Fix issue where queries can fail or omit OOO samples if OOO head compaction occurs between creating a querier and reading chunks (#13115)

* Add failing test. Signed-off-by: Charles Korn <charles.korn@grafana.com> * Don't run OOO head garbage collection while reads are running. Signed-off-by: Charles Korn <charles.korn@grafana.com> * Add further test cases for different order of operations. Signed-off-by: Charles Korn <charles.korn@grafana.com> * Ensure all queriers are closed if `DB.blockChunkQuerierForRange()` fails. Signed-off-by: Charles Korn <charles.korn@grafana.com> * Ensure all queriers are closed if `DB.Querier()` fails. Signed-off-by: Charles Korn <charles.korn@grafana.com> * Invert error handling in `DB.Querier()` and `DB.blockChunkQuerierForRange()` to make it clearer Signed-off-by: Charles Korn <charles.korn@grafana.com> * Ensure that queries that touch OOO data can't block OOO head garbage collection forever. Signed-off-by: Charles Korn <charles.korn@grafana.com> * Address PR feedback: fix parameter name in comment Co-authored-by: Jesus Vazquez <jesusvazquez@users.noreply.github.com> Signed-off-by: Charles Korn <charleskorn@users.noreply.github.com> * Address PR feedback: use `lastGarbageCollectedMmapRef` Signed-off-by: Charles Korn <charles.korn@grafana.com> * Address PR feedback: ensure pending reads are cleaned up if creating an OOO querier fails Signed-off-by: Charles Korn <charles.korn@grafana.com> --------- Signed-off-by: Charles Korn <charles.korn@grafana.com> Signed-off-by: Charles Korn <charleskorn@users.noreply.github.com> Co-authored-by: Jesus Vazquez <jesusvazquez@users.noreply.github.com>
2024-11-09 23:24:05 -08:00 · 2023-11-24 22:38:38 +11:00 · 2023-11-24 22:38:38 +11:00 · 59844498f7
parent 35a15e8f04
commit 59844498f7
9 changed files with 546 additions and 86 deletions
--- a/tsdb/db.go
+++ b/tsdb/db.go
@ -203,10 +203,14 @@ type DB struct {
 	compactor      Compactor
 	blocksToDelete BlocksToDeleteFunc

-	// Mutex for that must be held when modifying the general block layout.
+	// Mutex for that must be held when modifying the general block layout or lastGarbageCollectedMmapRef.
 	mtx    sync.RWMutex
 	blocks []*Block

+	// The last OOO chunk that was compacted and written to disk. New queriers must not read chunks less
+	// than or equal to this reference, as these chunks could be garbage collected at any time.
+	lastGarbageCollectedMmapRef chunks.ChunkDiskMapperRef
+
 	head *Head

 	compactc chan struct{}
@ -1243,6 +1247,20 @@ func (db *DB) compactOOOHead(ctx context.Context) error {

 	lastWBLFile, minOOOMmapRef := oooHead.LastWBLFile(), oooHead.LastMmapRef()
 	if lastWBLFile != 0 || minOOOMmapRef != 0 {
+		if minOOOMmapRef != 0 {
+			// Ensure that no more queriers are created that will reference chunks we're about to garbage collect.
+			// truncateOOO waits for any existing queriers that reference chunks we're about to garbage collect to
+			// complete before running garbage collection, so we don't need to do that here.
+			//
+			// We take mtx to ensure that Querier() and ChunkQuerier() don't miss blocks: without this, they could
+			// capture the list of blocks before the call to reloadBlocks() above runs, but then capture
+			// lastGarbageCollectedMmapRef after we update it here, and therefore not query either the blocks we've just
+			// written or the head chunks those blocks were created from.
+			db.mtx.Lock()
+			db.lastGarbageCollectedMmapRef = minOOOMmapRef
+			db.mtx.Unlock()
+		}
+
 		if err := db.head.truncateOOO(lastWBLFile, minOOOMmapRef); err != nil {
 			return errors.Wrap(err, "truncate ooo wbl")
 		}
@ -1869,7 +1887,7 @@ func (db *DB) Snapshot(dir string, withHead bool) error {
 }

 // Querier returns a new querier over the data partition for the given time range.
-func (db *DB) Querier(mint, maxt int64) (storage.Querier, error) {
+func (db *DB) Querier(mint, maxt int64) (_ storage.Querier, err error) {
 	var blocks []BlockReader

 	db.mtx.RLock()
@ -1880,11 +1898,23 @@ func (db *DB) Querier(mint, maxt int64) (storage.Querier, error) {
 			blocks = append(blocks, b)
 		}
 	}
-	var inOrderHeadQuerier storage.Querier
+
+	blockQueriers := make([]storage.Querier, 0, len(blocks)+2) // +2 to allow for possible in-order and OOO head queriers
+
+	defer func() {
+		if err != nil {
+			// If we fail, all previously opened queriers must be closed.
+			for _, q := range blockQueriers {
+				// TODO(bwplotka): Handle error.
+				_ = q.Close()
+			}
+		}
+	}()
+
 	if maxt >= db.head.MinTime() {
 		rh := NewRangeHead(db.head, mint, maxt)
 		var err error
-		inOrderHeadQuerier, err = NewBlockQuerier(rh, mint, maxt)
+		inOrderHeadQuerier, err := NewBlockQuerier(rh, mint, maxt)
 		if err != nil {
 			return nil, errors.Wrapf(err, "open block querier for head %s", rh)
 		}
@ -1906,44 +1936,40 @@ func (db *DB) Querier(mint, maxt int64) (storage.Querier, error) {
 				return nil, errors.Wrapf(err, "open block querier for head while getting new querier %s", rh)
 			}
 		}
+
+		if inOrderHeadQuerier != nil {
+			blockQueriers = append(blockQueriers, inOrderHeadQuerier)
+		}
 	}

