tests and refactor

go.mod

@@ -656,7 +656,6 @@ require (
 	go.opentelemetry.io/collector/confmap/provider/httpprovider v0.99.0 // indirect
 	go.opentelemetry.io/collector/confmap/provider/httpsprovider v0.99.0 // indirect
 	go.opentelemetry.io/collector/filter v0.99.0 // indirect
-	go.opentelemetry.io/collector/pdata/testdata v0.99.0 // indirect
 	go.opentelemetry.io/contrib/config v0.5.0 // indirect
 	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.50.0 // indirect
 	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.50.0 // indirect
@@ -715,9 +714,10 @@ replace (
 // * There is a release of Prometheus which contains
 // prometheus/prometheus#13002
 // and prometheus/prometheus#13497
+// and https://github.com/grafana/prometheus/pull/34
 // We use the last v1-related tag as the replace statement does not work for v2
 // tags without the v2 suffix to the module root.
-replace github.com/prometheus/prometheus => github.com/grafana/prometheus v1.8.2-0.20240130142130-51b39f24d406 // cmp_header_order branch
+replace github.com/prometheus/prometheus => github.com/grafana/prometheus v1.8.2-0.20240503114359-53f3689112ff // thampiotr/expose-disabling-staleness-markers
 
 replace gopkg.in/yaml.v2 => github.com/rfratto/go-yaml v0.0.0-20211119180816-77389c3526dc
 

go.sum

@@ -1066,8 +1066,8 @@ github.com/grafana/opentelemetry-collector/service v0.0.0-20240429170914-d1e1018
 github.com/grafana/opentelemetry-collector/service v0.0.0-20240429170914-d1e101852ba5/go.mod h1:0djU5YbUIZw4Y+KNYk07tZztXrMK/LNAZyfH8b7f7xA=
 github.com/grafana/postgres_exporter v0.15.1-0.20240417113938-9358270470dd h1:vNHdecaOmYgSHMEQRgyzWacV++N38Jp8qLZg0RCsfFo=
 github.com/grafana/postgres_exporter v0.15.1-0.20240417113938-9358270470dd/go.mod h1:kR16GJ0ZwWVQ2osW3pgtDJU1a/GXpufrwio0kLG14cg=
-github.com/grafana/prometheus v1.8.2-0.20240130142130-51b39f24d406 h1:LVIOYe5j92m10wluP5hgeHqSkOLnZzcPxhYCkdbLXCE=
-github.com/grafana/prometheus v1.8.2-0.20240130142130-51b39f24d406/go.mod h1:SRw624aMAxTfryAcP8rOjg4S/sHHaetx2lyJJ2nM83g=
+github.com/grafana/prometheus v1.8.2-0.20240503114359-53f3689112ff h1:t81+PfBE+XEserwuBJILA67hOv9g74X4Gs36IzCaRQA=
+github.com/grafana/prometheus v1.8.2-0.20240503114359-53f3689112ff/go.mod h1:SRw624aMAxTfryAcP8rOjg4S/sHHaetx2lyJJ2nM83g=
 github.com/grafana/pyroscope-go/godeltaprof v0.1.7 h1:C11j63y7gymiW8VugJ9ZW0pWfxTZugdSJyC48olk5KY=
 github.com/grafana/pyroscope-go/godeltaprof v0.1.7/go.mod h1:Tk376Nbldo4Cha9RgiU7ik8WKFkNpfds98aUzS8omLE=
 github.com/grafana/pyroscope/api v0.4.0 h1:J86DxoNeLOvtJhB1Cn65JMZkXe682D+RqeoIUiYc/eo=

internal/component/discovery/distributed_targets.go

@@ -1,8 +1,10 @@
 package discovery
 
 import (
-	"github.com/grafana/alloy/internal/service/cluster"
+	"github.com/grafana/ckit/peer"
 	"github.com/grafana/ckit/shard"
+
+	"github.com/grafana/alloy/internal/service/cluster"
 )
 
 // DistributedTargets uses the node's Lookup method to distribute discovery
@@ -17,17 +19,12 @@ type DistributedTargets struct {
 // NewDistributedTargets creates the abstraction that allows components to
 // dynamically shard targets between components.
 func NewDistributedTargets(clusteringEnabled bool, cluster cluster.Cluster, allTargets []Target) *DistributedTargets {
-	// TODO(@tpaschalis): Make this into a single code-path to simplify logic.
 	if !clusteringEnabled || cluster == nil {
-		return &DistributedTargets{
-			localTargets:     allTargets,
-			remoteTargetKeys: map[shard.Key]struct{}{},
-		}
+		cluster = disabledCluster{}
 	}
 
-	peerCount := len(cluster.Peers())
 	localCap := len(allTargets) + 1
-	if peerCount != 0 {
+	if peerCount := len(cluster.Peers()); peerCount != 0 {
 		localCap = (len(allTargets) + 1) / peerCount
 	}
 
