Add number tenants per query histogram

Signed-off-by: SungJin1212 <[email protected]>

CHANGELOG.md

@@ -18,6 +18,7 @@
 * [FEATURE] Chunk Cache: Support multi level cache and add metrics. #6249
 * [FEATURE] Distributor: Accept multiple HA Tracker pairs in the same request. #6256
 * [FEATURE] Ruler: Add support for per-user external labels #6340
+* [ENHANCEMENT] Querier: Add a `cortex_querier_federated_tenants_per_query` histogram to track the number of tenants per query when the `-tenant-federation.enabled=true`. #6447
 * [ENHANCEMENT] Query Frontend: Add a number of series in the query response to the query stat log. #6423
 * [ENHANCEMENT] Store Gateway: Add a hedged request to reduce the tail latency. #6388
 * [ENHANCEMENT] Ingester: Add metrics to track succeed/failed native histograms. #6370

pkg/cortex/modules.go

@@ -268,12 +268,13 @@ func (t *Cortex) initQueryable() (serv services.Service, err error) {
 
 // Enable merge querier if multi tenant query federation is enabled
 func (t *Cortex) initTenantFederation() (serv services.Service, err error) {
+
 	if t.Cfg.TenantFederation.Enabled {
 		// Make sure the mergeQuerier is only used for request with more than a
 		// single tenant. This allows for a less impactful enabling of tenant
 		// federation.
 		byPassForSingleQuerier := true
-		t.QuerierQueryable = querier.NewSampleAndChunkQueryable(tenantfederation.NewQueryable(t.QuerierQueryable, byPassForSingleQuerier))
+		t.QuerierQueryable = querier.NewSampleAndChunkQueryable(tenantfederation.NewQueryable(t.QuerierQueryable, byPassForSingleQuerier, prometheus.DefaultRegisterer))
 	}
 	return nil, nil
 }

pkg/querier/tenantfederation/merge_queryable.go

@@ -7,6 +7,8 @@ import (
 	"strings"
 
 	"github.com/pkg/errors"
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
 	"github.com/prometheus/prometheus/model/labels"
 	"github.com/prometheus/prometheus/storage"
 	"github.com/prometheus/prometheus/tsdb/chunkenc"
@@ -36,8 +38,8 @@ const (
 // If the label "__tenant_id__" is already existing, its value is overwritten
 // by the tenant ID and the previous value is exposed through a new label
 // prefixed with "original_". This behaviour is not implemented recursively.
-func NewQueryable(upstream storage.Queryable, byPassWithSingleQuerier bool) storage.Queryable {
-	return NewMergeQueryable(defaultTenantLabel, tenantQuerierCallback(upstream), byPassWithSingleQuerier)
+func NewQueryable(upstream storage.Queryable, byPassWithSingleQuerier bool, reg prometheus.Registerer) storage.Queryable {
+	return NewMergeQueryable(defaultTenantLabel, tenantQuerierCallback(upstream), byPassWithSingleQuerier, reg)
 }
 
 func tenantQuerierCallback(queryable storage.Queryable) MergeQuerierCallback {
@@ -79,18 +81,26 @@ type MergeQuerierCallback func(ctx context.Context, mint int64, maxt int64) (ids
 // If the label `idLabelName` is already existing, its value is overwritten and
 // the previous value is exposed through a new label prefixed with "original_".
 // This behaviour is not implemented recursively.
-func NewMergeQueryable(idLabelName string, callback MergeQuerierCallback, byPassWithSingleQuerier bool) storage.Queryable {
+func NewMergeQueryable(idLabelName string, callback MergeQuerierCallback, byPassWithSingleQuerier bool, reg prometheus.Registerer) storage.Queryable {
 	return &mergeQueryable{
 		idLabelName:             idLabelName,
 		callback:                callback,
 		byPassWithSingleQuerier: byPassWithSingleQuerier,
+
+		tenantsPerQuery: promauto.With(reg).NewHistogram(prometheus.HistogramOpts{
+			Namespace: "cortex",
+			Name:      "querier_federated_tenants_per_query",
+			Help:      "Number of tenants per query.",
+			Buckets:   []float64{1, 2, 4, 8, 16, 32, 64},
+		}),
 	}
 }
 
 type mergeQueryable struct {
 	idLabelName             string
 	byPassWithSingleQuerier bool
 	callback                MergeQuerierCallback
+	tenantsPerQuery         prometheus.Histogram
 }
 
 // Querier returns a new mergeQuerier, which aggregates results from multiple
@@ -102,6 +112,7 @@ func (m *mergeQueryable) Querier(mint int64, maxt int64) (storage.Querier, error
 		maxt:                    maxt,
 		byPassWithSingleQuerier: m.byPassWithSingleQuerier,
 		callback:                m.callback,
+		tenantsPerQuery:         m.tenantsPerQuery,
 	}, nil
 }
 
@@ -117,6 +128,7 @@ type mergeQuerier struct {
 	callback    MergeQuerierCallback
 
 	byPassWithSingleQuerier bool
+	tenantsPerQuery         prometheus.Histogram
 }
 
 // LabelValues returns all potential values for a label name.  It is not safe
@@ -130,6 +142,8 @@ func (m *mergeQuerier) LabelValues(ctx context.Context, name string, hints *stor
 		return nil, nil, err
 	}
 
+	m.tenantsPerQuery.Observe(float64(len(ids)))
+
 	// by pass when only single querier is returned
 	if m.byPassWithSingleQuerier && len(queriers) == 1 {
 		return queriers[0].LabelValues(ctx, name, hints, matchers...)
@@ -169,6 +183,8 @@ func (m *mergeQuerier) LabelNames(ctx context.Context, hints *storage.LabelHints
 		return nil, nil, err
 	}
 
+	m.tenantsPerQuery.Observe(float64(len(ids)))
+
 	// by pass when only single querier is returned
 	if m.byPassWithSingleQuerier && len(queriers) == 1 {
 		return queriers[0].LabelNames(ctx, hints, matchers...)
@@ -309,6 +325,8 @@ func (m *mergeQuerier) Select(ctx context.Context, sortSeries bool, hints *stora
 		return storage.ErrSeriesSet(err)
 	}
 
+	m.tenantsPerQuery.Observe(float64(len(ids)))
+
 	// by pass when only single querier is returned
 	if m.byPassWithSingleQuerier && len(queriers) == 1 {
 		return queriers[0].Select(ctx, sortSeries, hints, matchers...)