-	var outOfOrderHeadQuerier storage.Querier
 	if overlapsClosedInterval(mint, maxt, db.head.MinOOOTime(), db.head.MaxOOOTime()) {
-		rh := NewOOORangeHead(db.head, mint, maxt)
+		rh := NewOOORangeHead(db.head, mint, maxt, db.lastGarbageCollectedMmapRef)
 		var err error
-		outOfOrderHeadQuerier, err = NewBlockQuerier(rh, mint, maxt)
+		outOfOrderHeadQuerier, err := NewBlockQuerier(rh, mint, maxt)
 		if err != nil {
+			// If NewBlockQuerier() failed, make sure to clean up the pending read created by NewOOORangeHead.
+			rh.isoState.Close()
+
 			return nil, errors.Wrapf(err, "open block querier for ooo head %s", rh)
 		}
-	}

-	blockQueriers := make([]storage.Querier, 0, len(blocks))
-	for _, b := range blocks {
-		q, err := NewBlockQuerier(b, mint, maxt)
-		if err == nil {
-			blockQueriers = append(blockQueriers, q)
-			continue
-		}
-		// If we fail, all previously opened queriers must be closed.
-		for _, q := range blockQueriers {
-			// TODO(bwplotka): Handle error.
-			_ = q.Close()
-		}
-		return nil, errors.Wrapf(err, "open querier for block %s", b)
-	}
-	if inOrderHeadQuerier != nil {
-		blockQueriers = append(blockQueriers, inOrderHeadQuerier)
-	}
-	if outOfOrderHeadQuerier != nil {
 		blockQueriers = append(blockQueriers, outOfOrderHeadQuerier)
 	}
+
+	for _, b := range blocks {
+		q, err := NewBlockQuerier(b, mint, maxt)
+		if err != nil {
+			return nil, errors.Wrapf(err, "open querier for block %s", b)
+		}
+		blockQueriers = append(blockQueriers, q)
+	}
+
 	return storage.NewMergeQuerier(blockQueriers, nil, storage.ChainedSeriesMerge), nil
 }

 // blockChunkQuerierForRange returns individual block chunk queriers from the persistent blocks, in-order head block, and the
 // out-of-order head block, overlapping with the given time range.
-func (db *DB) blockChunkQuerierForRange(mint, maxt int64) ([]storage.ChunkQuerier, error) {
+func (db *DB) blockChunkQuerierForRange(mint, maxt int64) (_ []storage.ChunkQuerier, err error) {
 	var blocks []BlockReader

 	db.mtx.RLock()
@ -1954,11 +1980,22 @@ func (db *DB) blockChunkQuerierForRange(mint, maxt int64) ([]storage.ChunkQuerie
 			blocks = append(blocks, b)
 		}
 	}
-	var inOrderHeadQuerier storage.ChunkQuerier
+
+	blockQueriers := make([]storage.ChunkQuerier, 0, len(blocks)+2) // +2 to allow for possible in-order and OOO head queriers
+
+	defer func() {
+		if err != nil {
+			// If we fail, all previously opened queriers must be closed.
+			for _, q := range blockQueriers {
+				// TODO(bwplotka): Handle error.
+				_ = q.Close()
+			}
+		}
+	}()
+
 	if maxt >= db.head.MinTime() {
 		rh := NewRangeHead(db.head, mint, maxt)
-		var err error
-		inOrderHeadQuerier, err = NewBlockChunkQuerier(rh, mint, maxt)
+		inOrderHeadQuerier, err := NewBlockChunkQuerier(rh, mint, maxt)
 		if err != nil {
 			return nil, errors.Wrapf(err, "open querier for head %s", rh)
 		}
@ -1980,37 +2017,28 @@ func (db *DB) blockChunkQuerierForRange(mint, maxt int64) ([]storage.ChunkQuerie
 				return nil, errors.Wrapf(err, "open querier for head while getting new querier %s", rh)
 			}
 		}
+
+		if inOrderHeadQuerier != nil {
+			blockQueriers = append(blockQueriers, inOrderHeadQuerier)
+		}
 	}

-	var outOfOrderHeadQuerier storage.ChunkQuerier
 	if overlapsClosedInterval(mint, maxt, db.head.MinOOOTime(), db.head.MaxOOOTime()) {
-		rh := NewOOORangeHead(db.head, mint, maxt)
-		var err error
-		outOfOrderHeadQuerier, err = NewBlockChunkQuerier(rh, mint, maxt)
+		rh := NewOOORangeHead(db.head, mint, maxt, db.lastGarbageCollectedMmapRef)
+		outOfOrderHeadQuerier, err := NewBlockChunkQuerier(rh, mint, maxt)
 		if err != nil {
 			return nil, errors.Wrapf(err, "open block chunk querier for ooo head %s", rh)
 		}
+
+		blockQueriers = append(blockQueriers, outOfOrderHeadQuerier)
 	}