@@ -80,3 +77,15 @@ func (dt *DistributedTargets) MovedToRemoteInstance(prev *DistributedTargets) []
 func keyFor(tgt Target) shard.Key {
 	return shard.Key(tgt.NonMetaLabels().Hash())
 }
+
+type disabledCluster struct{}
+
+var _ cluster.Cluster = disabledCluster{}
+
+func (l disabledCluster) Lookup(key shard.Key, replicationFactor int, op shard.Op) ([]peer.Peer, error) {
+	return nil, nil
+}
+
+func (l disabledCluster) Peers() []peer.Peer {
+	return nil
+}

internal/component/discovery/distributed_targets_test.go

@@ -2,12 +2,14 @@ package discovery
 
 import (
 	"fmt"
-	"github.com/grafana/alloy/internal/service/cluster"
+	"math/rand"
+	"testing"
+
 	"github.com/grafana/ckit/peer"
 	"github.com/grafana/ckit/shard"
 	"github.com/stretchr/testify/require"
-	"math/rand"
-	"testing"
+
+	"github.com/grafana/alloy/internal/service/cluster"
 )
 
 var (
@@ -126,13 +128,13 @@ var movedToRemoteInstanceTestCases = []struct {
 	expectedMovedTargets []Target
 }{
 	{
-		name: "no previous targets distribution",
+		name:     "no previous targets distribution",
+		previous: nil,
 		current: testDistTargets(map[shard.Key][]peer.Peer{
 			keyFor(target1): {peer1Self},
 			keyFor(target2): {peer2},
 			keyFor(target3): {peer1Self},
 		}),
-		previous:             nil,
 		expectedMovedTargets: nil,
 	},
 	{
@@ -290,18 +292,18 @@ func mkTarget(kv ...string) Target {
 	return target
 }
 
-type fakeCluster struct {
-	lookupMap map[shard.Key][]peer.Peer
-	peers     []peer.Peer
-}
-
 func testDistTargets(lookupMap map[shard.Key][]peer.Peer) *DistributedTargets {
 	return NewDistributedTargets(true, &fakeCluster{
 		peers:     allTestPeers,
 		lookupMap: lookupMap,
 	}, allTestTargets)
 }
 
+type fakeCluster struct {
+	lookupMap map[shard.Key][]peer.Peer
+	peers     []peer.Peer
+}
+
 func (f *fakeCluster) Lookup(key shard.Key, _ int, _ shard.Op) ([]peer.Peer, error) {
 	if key == magicErrorKey {
 		return nil, fmt.Errorf("test error for magic error key")

internal/component/prometheus/scrape/scrape.go

@@ -8,6 +8,15 @@ import (
 	"time"
 
 	"github.com/alecthomas/units"
+	client_prometheus "github.com/prometheus/client_golang/prometheus"
+	config_util "github.com/prometheus/common/config"
+	"github.com/prometheus/common/model"
+	"github.com/prometheus/prometheus/config"
+	"github.com/prometheus/prometheus/discovery/targetgroup"
+	"github.com/prometheus/prometheus/model/labels"
+	"github.com/prometheus/prometheus/scrape"
+	"github.com/prometheus/prometheus/storage"
+
 	"github.com/grafana/alloy/internal/alloy/logging/level"
 	"github.com/grafana/alloy/internal/component"
 	component_config "github.com/grafana/alloy/internal/component/common/config"
@@ -18,14 +27,6 @@ import (
 	"github.com/grafana/alloy/internal/service/http"
 	"github.com/grafana/alloy/internal/service/labelstore"
 	"github.com/grafana/alloy/internal/useragent"
-	client_prometheus "github.com/prometheus/client_golang/prometheus"
-	config_util "github.com/prometheus/common/config"
-	"github.com/prometheus/common/model"
-	"github.com/prometheus/prometheus/config"
-	"github.com/prometheus/prometheus/discovery/targetgroup"
-	"github.com/prometheus/prometheus/model/labels"
-	"github.com/prometheus/prometheus/scrape"
-	"github.com/prometheus/prometheus/storage"
 )
 
 func init() {
@@ -239,7 +240,10 @@ func (c *Component) Run(ctx context.Context) error {
 
 			newTargetGroups, movedTargets := c.distributeTargets(targets, jobName, args)
 
-			// Make sure the targets that moved to another instance are not marked as stale.
+			// Make sure the targets that moved to another instance are NOT marked as stale. This is specific to how
+			// Prometheus handles marking series as stale: it is the client's responsibility to inject the
+			// staleness markers. In our case, for targets that moved to another instance in the cluster, we hand
+			// over this responsibility to the new owning instance. We must not inject staleness marker here.
 			c.scraper.DisableEndOfRunStalenessMarkers(jobName, movedTargets)
 
 			select {