
		multiErr syncMultiErrs
	)

	for _, datapoint := range datapoints {
		tagIter := tagIterator.Duplicate()
		// capture var
		datapoint := datapoint
		wg.Add(1)

		var (
			now                      = time.Now()
			deadline, deadlineExists = ctx.Deadline()
			timeout                  = minWriteWaitTimeout
		)
		if deadlineExists {
			if remain := deadline.Sub(now); remain >= timeout {
				timeout = remain
			}
		}
		spawned := s.opts.WriteWorkerPool().GoWithTimeout(func() {
			if err := s.writeSingle(query, datapoint, id, tagIter, namespace); err != nil {
				multiErr.add(err)
			}

			tagIter.Close()
			wg.Done()
		}, timeout)
		if !spawned {
			multiErr.add(fmt.Errorf("timeout exceeded waiting: %v", timeout))
		}
	}

	wg.Wait()
	return multiErr.lastError()
}

func (s *m3storage) Type() storage.Type {
	return storage.TypeLocalDC
}

func (s *m3storage) Close() error {
	return nil
}

func (s *m3storage) writeSingle(
	query *storage.WriteQuery,
	datapoint ts.Datapoint,
	identID ident.ID,
	iterator ident.TagIterator,
	namespace ClusterNamespace,
) error {
	namespaceID := namespace.NamespaceID()
	session := namespace.Session()
	return session.WriteTagged(namespaceID, identID, iterator,
		datapoint.Timestamp, datapoint.Value, query.Unit(), query.Annotation())
}

func narrowQueryOpts(o index.QueryOptions, namespace resolvedNamespace) index.QueryOptions {
	narrowed := o
	if !namespace.narrowing.start.IsZero() && namespace.narrowing.start.After(o.StartInclusive) {
		narrowed.StartInclusive = namespace.narrowing.start
	}
	if !namespace.narrowing.end.IsZero() && namespace.narrowing.end.Before(o.EndExclusive) {
		narrowed.EndExclusive = namespace.narrowing.end
	}

	return narrowed
}

func narrowAggOpts(o index.AggregationOptions, namespace resolvedNamespace) index.AggregationOptions {
	narrowed := o
	narrowed.QueryOptions = narrowQueryOpts(o.QueryOptions, namespace)

	return narrowed
}