-	blockQueriers := make([]storage.ChunkQuerier, 0, len(blocks))
 	for _, b := range blocks {
 		q, err := NewBlockChunkQuerier(b, mint, maxt)
-		if err == nil {
-			blockQueriers = append(blockQueriers, q)
-			continue
+		if err != nil {
+			return nil, errors.Wrapf(err, "open querier for block %s", b)
 		}
-		// If we fail, all previously opened queriers must be closed.
-		for _, q := range blockQueriers {
-			// TODO(bwplotka): Handle error.
-			_ = q.Close()
-		}
-		return nil, errors.Wrapf(err, "open querier for block %s", b)
-	}
-	if inOrderHeadQuerier != nil {
-		blockQueriers = append(blockQueriers, inOrderHeadQuerier)
-	}
-	if outOfOrderHeadQuerier != nil {
-		blockQueriers = append(blockQueriers, outOfOrderHeadQuerier)
+		blockQueriers = append(blockQueriers, q)
 	}

 	return blockQueriers, nil
--- a/tsdb/db_test.go
+++ b/tsdb/db_test.go
@ -38,6 +38,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 	prom_testutil "github.com/prometheus/client_golang/prometheus/testutil"
 	"github.com/stretchr/testify/require"
+	"go.uber.org/atomic"
 	"go.uber.org/goleak"

 	"github.com/prometheus/prometheus/config"
@ -3611,6 +3612,264 @@ func testChunkQuerierShouldNotPanicIfHeadChunkIsTruncatedWhileReadingQueriedChun
 	}
 }

