metrics: remove the Heartbeat distribution (#9010)

close #9003 Signed-off-by: lhy1024 <[email protected]>
tikv · Jan 17, 2025 · 2902597 · 2902597
1 parent a99abe3
commit 2902597
Show file tree

Hide file tree

Showing 9 changed files with 0 additions and 594 deletions.
diff --git a/metrics/grafana/pd.json b/metrics/grafana/pd.json
diff --git a/pkg/mcs/scheduling/server/cluster.go b/pkg/mcs/scheduling/server/cluster.go
@@ -414,9 +414,6 @@ func (c *Cluster) HandleStoreHeartbeat(heartbeat *schedulingpb.StoreHeartbeatReq
 	nowTime := time.Now()
 	newStore := store.Clone(core.SetStoreStats(stats), core.SetLastHeartbeatTS(nowTime))
 
-	if store := c.GetStore(storeID); store != nil {
-		statistics.UpdateStoreHeartbeatMetrics(store)
-	}
 	c.PutStore(newStore)
 	c.hotStat.Observe(storeID, newStore.GetStoreStats())
 	c.hotStat.FilterUnhealthyStore(c)

diff --git a/pkg/statistics/buckets/bucket_stat_informer.go b/pkg/statistics/buckets/bucket_stat_informer.go
@@ -208,10 +208,3 @@ func (b *BucketTreeItem) calculateHotDegree() {
 		}
 	}
 }
-
-// collectBucketsMetrics collects the metrics of the hot stats.
-func (b *BucketTreeItem) collectBucketsMetrics() {
-	for _, bucket := range b.stats {
-		bucketsHotDegreeHist.Observe(float64(bucket.HotDegree))
-	}
-}
diff --git a/pkg/statistics/buckets/hot_bucket_cache.go b/pkg/statistics/buckets/hot_bucket_cache.go
@@ -201,7 +201,6 @@ func (h *HotBucketCache) checkBucketsFlow(buckets *metapb.Buckets) (newItem *Buc
 	}
 	newItem.inherit(overlaps)
 	newItem.calculateHotDegree()
-	newItem.collectBucketsMetrics()
 	return newItem, overlaps
 }
 

diff --git a/pkg/statistics/buckets/metric.go b/pkg/statistics/buckets/metric.go
@@ -19,15 +19,6 @@ import (
 )
 
 var (
-	bucketsHotDegreeHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "buckets_hot_degree_hist",
-			Help:      "Bucketed histogram of bucket hot degree",
-			Buckets:   prometheus.LinearBuckets(-20, 2, 20), // [-20 20]
-		})
-
 	bucketsTaskDuration = prometheus.NewHistogramVec(
 		prometheus.HistogramOpts{
 			Namespace: "pd",
@@ -39,6 +30,5 @@ var (
 )
 
 func init() {
-	prometheus.MustRegister(bucketsHotDegreeHist)
 	prometheus.MustRegister(bucketsTaskDuration)
 }
diff --git a/pkg/statistics/hot_peer_cache.go b/pkg/statistics/hot_peer_cache.go
@@ -135,30 +135,6 @@ func (f *HotPeerCache) incMetrics(action utils.ActionType, storeID uint64) {
 	f.metrics[storeID][action].Inc()
 }
 
-func (f *HotPeerCache) collectPeerMetrics(loads []float64, interval uint64) {
-	regionHeartbeatIntervalHist.Observe(float64(interval))
-	if interval == 0 {
-		return
-	}
-	// TODO: use unified metrics. (keep backward compatibility at the same time)
-	for _, k := range f.kind.RegionStats() {
-		switch k {
-		case utils.RegionReadBytes:
-			readByteHist.Observe(loads[int(k)])
-		case utils.RegionReadKeys:
-			readKeyHist.Observe(loads[int(k)])
-		case utils.RegionWriteBytes:
-			writeByteHist.Observe(loads[int(k)])
-		case utils.RegionWriteKeys:
-			writeKeyHist.Observe(loads[int(k)])
-		case utils.RegionWriteQueryNum:
-			writeQueryHist.Observe(loads[int(k)])
-		case utils.RegionReadQueryNum:
-			readQueryHist.Observe(loads[int(k)])
-		}
-	}
-}
-
 // CollectExpiredItems collects expired items, mark them as needDelete and puts them into inherit items
 func (f *HotPeerCache) CollectExpiredItems(region *core.RegionInfo) []*HotPeerStat {
 	regionID := region.GetID()
@@ -185,7 +161,6 @@ func (f *HotPeerCache) CheckPeerFlow(region *core.RegionInfo, peers []*metapb.Pe
 		return nil
 	}
 
