Merge pull request #1773 from influxdb/race-condition-when-merging-multiple-series

Fix problem with merging series that have unequal number of points in gr...

CHANGELOG.md

@@ -5,6 +5,7 @@
 - [#1752](https://github.com/influxdb/influxdb/pull/1752): remove debug log output from collectd.
 - [#1720](https://github.com/influxdb/influxdb/pull/1720): Parse Series IDs as unsigned 32-bits.
 - [#1767](https://github.com/influxdb/influxdb/pull/1767): Drop Series was failing across shards.  Issue #1761.
+- [#1773](https://github.com/influxdb/influxdb/pull/1773): Fix bug when merging series together that have unequal number of points in a group by interval
 
 ### Features
 

influxql/engine.go

@@ -69,7 +69,7 @@ func NewPlanner(db DB) *Planner {
 
 // Plan creates an execution plan for the given SelectStatement and returns an Executor.
 func (p *Planner) Plan(stmt *SelectStatement) (*Executor, error) {
-	now := p.Now()
+	now := p.Now().UTC()
 
 	// Clone the statement to be planned.
 	// Replace instances of "now()" with the current time.
@@ -698,7 +698,9 @@ func MapMean(itr Iterator, e *Emitter, tmin int64) {
 		out.Count++
 		out.Sum += v.(float64)
 	}
-	e.Emit(Key{tmin, itr.Tags()}, out)
+	if out.Count > 0 {
+		e.Emit(Key{tmin, itr.Tags()}, out)
+	}
 }
 
 type meanMapOutput struct {
@@ -714,7 +716,9 @@ func ReduceMean(key Key, values []interface{}, e *Emitter) {
 		out.Count += val.Count
 		out.Sum += val.Sum
 	}
-	e.Emit(key, out.Sum/float64(out.Count))
+	if out.Count > 0 {
+		e.Emit(key, out.Sum/float64(out.Count))
+	}
 }
 
 // MapMin collects the values to pass to the reducer

server_test.go

@@ -1138,6 +1138,37 @@ func TestServer_DropSeriesFromMeasurement(t *testing.T) {
 	}
 }
 
+// Ensure that when merging many series together and some of them have a different number of points than others
+// in a group by interval the results are correct
+func TestServer_MergeManySeries(t *testing.T) {
+	c := NewMessagingClient()
+	s := OpenServer(c)
+	defer s.Close()
+	s.CreateDatabase("foo")
+	s.CreateRetentionPolicy("foo", &influxdb.RetentionPolicy{Name: "raw", Duration: 1 * time.Hour})
+	s.SetDefaultRetentionPolicy("foo", "raw")
+
+	for i := 1; i < 11; i++ {
+		for j := 1; j < 5+i%3; j++ {
+			tags := map[string]string{"host": fmt.Sprintf("server_%d", i)}
+			if index, err := s.WriteSeries("foo", "raw", []influxdb.Point{{Name: "cpu", Tags: tags, Timestamp: time.Unix(int64(j), int64(0)), Fields: map[string]interface{}{"value": float64(22)}}}); err != nil {
+				t.Fatalf("unexpected error: %s", err.Error())
+			} else if err = s.Sync(index); err != nil {
+				t.Fatalf("sync error: %s", err)
+			}
+		}
+	}
+
+	results := s.ExecuteQuery(MustParseQuery(`select count(value) from cpu group by time(1s)`), "foo", nil)
+	if res := results.Results[0]; res.Err != nil {
+		t.Fatalf("unexpected error: %s", res.Err)
+	} else if len(res.Series) != 1 {
+		t.Fatalf("unexpected row count: %d", len(res.Series))
+	} else if s := mustMarshalJSON(res); s != `{"series":[{"name":"cpu","columns":["time","count"],"values":[["1970-01-01T00:00:01Z",10],["1970-01-01T00:00:02Z",10],["1970-01-01T00:00:03Z",10],["1970-01-01T00:00:04Z",10],["1970-01-01T00:00:05Z",7],["1970-01-01T00:00:06Z",3]]}]}` {
+		t.Fatalf("unexpected row(0): %s", s)
+	}
+}
+
 // Ensure Drop Series can:
 // write to measurement cpu with tags region=uswest host=serverA
 // write to measurement cpu with tags region=uswest host=serverB

tx.go

@@ -2,6 +2,7 @@ package influxdb
 
 import (
 	"fmt"
+	"math"
 	"sort"
 	"sync"
 	"time"
@@ -309,10 +310,12 @@ func (i *shardIterator) Tags() string { return i.tags }
 
 func (i *shardIterator) Next() (key int64, data []byte, value interface{}) {
 	min := -1
+	minKey := int64(math.MaxInt64)
 
 	for ind, kv := range i.keyValues {
-		if kv.key != 0 && kv.key < i.tmax {
+		if kv.key != 0 && kv.key < i.tmax && kv.key < minKey {
 			min = ind
+			minKey = kv.key
 		}
 	}