fix Makefile proto gen conditions

Makefile

@@ -324,11 +324,11 @@ GENERATE_FILES ?= true
 
 %.pb.go: %.proto
 ifeq ($(GENERATE_FILES),true)
-	if [ "$@" != "*_custom_types.proto" ]; then \
+	if case "$@" in *_custom_types.pb.go) false ;; *) true ;; esac; then \
 		protoc -I $(GOPATH)/src:./vendor/github.com/gogo/protobuf:./vendor:./$(@D):./pkg/storegateway/storepb --gogoslick_out=plugins=grpc,Mgoogle/protobuf/any.proto=github.com/gogo/protobuf/types,:./$(@D) ./$(patsubst %.pb.go,%.proto,$@); \
-    else \
+	else \
 		echo "Skipping $@"; \
-    fi
+	fi
 else
 	@echo "Warning: generating files has been disabled, but the following file needs to be regenerated: $@"
 	@echo "If this is unexpected, check if the last modified timestamps on $@ and $(patsubst %.pb.go,%.proto,$@) are correct."

pkg/mimirpb/mimir.proto

@@ -13,7 +13,7 @@ package cortexpb;
 option go_package = "mimirpb";
 
 import "github.com/gogo/protobuf/gogoproto/gogo.proto";
-import "github.com/grafana/mimir/pkg/mimirpb/mimir_custom_types.proto";
+import "mimir_custom_types.proto";
 
 option (gogoproto.marshaler_all) = true;
 option (gogoproto.unmarshaler_all) = true;
@@ -66,11 +66,6 @@ message TimeSeries {
   repeated Histogram histograms = 4 [(gogoproto.nullable) = false];
 }
 
-message LabelPair {
-  bytes name  = 1;
-  bytes value = 2;
-}
-
 message Sample {
 	// Fields order MUST match promql.FPoint so that we can cast types between them.
   int64 timestamp_ms = 2;

pkg/scheduler/queue/baseline.txt

@@ -0,0 +1,24 @@
+goos: linux
+goarch: amd64
+pkg: github.com/grafana/mimir/pkg/scheduler/queue
+cpu: AMD Ryzen 9 PRO 6950H with Radeon Graphics     
+BenchmarkConcurrentQueueOperations/1_tenants/10_concurrent_producers/16_concurrent_consumers-16         	 3375270	      3582 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/10_concurrent_producers/160_concurrent_consumers-16        	 3158005	      3817 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/10_concurrent_producers/1600_concurrent_consumers-16       	 2469280	      4872 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/25_concurrent_producers/16_concurrent_consumers-16         	 3133611	      3877 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/25_concurrent_producers/160_concurrent_consumers-16        	 2882362	      4214 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/25_concurrent_producers/1600_concurrent_consumers-16       	 2554503	      4684 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/10_concurrent_producers/16_concurrent_consumers-16        	 3180195	      3682 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/10_concurrent_producers/160_concurrent_consumers-16       	 3052290	      3901 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/10_concurrent_producers/1600_concurrent_consumers-16      	 2420862	      4755 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/25_concurrent_producers/16_concurrent_consumers-16        	 3111976	      3767 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/25_concurrent_producers/160_concurrent_consumers-16       	 3040722	      3943 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/25_concurrent_producers/1600_concurrent_consumers-16      	 2566794	      4633 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/10_concurrent_producers/16_concurrent_consumers-16      	 3222010	      3926 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/10_concurrent_producers/160_concurrent_consumers-16     	 2796484	      4410 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/10_concurrent_producers/1600_concurrent_consumers-16    	 2399295	      4669 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/25_concurrent_producers/16_concurrent_consumers-16      	 3091362	      3847 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/25_concurrent_producers/160_concurrent_consumers-16     	 2992759	      3959 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/25_concurrent_producers/1600_concurrent_consumers-16    	 2485108	      4807 ns/op
+PASS
+ok  	github.com/grafana/mimir/pkg/scheduler/queue	292.419s