+func TestQuerierShouldNotFailIfOOOCompactionOccursAfterRetrievingQuerier(t *testing.T) {
+	opts := DefaultOptions()
+	opts.OutOfOrderTimeWindow = 3 * DefaultBlockDuration
+	db := openTestDB(t, opts, nil)
+	defer func() {
+		require.NoError(t, db.Close())
+	}()
+
+	// Disable compactions so we can control it.
+	db.DisableCompactions()
+
+	metric := labels.FromStrings(labels.MetricName, "test_metric")
+	ctx := context.Background()
+	interval := int64(15 * time.Second / time.Millisecond)
+	ts := int64(0)
+	samplesWritten := 0
+
+	// Capture the first timestamp - this will be the timestamp of the OOO sample we'll append below.
+	oooTS := ts
+	ts += interval
+
+	// Push samples after the OOO sample we'll write below.
+	for ; ts < 10*interval; ts += interval {
+		app := db.Appender(ctx)
+		_, err := app.Append(0, metric, ts, float64(ts))
+		require.NoError(t, err)
+		require.NoError(t, app.Commit())
+		samplesWritten++
+	}
+
+	// Push a single OOO sample.
+	app := db.Appender(ctx)
+	_, err := app.Append(0, metric, oooTS, float64(ts))
+	require.NoError(t, err)
+	require.NoError(t, app.Commit())
+	samplesWritten++
+
+	// Get a querier.
+	querierCreatedBeforeCompaction, err := db.ChunkQuerier(0, math.MaxInt64)
+	require.NoError(t, err)
+
+	// Start OOO head compaction.
+	compactionComplete := atomic.NewBool(false)
+	go func() {
+		defer compactionComplete.Store(true)
+
+		require.NoError(t, db.CompactOOOHead(ctx))
+		require.Equal(t, float64(1), prom_testutil.ToFloat64(db.Head().metrics.chunksRemoved))
+	}()
+
+	// Give CompactOOOHead time to start work.
+	// If it does not wait for querierCreatedBeforeCompaction to be closed, then the query will return incorrect results or fail.
+	time.Sleep(time.Second)
+	require.False(t, compactionComplete.Load(), "compaction completed before reading chunks or closing querier created before compaction")
+
+	// Get another querier. This one should only use the compacted blocks from disk and ignore the chunks that will be garbage collected.
+	querierCreatedAfterCompaction, err := db.ChunkQuerier(0, math.MaxInt64)
+	require.NoError(t, err)
+
+	testQuerier := func(q storage.ChunkQuerier) {
+		// Query back the series.
+		hints := &storage.SelectHints{Start: 0, End: math.MaxInt64, Step: interval}
+		seriesSet := q.Select(ctx, true, hints, labels.MustNewMatcher(labels.MatchEqual, labels.MetricName, "test_metric"))
+
+		// Collect the iterator for the series.
+		var iterators []chunks.Iterator
+		for seriesSet.Next() {
+			iterators = append(iterators, seriesSet.At().Iterator(nil))
+		}
+		require.NoError(t, seriesSet.Err())
+		require.Len(t, iterators, 1)
+		iterator := iterators[0]
+
+		// Check that we can still successfully read all samples.
+		samplesRead := 0
+		for iterator.Next() {
+			samplesRead += iterator.At().Chunk.NumSamples()
+		}
+
+		require.NoError(t, iterator.Err())
+		require.Equal(t, samplesWritten, samplesRead)
+	}
+
+	testQuerier(querierCreatedBeforeCompaction)
+
+	require.False(t, compactionComplete.Load(), "compaction completed before closing querier created before compaction")
+	require.NoError(t, querierCreatedBeforeCompaction.Close())
+	require.Eventually(t, compactionComplete.Load, time.Second, 10*time.Millisecond, "compaction should complete after querier created before compaction was closed, and not wait for querier created after compaction")
+
+	// Use the querier created after compaction and confirm it returns the expected results (ie. from the disk block created from OOO head and in-order head) without error.
+	testQuerier(querierCreatedAfterCompaction)
+	require.NoError(t, querierCreatedAfterCompaction.Close())
+}
+
+func TestQuerierShouldNotFailIfOOOCompactionOccursAfterSelecting(t *testing.T) {
+	opts := DefaultOptions()
+	opts.OutOfOrderTimeWindow = 3 * DefaultBlockDuration
+	db := openTestDB(t, opts, nil)
+	defer func() {
+		require.NoError(t, db.Close())
+	}()
+
+	// Disable compactions so we can control it.
+	db.DisableCompactions()
+
+	metric := labels.FromStrings(labels.MetricName, "test_metric")
+	ctx := context.Background()
+	interval := int64(15 * time.Second / time.Millisecond)
+	ts := int64(0)
+	samplesWritten := 0
+
+	// Capture the first timestamp - this will be the timestamp of the OOO sample we'll append below.
+	oooTS := ts
+	ts += interval
+
+	// Push samples after the OOO sample we'll write below.
+	for ; ts < 10*interval; ts += interval {
+		app := db.Appender(ctx)
+		_, err := app.Append(0, metric, ts, float64(ts))
+		require.NoError(t, err)
+		require.NoError(t, app.Commit())
+		samplesWritten++
+	}
+
+	// Push a single OOO sample.
+	app := db.Appender(ctx)
+	_, err := app.Append(0, metric, oooTS, float64(ts))
+	require.NoError(t, err)
+	require.NoError(t, app.Commit())
+	samplesWritten++
+
+	// Get a querier.
+	querier, err := db.ChunkQuerier(0, math.MaxInt64)
+	require.NoError(t, err)
+
+	// Query back the series.
+	hints := &storage.SelectHints{Start: 0, End: math.MaxInt64, Step: interval}
+	seriesSet := querier.Select(ctx, true, hints, labels.MustNewMatcher(labels.MatchEqual, labels.MetricName, "test_metric"))
+
+	// Start OOO head compaction.
+	compactionComplete := atomic.NewBool(false)
+	go func() {
+		defer compactionComplete.Store(true)
+
+		require.NoError(t, db.CompactOOOHead(ctx))
+		require.Equal(t, float64(1), prom_testutil.ToFloat64(db.Head().metrics.chunksRemoved))
+	}()
+
+	// Give CompactOOOHead time to start work.
+	// If it does not wait for the querier to be closed, then the query will return incorrect results or fail.
+	time.Sleep(time.Second)
+	require.False(t, compactionComplete.Load(), "compaction completed before reading chunks or closing querier")
+
+	// Collect the iterator for the series.
+	var iterators []chunks.Iterator
+	for seriesSet.Next() {
+		iterators = append(iterators, seriesSet.At().Iterator(nil))
+	}
+	require.NoError(t, seriesSet.Err())
+	require.Len(t, iterators, 1)
+	iterator := iterators[0]
+
+	// Check that we can still successfully read all samples.
+	samplesRead := 0
+	for iterator.Next() {
+		samplesRead += iterator.At().Chunk.NumSamples()
+	}
+
+	require.NoError(t, iterator.Err())
+	require.Equal(t, samplesWritten, samplesRead)
+
+	require.False(t, compactionComplete.Load(), "compaction completed before closing querier")
+	require.NoError(t, querier.Close())
+	require.Eventually(t, compactionComplete.Load, time.Second, 10*time.Millisecond, "compaction should complete after querier was closed")
+}
+
+func TestQuerierShouldNotFailIfOOOCompactionOccursAfterRetrievingIterators(t *testing.T) {
+	opts := DefaultOptions()
+	opts.OutOfOrderTimeWindow = 3 * DefaultBlockDuration
+	db := openTestDB(t, opts, nil)
+	defer func() {
+		require.NoError(t, db.Close())
+	}()
+
+	// Disable compactions so we can control it.
+	db.DisableCompactions()
+
+	metric := labels.FromStrings(labels.MetricName, "test_metric")
+	ctx := context.Background()
+	interval := int64(15 * time.Second / time.Millisecond)
+	ts := int64(0)
+	samplesWritten := 0
+
+	// Capture the first timestamp - this will be the timestamp of the OOO sample we'll append below.
+	oooTS := ts
+	ts += interval
+
+	// Push samples after the OOO sample we'll write below.
+	for ; ts < 10*interval; ts += interval {
+		app := db.Appender(ctx)
+		_, err := app.Append(0, metric, ts, float64(ts))
+		require.NoError(t, err)
+		require.NoError(t, app.Commit())
+		samplesWritten++
+	}
+
+	// Push a single OOO sample.
+	app := db.Appender(ctx)
+	_, err := app.Append(0, metric, oooTS, float64(ts))
+	require.NoError(t, err)
+	require.NoError(t, app.Commit())
+	samplesWritten++
+
+	// Get a querier.
+	querier, err := db.ChunkQuerier(0, math.MaxInt64)
+	require.NoError(t, err)
+
+	// Query back the series.
+	hints := &storage.SelectHints{Start: 0, End: math.MaxInt64, Step: interval}
+	seriesSet := querier.Select(ctx, true, hints, labels.MustNewMatcher(labels.MatchEqual, labels.MetricName, "test_metric"))
+
+	// Collect the iterator for the series.
+	var iterators []chunks.Iterator
+	for seriesSet.Next() {
+		iterators = append(iterators, seriesSet.At().Iterator(nil))
+	}
+	require.NoError(t, seriesSet.Err())
+	require.Len(t, iterators, 1)
+	iterator := iterators[0]
+
+	// Start OOO head compaction.
+	compactionComplete := atomic.NewBool(false)
+	go func() {
+		defer compactionComplete.Store(true)
+
+		require.NoError(t, db.CompactOOOHead(ctx))
+		require.Equal(t, float64(1), prom_testutil.ToFloat64(db.Head().metrics.chunksRemoved))
+	}()
+
+	// Give CompactOOOHead time to start work.
+	// If it does not wait for the querier to be closed, then the query will return incorrect results or fail.
+	time.Sleep(time.Second)
+	require.False(t, compactionComplete.Load(), "compaction completed before reading chunks or closing querier")
+
+	// Check that we can still successfully read all samples.
+	samplesRead := 0
+	for iterator.Next() {
+		samplesRead += iterator.At().Chunk.NumSamples()
+	}
+
+	require.NoError(t, iterator.Err())
+	require.Equal(t, samplesWritten, samplesRead)
+
+	require.False(t, compactionComplete.Load(), "compaction completed before closing querier")
+	require.NoError(t, querier.Close())
+	require.Eventually(t, compactionComplete.Load, time.Second, 10*time.Millisecond, "compaction should complete after querier was closed")
+}
+
 func newTestDB(t *testing.T) *DB {
 	dir := t.TempDir()

