statistics: add metrics for statscache lru (#34568)

ref pingcap/tidb#34052
2022-05-11 17:26:35 +08:00
parent 9e7c9519da
commit ea4fda51aa
6 changed files with 272 additions and 1 deletions
--- a/metrics/grafana/tidb.json
+++ b/metrics/grafana/tidb.json
@ -11568,6 +11568,224 @@
            "align": false,
            "alignLevel": null
          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_TEST-CLUSTER}",
+          "description": "TiDB managing stats cache by lru",
+          "fieldConfig": {
+            "defaults": {},
+            "overrides": []
+          },
+          "fill": 1,
+          "fillGradient": 0,
+          "gridPos": {
+            "h": 7,
+            "w": 8,
+            "x": 16,
+            "y": 89
+          },
+          "hiddenSeries": false,
+          "id": 246,
+          "legend": {
+            "alignAsTable": true,
+            "avg": true,
+            "current": true,
+            "max": true,
+            "min": false,
+            "rightSide": true,
+            "show": true,
+            "total": false,
+            "values": true
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "options": {
+            "alertThreshold": true
+          },
+          "percentage": false,
+          "pluginVersion": "7.5.11",
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "exemplar": true,
+              "expr": "sum(rate(tidb_statistics_stats_cache_lru_op{k8s_cluster=\"$k8s_cluster\", tidb_cluster=\"$tidb_cluster\"}[1m])) by (type)",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{type}}",
+              "refId": "A",
+              "step": 30
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeRegions": [],
+          "timeShift": null,
+          "title": "Stats Cache LRU OPS",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "$$hashKey": "object:90",
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "$$hashKey": "object:91",
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_TEST-CLUSTER}",
+          "description": "TiDB managing stats cache by lru",
+          "editable": true,
+          "error": false,
+          "fieldConfig": {
+            "defaults": {},
+            "overrides": []
+          },
+          "fill": 0,
+          "fillGradient": 0,
+          "grid": {},
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 96
+          },
+          "hiddenSeries": false,
+          "id": 247,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": true,
+            "hideEmpty": true,
+            "hideZero": false,
+            "max": true,
+            "min": false,
+            "rightSide": true,
+            "show": true,
+            "sideWidth": null,
+            "total": false,
+            "values": true
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null as zero",
+          "options": {
+            "alertThreshold": true
+          },
+          "percentage": false,
+          "pluginVersion": "7.5.11",
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "exemplar": true,
+              "expr": "tidb_statistics_stats_cache_lru_val{k8s_cluster=\"$k8s_cluster\", tidb_cluster=\"$tidb_cluster\", type=\"track\"}",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 1,
+              "legendFormat": "track-{{instance}}",
+              "refId": "A"
+            },
+            {
+              "exemplar": true,
+              "expr": "tidb_statistics_stats_cache_lru_val{k8s_cluster=\"$k8s_cluster\", tidb_cluster=\"$tidb_cluster\", type=\"capacity\"}",
+              "hide": true,
+              "interval": "",
+              "legendFormat": "capacity--{{instance}}",
+              "refId": "B"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeRegions": [],
+          "timeShift": null,
+          "title": "Stats Cache LRU Cost",
+          "tooltip": {
+            "msResolution": true,
+            "shared": true,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "bytes",
+              "label": "",
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": "",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": false
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
        }
      ],
      "repeat": null,
--- a/metrics/metrics.go
+++ b/metrics/metrics.go
@ -176,6 +176,8 @@ func RegisterMetrics() {
 	prometheus.MustRegister(LoadTableCacheDurationHistogram)
 	prometheus.MustRegister(NonTransactionalDeleteCount)
 	prometheus.MustRegister(MemoryUsage)
+	prometheus.MustRegister(StatsCacheLRUCounter)
+	prometheus.MustRegister(StatsCacheLRUGauge)

 	tikvmetrics.InitMetrics(TiDB, TiKVClient)
 	tikvmetrics.RegisterMetrics()
--- a/metrics/stats.go
+++ b/metrics/stats.go
@ -128,4 +128,19 @@ var (
 			Help:      "Bucketed histogram of latency time (ms) of stats read during sync-load.",
 			Buckets:   prometheus.ExponentialBuckets(1, 2, 22), // 1ms ~ 1h
 		})
+
+	StatsCacheLRUCounter = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Namespace: "tidb",
+			Subsystem: "statistics",
+			Name:      "stats_cache_lru_op",
+			Help:      "Counter of lru for statsCache operation",
+		}, []string{LblType})
+
+	StatsCacheLRUGauge = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: "tidb",
+		Subsystem: "statistics",
+		Name:      "stats_cache_lru_val",
+		Help:      "gauge of stats cache lru value",
+	}, []string{LblType})
 )
--- a/statistics/handle/handle.go
+++ b/statistics/handle/handle.go
@ -544,12 +544,17 @@ func (h *Handle) GetPartitionStats(tblInfo *model.TableInfo, pid int64, opts ...
 func (h *Handle) updateStatsCache(newCache statsCache) (updated bool) {
 	h.statsCache.Lock()
 	oldCache := h.statsCache.Load().(statsCache)
+	enableQuota := oldCache.EnableQuota()
+	newCost := newCache.Cost()
 	if oldCache.version < newCache.version || (oldCache.version == newCache.version && oldCache.minorVersion < newCache.minorVersion) {
-		h.statsCache.memTracker.Consume(newCache.Cost() - oldCache.Cost())
+		h.statsCache.memTracker.Consume(newCost - oldCache.Cost())
 		h.statsCache.Store(newCache)
 		updated = true
 	}
 	h.statsCache.Unlock()
+	if updated && enableQuota {
+		costGauge.Set(float64(newCost))
+	}
 	return
 }