pkg/scheduler/queue/break.txt

@@ -0,0 +1,24 @@
+goos: linux
+goarch: amd64
+pkg: github.com/grafana/mimir/pkg/scheduler/queue
+cpu: AMD Ryzen 9 PRO 6950H with Radeon Graphics     
+BenchmarkConcurrentQueueOperations/1_tenants/10_concurrent_producers/16_concurrent_consumers-16         	 3375884	      3474 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/10_concurrent_producers/160_concurrent_consumers-16        	 3243482	      3666 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/10_concurrent_producers/1600_concurrent_consumers-16       	 2816721	      4222 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/25_concurrent_producers/16_concurrent_consumers-16         	 3460167	      3495 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/25_concurrent_producers/160_concurrent_consumers-16        	 3213451	      3730 ns/op
+BenchmarkConcurrentQueueOperations/1_tenants/25_concurrent_producers/1600_concurrent_consumers-16       	 2842573	      4287 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/10_concurrent_producers/16_concurrent_consumers-16        	 3229412	      3664 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/10_concurrent_producers/160_concurrent_consumers-16       	 3111933	      3913 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/10_concurrent_producers/1600_concurrent_consumers-16      	 2624695	      4583 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/25_concurrent_producers/16_concurrent_consumers-16        	 3245578	      3680 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/25_concurrent_producers/160_concurrent_consumers-16       	 3026472	      3864 ns/op
+BenchmarkConcurrentQueueOperations/10_tenants/25_concurrent_producers/1600_concurrent_consumers-16      	 2503609	      4762 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/10_concurrent_producers/16_concurrent_consumers-16      	 3292596	      3629 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/10_concurrent_producers/160_concurrent_consumers-16     	 3137596	      3817 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/10_concurrent_producers/1600_concurrent_consumers-16    	 2726665	      4568 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/25_concurrent_producers/16_concurrent_consumers-16      	 3080842	      3869 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/25_concurrent_producers/160_concurrent_consumers-16     	 2937258	      3981 ns/op
+BenchmarkConcurrentQueueOperations/1000_tenants/25_concurrent_producers/1600_concurrent_consumers-16    	 2483106	      4968 ns/op
+PASS
+ok  	github.com/grafana/mimir/pkg/scheduler/queue	289.483s