--- a/tsdb/head.go
+++ b/tsdb/head.go
@ -106,6 +106,8 @@ type Head struct {

 	iso *isolation

+	oooIso *oooIsolation
+
 	cardinalityMutex      sync.Mutex
 	cardinalityCache      *index.PostingsStats // Posting stats cache which will expire after 30sec.
 	lastPostingsStatsCall time.Duration        // Last posting stats call (PostingsCardinalityStats()) time for caching.
@ -300,6 +302,7 @@ func (h *Head) resetInMemoryState() error {
 	}

 	h.iso = newIsolation(h.opts.IsolationDisabled)
+	h.oooIso = newOOOIsolation()

 	h.exemplarMetrics = em
 	h.exemplars = es
@ -1133,6 +1136,14 @@ func (h *Head) WaitForPendingReadersInTimeRange(mint, maxt int64) {
 	}
 }

+// WaitForPendingReadersForOOOChunksAtOrBefore is like WaitForPendingReadersInTimeRange, except it waits for
+// queries touching OOO chunks less than or equal to chunk to finish querying.
+func (h *Head) WaitForPendingReadersForOOOChunksAtOrBefore(chunk chunks.ChunkDiskMapperRef) {
+	for h.oooIso.HasOpenReadsAtOrBefore(chunk) {
+		time.Sleep(500 * time.Millisecond)
+	}
+}
+
 // WaitForAppendersOverlapping waits for appends overlapping maxt to finish.
 func (h *Head) WaitForAppendersOverlapping(maxt int64) {
 	for maxt >= h.iso.lowestAppendTime() {
@ -1271,13 +1282,19 @@ func (h *Head) truncateWAL(mint int64) error {
 }

 // truncateOOO
+//   - waits for any pending reads that potentially touch chunks less than or equal to newMinOOOMmapRef
 //   - truncates the OOO WBL files whose index is strictly less than lastWBLFile.
-//   - garbage collects all the m-map chunks from the memory that are less than or equal to minOOOMmapRef
+//   - garbage collects all the m-map chunks from the memory that are less than or equal to newMinOOOMmapRef
 //     and then deletes the series that do not have any data anymore.
-func (h *Head) truncateOOO(lastWBLFile int, minOOOMmapRef chunks.ChunkDiskMapperRef) error {
+//
+// The caller is responsible for ensuring that no further queriers will be created that reference chunks less
+// than or equal to newMinOOOMmapRef before calling truncateOOO.
+func (h *Head) truncateOOO(lastWBLFile int, newMinOOOMmapRef chunks.ChunkDiskMapperRef) error {
 	curMinOOOMmapRef := chunks.ChunkDiskMapperRef(h.minOOOMmapRef.Load())
-	if minOOOMmapRef.GreaterThan(curMinOOOMmapRef) {
-		h.minOOOMmapRef.Store(uint64(minOOOMmapRef))
+	if newMinOOOMmapRef.GreaterThan(curMinOOOMmapRef) {
+		h.WaitForPendingReadersForOOOChunksAtOrBefore(newMinOOOMmapRef)
+		h.minOOOMmapRef.Store(uint64(newMinOOOMmapRef))
+
 		if err := h.truncateSeriesAndChunkDiskMapper("truncateOOO"); err != nil {
 			return err
 		}
--- a/tsdb/ooo_head.go
+++ b/tsdb/ooo_head.go
@ -20,6 +20,7 @@ import (
 	"github.com/oklog/ulid"

 	"github.com/prometheus/prometheus/tsdb/chunkenc"
+	"github.com/prometheus/prometheus/tsdb/chunks"
 	"github.com/prometheus/prometheus/tsdb/tombstones"
 )

@ -113,22 +114,27 @@ type OOORangeHead struct {
 	// the timerange of the query and having preexisting pointers to the first
 	// and last timestamp help with that.
 	mint, maxt int64
+
+	isoState *oooIsolationState
 }