--- a/statistics/handle/lru_cache.go
+++ b/statistics/handle/lru_cache.go
@ -19,9 +19,20 @@ import (
 	"math"
 	"sync"

+	"github.com/pingcap/tidb/metrics"
 	"github.com/pingcap/tidb/statistics"
 )

+var (
+	missCounter   = metrics.StatsCacheLRUCounter.WithLabelValues("miss")
+	hitCounter    = metrics.StatsCacheLRUCounter.WithLabelValues("hit")
+	updateCounter = metrics.StatsCacheLRUCounter.WithLabelValues("update")
+	delCounter    = metrics.StatsCacheLRUCounter.WithLabelValues("del")
+	evictCounter  = metrics.StatsCacheLRUCounter.WithLabelValues("evict")
+	costGauge     = metrics.StatsCacheLRUGauge.WithLabelValues("track")
+	capacityGauge = metrics.StatsCacheLRUGauge.WithLabelValues("capacity")
+)
+
 type statsInnerCache struct {
 	sync.RWMutex
 	elements map[int64]*lruMapElement
@ -50,6 +61,7 @@ func newInnerLruCache(c int64) *innerItemLruCache {
 	if c < 1 {
 		c = math.MaxInt64
 	}
+	capacityGauge.Set(float64(c))
 	return &innerItemLruCache{
 		capacity: c,
 		cache:    list.New(),
@ -261,6 +273,11 @@ func (s *statsInnerCache) SetCapacity(c int64) {
 	s.lru.setCapacity(c)
 }

+// EnableQuota implements statsCacheInner
+func (s *statsInnerCache) EnableQuota() bool {
+	return true
+}
+
 func (s *statsInnerCache) onEvict(tblID int64) {
 	element, exist := s.elements[tblID]
 	if !exist {
@ -283,12 +300,15 @@ func (s *statsInnerCache) capacity() int64 {
 func (c *innerItemLruCache) get(tblID, id int64) (*lruCacheItem, bool) {
 	v, ok := c.elements[tblID]
 	if !ok {
+		missCounter.Inc()
 		return nil, false
 	}
 	ele, ok := v[id]
 	if !ok {
+		missCounter.Inc()
 		return nil, false
 	}
+	hitCounter.Inc()
 	c.cache.MoveToFront(ele)
 	return ele.Value.(*lruCacheItem), true
 }
@ -302,6 +322,7 @@ func (c *innerItemLruCache) del(tblID, id int64) {
 	if !ok {
 		return
 	}
+	delCounter.Inc()
 	delete(c.elements[tblID], id)
 	c.cache.Remove(ele)
 }
@ -309,6 +330,7 @@ func (c *innerItemLruCache) del(tblID, id int64) {
 func (c *innerItemLruCache) put(tblID, id int64, item statistics.TableCacheItem, itemMem statistics.CacheItemMemoryUsage,
 	needEvict, needMove bool) {
 	defer func() {
+		updateCounter.Inc()
 		if needEvict {
 			c.evictIfNeeded()
 		}
@ -344,6 +366,7 @@ func (c *innerItemLruCache) put(tblID, id int64, item statistics.TableCacheItem,
 func (c *innerItemLruCache) evictIfNeeded() {
 	curr := c.cache.Back()
 	for c.trackingCost > c.capacity {
+		evictCounter.Inc()
 		prev := curr.Prev()
 		item := curr.Value.(*lruCacheItem)
 		oldMem := item.innerMemUsage
@ -381,5 +404,6 @@ func (c *innerItemLruCache) setCapacity(capacity int64) {
 		capacity = math.MaxInt64
 	}
 	c.capacity = capacity
+	capacityGauge.Set(float64(c.capacity))
 	c.evictIfNeeded()
 }
--- a/statistics/handle/statscache.go
+++ b/statistics/handle/statscache.go
@ -38,6 +38,7 @@ type statsCacheInner interface {
 	FreshMemUsage()
 	Copy() statsCacheInner
 	SetCapacity(int64)
+	EnableQuota() bool
 }

 func newStatsCache() statsCache {
@ -245,4 +246,10 @@ func (m *mapCache) Copy() statsCacheInner {
 	return newM
 }

+// SetCapacity implements statsCacheInner
 func (m *mapCache) SetCapacity(int64) {}
+
+// EnableQuota implements statsCacheInner
+func (m *mapCache) EnableQuota() bool {
+	return false
+}