pkg/scheduler/queue/tree/tenant_deletion_benchmark_test.go

@@ -0,0 +1,234 @@
+package tree
+
+import (
+	"fmt"
+	"math/rand"
+	"slices"
+	"strconv"
+	"testing"
+)
+
+type fullTenantDeleteFunc func(tenantNodes map[string][]*Node, tenantIDOrder []string, tenant string) (map[string][]*Node, []string)
+
+func baseline(tenantNodes map[string][]*Node, tenantIDOrder []string, tenant string) (map[string][]*Node, []string) {
+	// delete from shared tenantNodes
+	for i := range tenantNodes[tenant] {
+		// only ever going to have a slice of length one here for simplicity
+		// and best representation of the actual case we are benchmarking;
+		// no need to check if this is the dequeuedFrom node
+		//if tenantNode == dequeuedFrom {
+		tenantNodes[tenant] = append(tenantNodes[tenant][:i], tenantNodes[tenant][i+1:]...)
+		//}
+	}
+
+	for idx, name := range tenantIDOrder {
+		if name == tenant {
+			tenantIDOrder[idx] = string(emptyTenantID)
+		}
+	}
+
+	// clear all sequential empty elements from tenantIDOrder
+	lastElementIndex := len(tenantIDOrder) - 1
+	for i := lastElementIndex; i >= 0 && tenantIDOrder[i] == ""; i-- {
+		tenantIDOrder = tenantIDOrder[:i]
+	}
+	return tenantNodes, tenantIDOrder
+}
+
+func breakEarly(tenantNodes map[string][]*Node, tenantIDOrder []string, tenant string) (map[string][]*Node, []string) {
+	// delete from shared tenantNodes
+	for i := range tenantNodes[tenant] {
+		// only ever going to have a slice of length one here for simplicity
+		// and best representation of the actual case we are benchmarking;
+		// no need to check if this is the dequeuedFrom node
+		//if tenantNode == dequeuedFrom {
+		tenantNodes[tenant] = append(tenantNodes[tenant][:i], tenantNodes[tenant][i+1:]...)
+		//}
+	}
+
+	for idx, name := range tenantIDOrder {
+		if name == tenant {
+			tenantIDOrder[idx] = string(emptyTenantID)
+			break
+		}
+	}
+
+	// clear all sequential empty elements from tenantIDOrder
+	lastElementIndex := len(tenantIDOrder) - 1
+	for i := lastElementIndex; i >= 0 && tenantIDOrder[i] == ""; i-- {
+		tenantIDOrder = tenantIDOrder[:i]
+	}
+	return tenantNodes, tenantIDOrder
+}
+
+func breakEarlyShrinkOnce(tenantNodes map[string][]*Node, tenantIDOrder []string, tenant string) (map[string][]*Node, []string) {
+	// delete from shared tenantNodes
+	for i := range tenantNodes[tenant] {
+		// only ever going to have a slice of length one here for simplicity
+		// and best representation of the actual case we are benchmarking;
+		// no need to check if this is the dequeuedFrom node
+		//if tenantNode == dequeuedFrom {
+		tenantNodes[tenant] = slices.Delete(tenantNodes[tenant], i, i+1)
+		//}
+	}
+
+	for idx, name := range tenantIDOrder {
+		if name == tenant {
+			tenantIDOrder[idx] = string(emptyTenantID)
+			break
+		}
+	}
+
+	emptyTenantIDsAtEnd := 0
+	for i := len(tenantIDOrder) - 1; i >= 0 && tenantIDOrder[i] == ""; i-- {
+		emptyTenantIDsAtEnd++
+	}
+	tenantIDOrder = slices.Delete(
+		tenantIDOrder,
+		len(tenantIDOrder)-emptyTenantIDsAtEnd,
+		len(tenantIDOrder),
+	)
+
+	return tenantNodes, tenantIDOrder
+}
+
+func breakEarlyShrinkOnceNoSlices(tenantNodes map[string][]*Node, tenantIDOrder []string, tenant string) (map[string][]*Node, []string) {
+	// delete from shared tenantNodes
+	for i := range tenantNodes[tenant] {
+		// only ever going to have a slice of length one here for simplicity
+		// and best representation of the actual case we are benchmarking;
+		// no need to check if this is the dequeuedFrom node
+		//if tenantNode == dequeuedFrom {
+		tenantNodes[tenant] = append(tenantNodes[tenant][:i], tenantNodes[tenant][i+1:]...)
+		//}
+	}
+
+	for idx, name := range tenantIDOrder {
+		if name == tenant {
+			tenantIDOrder[idx] = string(emptyTenantID)
+			break
+		}
+	}
+
+	emptyTenantIDsAtEnd := 0
+	for i := len(tenantIDOrder) - 1; i >= 0 && tenantIDOrder[i] == ""; i-- {
+		emptyTenantIDsAtEnd++
+	}
+	tenantIDOrder = tenantIDOrder[:len(tenantIDOrder)-emptyTenantIDsAtEnd]
+
+	return nil, tenantIDOrder
+}
+
+func BenchmarkFullTenantDeletion(b *testing.B) {
+	tenantCount := 10000
+
+	var testCases = []struct {
+		name                string
+		deleteFunc          fullTenantDeleteFunc
+		tenantNodes         map[string][]*Node
+		tenantRotationOrder []string
+		tenantDeletionOrder []string
+	}{
+		{"baseline", baseline, nil, nil, nil},
+		{"breakEarlyShrinkOnce", breakEarlyShrinkOnce, nil, nil, nil},
+		{"breakEarlyShrinkOnceNoSlices", breakEarlyShrinkOnceNoSlices, nil, nil, nil},
+	}
+
+	makeStubs := func() (tenantNodes map[string][]*Node, tenantRotationOrder, tenantDeletionOrder []string) {
+		tenantNodes = make(map[string][]*Node, tenantCount)
+		tenantRotationOrder = make([]string, tenantCount)
+
+		for i := range tenantCount {
+			tenantRotationOrder[i] = strconv.Itoa(i)
+			tenantNodes[strconv.Itoa(i)] = append(tenantNodes[strconv.Itoa(i)], &Node{})
+		}
+
+		tenantDeletionOrder = make([]string, tenantCount)
+		copy(tenantDeletionOrder, tenantRotationOrder)
+		rand.Shuffle(len(tenantDeletionOrder), func(i, j int) {
+			tenantDeletionOrder[i], tenantDeletionOrder[j] = tenantDeletionOrder[j], tenantDeletionOrder[i]
+		})
+
+		return tenantNodes, tenantRotationOrder, tenantDeletionOrder
+	}
+
+	for _, testCase := range testCases {
+		b.Run(fmt.Sprintf("delete_tenant_func_%s", testCase.name), func(b *testing.B) {
+			for i := 0; i < b.N; i++ {
+				tenantNodes, tenantRotationOrder, tenantDeletionOrder := makeStubs()
+				testCase.tenantNodes = tenantNodes
+				testCase.tenantRotationOrder = tenantRotationOrder
+				testCase.tenantDeletionOrder = tenantDeletionOrder
+
+				tenantDeleteIdx := 0
+				for len(testCase.tenantRotationOrder) > 0 {
+					//fmt.Println("len(testCase.tenantRotationOrder): ", len(testCase.tenantRotationOrder))
+					testCase.tenantNodes, testCase.tenantRotationOrder = testCase.deleteFunc(
+						testCase.tenantNodes, testCase.tenantRotationOrder, testCase.tenantDeletionOrder[tenantDeleteIdx],
+					)
+					tenantDeleteIdx++
+				}
+			}
+		})
+	}
+}
+
+type makeQueuePathFunc func(component, tenant string) QueuePath
+
+func baselineMakeQueuePath(component, tenant string) QueuePath {
+	return append([]string{component}, tenant)
+}
+
+func noAppendMakeQueuePath(component, tenant string) QueuePath {
+	return QueuePath{component, tenant}
+}
+
+func BenchmarkMakeQueuePath(b *testing.B) {
+	tenantCount := 10000
+
+	tenants := make([]string, tenantCount)
+	queueComponents := make([]string, tenantCount)
+	for i := range tenantCount {
+		tenants[i] = strconv.Itoa(i)
+		queueComponents[i] = randAdditionalQueueDimension()
+	}
+
+	var testCases = []struct {
+		name     string
+		pathFunc makeQueuePathFunc
+	}{
+		{"baselineMakeQueuePath", baselineMakeQueuePath},
+		{"noAppendMakeQueuePath", noAppendMakeQueuePath},
+	}
+
+	for _, testCase := range testCases {
+		b.Run(fmt.Sprintf("queue_path_func_%s", testCase.name), func(b *testing.B) {
+			for i := 0; i < b.N; i++ {
+				for tenantIdx := range tenantCount {
+					_ = testCase.pathFunc(
+						queueComponents[tenantIdx], tenants[tenantIdx],
+					)
+				}
+			}
+		})
+	}
+}
+
+const ingesterQueueDimension = "ingester"
+const storeGatewayQueueDimension = "store-gateway"
+const ingesterAndStoreGatewayQueueDimension = "ingester-and-store-gateway"
+const unknownQueueDimension = "unknown"
+
+var secondQueueDimensionOptions = []string{
+	ingesterQueueDimension,
+	storeGatewayQueueDimension,
+	ingesterAndStoreGatewayQueueDimension,
+	unknownQueueDimension,
+}
+
+func randAdditionalQueueDimension() string {
+	maxIdx := len(secondQueueDimensionOptions) - 1
+
+	idx := rand.Intn(maxIdx)
+	return secondQueueDimensionOptions[idx]
+}