-func NewOOORangeHead(head *Head, mint, maxt int64) *OOORangeHead {
+func NewOOORangeHead(head *Head, mint, maxt int64, minRef chunks.ChunkDiskMapperRef) *OOORangeHead {
+	isoState := head.oooIso.TrackReadAfter(minRef)
+
 	return &OOORangeHead{
-		head: head,
-		mint: mint,
-		maxt: maxt,
+		head:     head,
+		mint:     mint,
+		maxt:     maxt,
+		isoState: isoState,
 	}
 }

 func (oh *OOORangeHead) Index() (IndexReader, error) {
-	return NewOOOHeadIndexReader(oh.head, oh.mint, oh.maxt), nil
+	return NewOOOHeadIndexReader(oh.head, oh.mint, oh.maxt, oh.isoState.minRef), nil
 }

 func (oh *OOORangeHead) Chunks() (ChunkReader, error) {
-	return NewOOOHeadChunkReader(oh.head, oh.mint, oh.maxt), nil
+	return NewOOOHeadChunkReader(oh.head, oh.mint, oh.maxt, oh.isoState), nil
 }

 func (oh *OOORangeHead) Tombstones() (tombstones.Reader, error) {
--- a/tsdb/ooo_head_read.go
+++ b/tsdb/ooo_head_read.go
@ -38,26 +38,29 @@ var _ IndexReader = &OOOHeadIndexReader{}
 // decided to do this to avoid code duplication.
 // The only methods that change are the ones about getting Series and Postings.
 type OOOHeadIndexReader struct {
-	*headIndexReader // A reference to the headIndexReader so we can reuse as many interface implementation as possible.
+	*headIndexReader            // A reference to the headIndexReader so we can reuse as many interface implementation as possible.
+	lastGarbageCollectedMmapRef chunks.ChunkDiskMapperRef
 }

-func NewOOOHeadIndexReader(head *Head, mint, maxt int64) *OOOHeadIndexReader {
+func NewOOOHeadIndexReader(head *Head, mint, maxt int64, lastGarbageCollectedMmapRef chunks.ChunkDiskMapperRef) *OOOHeadIndexReader {
 	hr := &headIndexReader{
 		head: head,
 		mint: mint,
 		maxt: maxt,
 	}
-	return &OOOHeadIndexReader{hr}
+	return &OOOHeadIndexReader{hr, lastGarbageCollectedMmapRef}
 }

 func (oh *OOOHeadIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta) error {
-	return oh.series(ref, builder, chks, 0)
+	return oh.series(ref, builder, chks, oh.lastGarbageCollectedMmapRef, 0)
 }

-// The passed lastMmapRef tells upto what max m-map chunk that we can consider.
-// If it is 0, it means all chunks need to be considered.
-// If it is non-0, then the oooHeadChunk must not be considered.
-func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta, lastMmapRef chunks.ChunkDiskMapperRef) error {
+// lastGarbageCollectedMmapRef gives the last mmap chunk that may be being garbage collected and so
+// any chunk at or before this ref will not be considered. 0 disables this check.
+//
+// maxMmapRef tells upto what max m-map chunk that we can consider. If it is non-0, then
+// the oooHeadChunk will not be considered.
+func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta, lastGarbageCollectedMmapRef, maxMmapRef chunks.ChunkDiskMapperRef) error {
 	s := oh.head.series.getByID(chunks.HeadSeriesRef(ref))

 	if s == nil {
@ -112,14 +115,14 @@ func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.Scra
 	// so we can set the correct markers.
 	if s.ooo.oooHeadChunk != nil {
 		c := s.ooo.oooHeadChunk
-		if c.OverlapsClosedInterval(oh.mint, oh.maxt) && lastMmapRef == 0 {
+		if c.OverlapsClosedInterval(oh.mint, oh.maxt) && maxMmapRef == 0 {
 			ref := chunks.ChunkRef(chunks.NewHeadChunkRef(s.ref, s.oooHeadChunkID(len(s.ooo.oooMmappedChunks))))
 			addChunk(c.minTime, c.maxTime, ref)
 		}
 	}
 	for i := len(s.ooo.oooMmappedChunks) - 1; i >= 0; i-- {
 		c := s.ooo.oooMmappedChunks[i]
-		if c.OverlapsClosedInterval(oh.mint, oh.maxt) && (lastMmapRef == 0 || lastMmapRef.GreaterThanOrEqualTo(c.ref)) {
+		if c.OverlapsClosedInterval(oh.mint, oh.maxt) && (maxMmapRef == 0 || maxMmapRef.GreaterThanOrEqualTo(c.ref)) && (lastGarbageCollectedMmapRef == 0 || c.ref.GreaterThan(lastGarbageCollectedMmapRef)) {
 			ref := chunks.ChunkRef(chunks.NewHeadChunkRef(s.ref, s.oooHeadChunkID(i)))
 			addChunk(c.minTime, c.maxTime, ref)
 		}
@ -232,13 +235,15 @@ func (oh *OOOHeadIndexReader) Postings(ctx context.Context, name string, values
 type OOOHeadChunkReader struct {
 	head       *Head
 	mint, maxt int64
+	isoState   *oooIsolationState
 }

-func NewOOOHeadChunkReader(head *Head, mint, maxt int64) *OOOHeadChunkReader {
+func NewOOOHeadChunkReader(head *Head, mint, maxt int64, isoState *oooIsolationState) *OOOHeadChunkReader {
 	return &OOOHeadChunkReader{
-		head: head,
-		mint: mint,
-		maxt: maxt,
+		head:     head,
+		mint:     mint,
+		maxt:     maxt,
+		isoState: isoState,
 	}
 }

@ -272,6 +277,9 @@ func (cr OOOHeadChunkReader) Chunk(meta chunks.Meta) (chunkenc.Chunk, error) {
 }

 func (cr OOOHeadChunkReader) Close() error {
+	if cr.isoState != nil {
+		cr.isoState.Close()
+	}
 	return nil
 }

@ -306,7 +314,7 @@ func NewOOOCompactionHead(ctx context.Context, head *Head) (*OOOCompactionHead,
 		ch.lastWBLFile = lastWBLFile
 	}

