Fix bug where querier may not be able to achieve max-concurrent (cortexproject#4417)

* Fix bug where querier may not be able to achieve max-concurrent

Signed-off-by: Alvin Lin <[email protected]>

* Update change log

Signed-off-by: Alvin Lin <[email protected]>

* Update CHANGELOG.md

Co-authored-by: Bryan Boreham <[email protected]>
Signed-off-by: Alvin Lin <[email protected]>

* Update pkg/querier/worker/worker_test.go

Co-authored-by: Arve Knudsen <[email protected]>
Signed-off-by: Alvin Lin <[email protected]>

* Address PR comments

Signed-off-by: Alvin Lin <[email protected]>

Co-authored-by: Bryan Boreham <[email protected]>
Co-authored-by: Arve Knudsen <[email protected]>
Signed-off-by: Alvin Lin <[email protected]>

Author: 3 people

CHANGELOG.md

@@ -38,8 +38,12 @@
 * [BUGFIX] Ruler: fixed counting of PromQL evaluation errors as user-errors when updating `cortex_ruler_queries_failed_total`. #4335
 * [BUGFIX] Ingester: When using block storage, prevent any reads or writes while the ingester is stopping. This will prevent accessing TSDB blocks once they have been already closed. #4304
 * [BUGFIX] Ingester: fixed ingester stuck on start up (LEAVING ring state) when `-ingester.heartbeat-period=0` and `-ingester.unregister-on-shutdown=false`. #4366
+<<<<<<< HEAD
 =======
 >>>>>>> Update for 1.10.0-rc.0 release candidate
+=======
+* [BUGFIX] Querier: After query-frontend restart, querier may have lower than configured concurrency. #4417
+>>>>>>> Fix bug where querier may not be able to achieve max-concurrent (#4417)
 
 ## 1.10.0 / 2021-08-03
 

pkg/querier/worker/worker.go

@@ -208,6 +208,8 @@ func (w *querierWorker) AddressRemoved(address string) {
 	w.mu.Lock()
 	p := w.managers[address]
 	delete(w.managers, address)
+	// Called with lock.
+	w.resetConcurrency()
 	w.mu.Unlock()
 
 	if p != nil {

pkg/querier/worker/worker_test.go

@@ -3,6 +3,7 @@ package worker
 import (
 	"context"
 	"fmt"
+	"math/rand"
 	"testing"
 	"time"
 
@@ -17,46 +18,52 @@ import (
 
 func TestResetConcurrency(t *testing.T) {
 	tests := []struct {
-		name                string
-		parallelism         int
-		maxConcurrent       int
-		numTargets          int
-		expectedConcurrency int
+		name                                  string
+		parallelism                           int
+		maxConcurrent                         int
+		numTargets                            int
+		expectedConcurrency                   int
+		expectedConcurrencyAfterTargetRemoval int
 	}{
 		{
-			name:                "Test create at least one processor per target",
-			parallelism:         0,
-			maxConcurrent:       0,
-			numTargets:          2,
-			expectedConcurrency: 2,
+			name:                                  "Test create at least one processor per target",
+			parallelism:                           0,
+			maxConcurrent:                         0,
+			numTargets:                            2,
+			expectedConcurrency:                   2,
+			expectedConcurrencyAfterTargetRemoval: 1,
 		},
 		{
-			name:                "Test parallelism per target",
-			parallelism:         4,
-			maxConcurrent:       0,
-			numTargets:          2,
-			expectedConcurrency: 8,
+			name:                                  "Test parallelism per target",
+			parallelism:                           4,
+			maxConcurrent:                         0,
+			numTargets:                            2,
+			expectedConcurrency:                   8,
+			expectedConcurrencyAfterTargetRemoval: 4,
 		},
 		{
-			name:                "Test Total Parallelism with a remainder",
-			parallelism:         1,
-			maxConcurrent:       7,
-			numTargets:          4,
-			expectedConcurrency: 7,
+			name:                                  "Test Total Parallelism with a remainder",
+			parallelism:                           1,
+			maxConcurrent:                         7,
+			numTargets:                            4,
+			expectedConcurrency:                   7,
+			expectedConcurrencyAfterTargetRemoval: 7,
 		},
 		{
-			name:                "Test Total Parallelism dividing evenly",
-			parallelism:         1,
-			maxConcurrent:       6,
-			numTargets:          2,
-			expectedConcurrency: 6,
+			name:                                  "Test Total Parallelism dividing evenly",
+			parallelism:                           1,
+			maxConcurrent:                         6,
+			numTargets:                            2,
+			expectedConcurrency:                   6,
+			expectedConcurrencyAfterTargetRemoval: 6,
 		},
 		{
-			name:                "Test Total Parallelism at least one worker per target",
-			parallelism:         1,
-			maxConcurrent:       3,
-			numTargets:          6,
-			expectedConcurrency: 6,
+			name:                                  "Test Total Parallelism at least one worker per target",
+			parallelism:                           1,
+			maxConcurrent:                         3,
+			numTargets:                            6,
+			expectedConcurrency:                   6,
+			expectedConcurrencyAfterTargetRemoval: 5,
 		},
 	}
 
@@ -82,6 +89,12 @@ func TestResetConcurrency(t *testing.T) {
 				return getConcurrentProcessors(w)
 			})
 
+			// now we remove an address and ensure we still have the expected concurrency
+			w.AddressRemoved(fmt.Sprintf("127.0.0.1:%d", rand.Intn(tt.numTargets)))
+			test.Poll(t, 250*time.Millisecond, tt.expectedConcurrencyAfterTargetRemoval, func() interface{} {
+				return getConcurrentProcessors(w)
+			})
+
 			require.NoError(t, services.StopAndAwaitTerminated(context.Background(), w))
 			assert.Equal(t, 0, getConcurrentProcessors(w))
 		})