pkg/scheduler/queue/vs.txt

@@ -0,0 +1,27 @@
+goos: linux
+goarch: amd64
+pkg: github.com/grafana/mimir/pkg/scheduler/queue
+cpu: AMD Ryzen 9 PRO 6950H with Radeon Graphics     
+                                                                                            │ baseline.txt │              break.txt               │
+                                                                                            │    sec/op    │    sec/op     vs base                │
+ConcurrentQueueOperations/1_tenants/10_concurrent_producers/16_concurrent_consumers-16        3.582µ ± ∞ ¹   3.474µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1_tenants/10_concurrent_producers/160_concurrent_consumers-16       3.817µ ± ∞ ¹   3.666µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1_tenants/10_concurrent_producers/1600_concurrent_consumers-16      4.872µ ± ∞ ¹   4.222µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1_tenants/25_concurrent_producers/16_concurrent_consumers-16        3.877µ ± ∞ ¹   3.495µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1_tenants/25_concurrent_producers/160_concurrent_consumers-16       4.214µ ± ∞ ¹   3.730µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1_tenants/25_concurrent_producers/1600_concurrent_consumers-16      4.684µ ± ∞ ¹   4.287µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/10_tenants/10_concurrent_producers/16_concurrent_consumers-16       3.682µ ± ∞ ¹   3.664µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/10_tenants/10_concurrent_producers/160_concurrent_consumers-16      3.901µ ± ∞ ¹   3.913µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/10_tenants/10_concurrent_producers/1600_concurrent_consumers-16     4.755µ ± ∞ ¹   4.583µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/10_tenants/25_concurrent_producers/16_concurrent_consumers-16       3.767µ ± ∞ ¹   3.680µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/10_tenants/25_concurrent_producers/160_concurrent_consumers-16      3.943µ ± ∞ ¹   3.864µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/10_tenants/25_concurrent_producers/1600_concurrent_consumers-16     4.633µ ± ∞ ¹   4.762µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1000_tenants/10_concurrent_producers/16_concurrent_consumers-16     3.926µ ± ∞ ¹   3.629µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1000_tenants/10_concurrent_producers/160_concurrent_consumers-16    4.410µ ± ∞ ¹   3.817µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1000_tenants/10_concurrent_producers/1600_concurrent_consumers-16   4.669µ ± ∞ ¹   4.568µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1000_tenants/25_concurrent_producers/16_concurrent_consumers-16     3.847µ ± ∞ ¹   3.869µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1000_tenants/25_concurrent_producers/160_concurrent_consumers-16    3.959µ ± ∞ ¹   3.981µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+ConcurrentQueueOperations/1000_tenants/25_concurrent_producers/1600_concurrent_consumers-16   4.807µ ± ∞ ¹   4.968µ ± ∞ ¹       ~ (p=1.000 n=1) ²
+geomean                                                                                       4.164µ         3.987µ        -4.26%
+¹ need >= 6 samples for confidence interval at level 0.95
+² need >= 4 samples to detect a difference at alpha level 0.05