-	ch.oooIR = NewOOOHeadIndexReader(head, math.MinInt64, math.MaxInt64)
+	ch.oooIR = NewOOOHeadIndexReader(head, math.MinInt64, math.MaxInt64, 0)
 	n, v := index.AllPostingsKey()

 	// TODO: verify this gets only ooo samples.
@ -365,7 +373,7 @@ func (ch *OOOCompactionHead) Index() (IndexReader, error) {
 }

 func (ch *OOOCompactionHead) Chunks() (ChunkReader, error) {
-	return NewOOOHeadChunkReader(ch.oooIR.head, ch.oooIR.mint, ch.oooIR.maxt), nil
+	return NewOOOHeadChunkReader(ch.oooIR.head, ch.oooIR.mint, ch.oooIR.maxt, nil), nil
 }

 func (ch *OOOCompactionHead) Tombstones() (tombstones.Reader, error) {
@ -391,7 +399,7 @@ func (ch *OOOCompactionHead) Meta() BlockMeta {
 // Only the method of BlockReader interface are valid for the cloned OOOCompactionHead.
 func (ch *OOOCompactionHead) CloneForTimeRange(mint, maxt int64) *OOOCompactionHead {
 	return &OOOCompactionHead{
-		oooIR:       NewOOOHeadIndexReader(ch.oooIR.head, mint, maxt),
+		oooIR:       NewOOOHeadIndexReader(ch.oooIR.head, mint, maxt, 0),
 		lastMmapRef: ch.lastMmapRef,
 		postings:    ch.postings,
 		chunkRange:  ch.chunkRange,
@ -433,7 +441,7 @@ func (ir *OOOCompactionHeadIndexReader) SortedPostings(p index.Postings) index.P
 }

 func (ir *OOOCompactionHeadIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta) error {
-	return ir.ch.oooIR.series(ref, builder, chks, ir.ch.lastMmapRef)
+	return ir.ch.oooIR.series(ref, builder, chks, 0, ir.ch.lastMmapRef)
 }

 func (ir *OOOCompactionHeadIndexReader) SortedLabelValues(_ context.Context, name string, matchers ...*labels.Matcher) ([]string, error) {
--- a/tsdb/ooo_head_read_test.go
+++ b/tsdb/ooo_head_read_test.go
@ -356,7 +356,7 @@ func TestOOOHeadIndexReader_Series(t *testing.T) {
 						})
 					}

-					ir := NewOOOHeadIndexReader(h, tc.queryMinT, tc.queryMaxT)
+					ir := NewOOOHeadIndexReader(h, tc.queryMinT, tc.queryMaxT, 0)

 					var chks []chunks.Meta
 					var b labels.ScratchBuilder
@ -437,7 +437,7 @@ func TestOOOHeadChunkReader_LabelValues(t *testing.T) {
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
 			// We first want to test using a head index reader that covers the biggest query interval
-			oh := NewOOOHeadIndexReader(head, tc.queryMinT, tc.queryMaxT)
+			oh := NewOOOHeadIndexReader(head, tc.queryMinT, tc.queryMaxT, 0)
 			matchers := []*labels.Matcher{labels.MustNewMatcher(labels.MatchEqual, "foo", "bar1")}
 			values, err := oh.LabelValues(ctx, "foo", matchers...)
 			sort.Strings(values)
@ -484,7 +484,8 @@ func TestOOOHeadChunkReader_Chunk(t *testing.T) {
 	t.Run("Getting a non existing chunk fails with not found error", func(t *testing.T) {
 		db := newTestDBWithOpts(t, opts)

-		cr := NewOOOHeadChunkReader(db.head, 0, 1000)
+		cr := NewOOOHeadChunkReader(db.head, 0, 1000, nil)
+		defer cr.Close()
 		c, err := cr.Chunk(chunks.Meta{
 			Ref: 0x1000000, Chunk: chunkenc.Chunk(nil), MinTime: 100, MaxTime: 300,
 		})
@ -842,14 +843,15 @@ func TestOOOHeadChunkReader_Chunk(t *testing.T) {

 			// The Series method is the one that populates the chunk meta OOO
 			// markers like OOOLastRef. These are then used by the ChunkReader.
-			ir := NewOOOHeadIndexReader(db.head, tc.queryMinT, tc.queryMaxT)
+			ir := NewOOOHeadIndexReader(db.head, tc.queryMinT, tc.queryMaxT, 0)
 			var chks []chunks.Meta
 			var b labels.ScratchBuilder
 			err := ir.Series(s1Ref, &b, &chks)
 			require.NoError(t, err)
 			require.Equal(t, len(tc.expChunksSamples), len(chks))

-			cr := NewOOOHeadChunkReader(db.head, tc.queryMinT, tc.queryMaxT)
+			cr := NewOOOHeadChunkReader(db.head, tc.queryMinT, tc.queryMaxT, nil)
+			defer cr.Close()
 			for i := 0; i < len(chks); i++ {
 				c, err := cr.Chunk(chks[i])
 				require.NoError(t, err)
@ -1005,7 +1007,7 @@ func TestOOOHeadChunkReader_Chunk_ConsistentQueryResponseDespiteOfHeadExpanding(

 			// The Series method is the one that populates the chunk meta OOO
 			// markers like OOOLastRef. These are then used by the ChunkReader.
-			ir := NewOOOHeadIndexReader(db.head, tc.queryMinT, tc.queryMaxT)
+			ir := NewOOOHeadIndexReader(db.head, tc.queryMinT, tc.queryMaxT, 0)
 			var chks []chunks.Meta
 			var b labels.ScratchBuilder
 			err := ir.Series(s1Ref, &b, &chks)
@ -1020,7 +1022,8 @@ func TestOOOHeadChunkReader_Chunk_ConsistentQueryResponseDespiteOfHeadExpanding(
 			}
 			require.NoError(t, app.Commit())