-	f.collectPeerMetrics(deltaLoads, interval) // update metrics
 	regionID := region.GetID()
 
 	regionPeers := region.GetPeers()

diff --git a/pkg/statistics/metrics.go b/pkg/statistics/metrics.go
@@ -81,70 +81,6 @@ var (
 			Name:      "label_level",
 			Help:      "Number of regions in the different label level.",
 		}, []string{"type"})
-	readByteHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "read_byte_hist",
-			Help:      "The distribution of region read bytes",
-			Buckets:   prometheus.ExponentialBuckets(1, 8, 12),
-		})
-	writeByteHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "write_byte_hist",
-			Help:      "The distribution of region write bytes",
-			Buckets:   prometheus.ExponentialBuckets(1, 8, 12),
-		})
-	readKeyHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "read_key_hist",
-			Help:      "The distribution of region read keys",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 18),
-		})
-	writeKeyHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "write_key_hist",
-			Help:      "The distribution of region write keys",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 18),
-		})
-	readQueryHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "read_query_hist",
-			Help:      "The distribution of region read query",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 12),
-		})
-	writeQueryHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "write_query_hist",
-			Help:      "The distribution of region write query",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 12),
-		})
-	regionHeartbeatIntervalHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "region_heartbeat_interval_hist",
-			Help:      "Bucketed histogram of the batch size of handled requests.",
-			Buckets:   prometheus.LinearBuckets(0, 30, 20),
-		})
-	storeHeartbeatIntervalHist = prometheus.NewHistogram(
-		prometheus.HistogramOpts{
-			Namespace: "pd",
-			Subsystem: "scheduler",
-			Name:      "store_heartbeat_interval_hist",
-			Help:      "Bucketed histogram of the batch size of handled requests.",
-			Buckets:   prometheus.LinearBuckets(0, 5, 12),
-		})
 
 	regionAbnormalPeerDuration = prometheus.NewHistogramVec(
 		prometheus.HistogramOpts{
@@ -187,12 +123,6 @@ func init() {
 	prometheus.MustRegister(configStatusGauge)
 	prometheus.MustRegister(StoreLimitGauge)
 	prometheus.MustRegister(regionLabelLevelGauge)
-	prometheus.MustRegister(readByteHist)
-	prometheus.MustRegister(readKeyHist)
-	prometheus.MustRegister(writeKeyHist)
-	prometheus.MustRegister(writeByteHist)
-	prometheus.MustRegister(regionHeartbeatIntervalHist)
-	prometheus.MustRegister(storeHeartbeatIntervalHist)
 	prometheus.MustRegister(regionAbnormalPeerDuration)
 	prometheus.MustRegister(hotCacheFlowQueueStatusGauge)
 	prometheus.MustRegister(hotPeerSummary)

diff --git a/pkg/statistics/store.go b/pkg/statistics/store.go
@@ -129,11 +129,6 @@ func (s *StoresStats) FilterUnhealthyStore(cluster core.StoreSetInformer) {
 	}
 }
 
-// UpdateStoreHeartbeatMetrics is used to update store heartbeat interval metrics
-func UpdateStoreHeartbeatMetrics(store *core.StoreInfo) {
-	storeHeartbeatIntervalHist.Observe(time.Since(store.GetLastHeartbeatTS()).Seconds())
-}
-
 // RollingStoreStats are multiple sets of recent historical records with specified windows size.
 type RollingStoreStats struct {
 	syncutil.RWMutex

diff --git a/server/cluster/cluster.go b/server/cluster/cluster.go
@@ -1040,7 +1040,6 @@ func (c *RaftCluster) HandleStoreHeartbeat(heartbeat *pdpb.StoreHeartbeatRequest
 			newStore = newStore.Clone(core.SetLastPersistTime(nowTime))
 		}
 	}
-	statistics.UpdateStoreHeartbeatMetrics(store)
 	// Supply NodeState in the response to help the store handle special cases
 	// more conveniently, such as avoiding calling `remove_peer` redundantly under
 	// NodeState_Removing.