-			cr := NewOOOHeadChunkReader(db.head, tc.queryMinT, tc.queryMaxT)
+			cr := NewOOOHeadChunkReader(db.head, tc.queryMinT, tc.queryMaxT, nil)
+			defer cr.Close()
 			for i := 0; i < len(chks); i++ {
 				c, err := cr.Chunk(chks[i])
 				require.NoError(t, err)
--- a/tsdb/ooo_isolation.go
+++ b/tsdb/ooo_isolation.go
@ -0,0 +1,79 @@
+// Copyright 2023 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 (
+	"container/list"
+	"sync"
+
+	"github.com/prometheus/prometheus/tsdb/chunks"
+)
+
+type oooIsolation struct {
+	mtx       sync.RWMutex
+	openReads *list.List
+}
+
+type oooIsolationState struct {
+	i *oooIsolation
+	e *list.Element
+
+	minRef chunks.ChunkDiskMapperRef
+}
+
+func newOOOIsolation() *oooIsolation {
+	return &oooIsolation{
+		openReads: list.New(),
+	}
+}
+
+// HasOpenReadsAtOrBefore returns true if this oooIsolation is aware of any reads that use
+// chunks with reference at or before ref.
+func (i *oooIsolation) HasOpenReadsAtOrBefore(ref chunks.ChunkDiskMapperRef) bool {
+	i.mtx.RLock()
+	defer i.mtx.RUnlock()
+
+	for e := i.openReads.Front(); e != nil; e = e.Next() {
+		s := e.Value.(*oooIsolationState)
+
+		if ref.GreaterThan(s.minRef) {
+			return true
+		}
+	}
+
+	return false
+}
+
+// TrackReadAfter records a read that uses chunks with reference after minRef.
+//
+// The caller must ensure that the returned oooIsolationState is eventually closed when
+// the read is complete.
+func (i *oooIsolation) TrackReadAfter(minRef chunks.ChunkDiskMapperRef) *oooIsolationState {
+	s := &oooIsolationState{
+		i:      i,
+		minRef: minRef,
+	}
+
+	i.mtx.Lock()
+	s.e = i.openReads.PushBack(s)
+	i.mtx.Unlock()
+
+	return s
+}
+
+func (s oooIsolationState) Close() {
+	s.i.mtx.Lock()
+	s.i.openReads.Remove(s.e)
+	s.i.mtx.Unlock()
+}
--- a/tsdb/ooo_isolation_test.go
+++ b/tsdb/ooo_isolation_test.go
@ -0,0 +1,60 @@
+// Copyright 2023 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 (
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestOOOIsolation(t *testing.T) {
+	i := newOOOIsolation()
+
+	// Empty state shouldn't have any open reads.
+	require.False(t, i.HasOpenReadsAtOrBefore(0))
+	require.False(t, i.HasOpenReadsAtOrBefore(1))
+	require.False(t, i.HasOpenReadsAtOrBefore(2))
+	require.False(t, i.HasOpenReadsAtOrBefore(3))
+
+	// Add a read.
+	read1 := i.TrackReadAfter(1)
+	require.False(t, i.HasOpenReadsAtOrBefore(0))
+	require.False(t, i.HasOpenReadsAtOrBefore(1))
+	require.True(t, i.HasOpenReadsAtOrBefore(2))
+
+	// Add another overlapping read.
+	read2 := i.TrackReadAfter(0)
+	require.False(t, i.HasOpenReadsAtOrBefore(0))
+	require.True(t, i.HasOpenReadsAtOrBefore(1))
+	require.True(t, i.HasOpenReadsAtOrBefore(2))
+
+	// Close the second read, should now only report open reads for the first read's ref.
+	read2.Close()
+	require.False(t, i.HasOpenReadsAtOrBefore(0))
+	require.False(t, i.HasOpenReadsAtOrBefore(1))
+	require.True(t, i.HasOpenReadsAtOrBefore(2))
+
+	// Close the second read again: this should do nothing and ensures we can safely call Close() multiple times.
+	read2.Close()
+	require.False(t, i.HasOpenReadsAtOrBefore(0))
+	require.False(t, i.HasOpenReadsAtOrBefore(1))
+	require.True(t, i.HasOpenReadsAtOrBefore(2))
+
+	// Closing the first read should indicate no further open reads.
+	read1.Close()
+	require.False(t, i.HasOpenReadsAtOrBefore(0))
+	require.False(t, i.HasOpenReadsAtOrBefore(1))
+	require.False(t, i.HasOpenReadsAtOrBefore(2))
+}
--- a/tsdb/querier_test.go
+++ b/tsdb/querier_test.go
@ -2803,7 +2803,7 @@ func BenchmarkQueries(b *testing.B) {

 					qHead, err := NewBlockQuerier(NewRangeHead(head, 1, nSamples), 1, nSamples)
 					require.NoError(b, err)
-					qOOOHead, err := NewBlockQuerier(NewOOORangeHead(head, 1, nSamples), 1, nSamples)
+					qOOOHead, err := NewBlockQuerier(NewOOORangeHead(head, 1, nSamples, 0), 1, nSamples)
 					require.NoError(b, err)

 					queryTypes = append(queryTypes, qt{