metrics: adjust metrics and its comments (pingcap#16429)

glorv · Apr 15, 2020 · 2c8afe6 · 2c8afe6
1 parent 62c1c8d
commit 2c8afe6
Show file tree

Hide file tree

Showing 14 changed files with 49 additions and 50 deletions.
diff --git a/distsql/select_result.go b/distsql/select_result.go
@@ -107,7 +107,7 @@ func (r *selectResult) fetchResp(ctx context.Context) error {
 				// final round of fetch
 				// TODO: Add a label to distinguish between success or failure.
 				// https://github.com/pingcap/tidb/issues/11397
-				metrics.DistSQLQueryHistgram.WithLabelValues(r.label, r.sqlType).Observe(r.fetchDuration.Seconds())
+				metrics.DistSQLQueryHistogram.WithLabelValues(r.label, r.sqlType).Observe(r.fetchDuration.Seconds())
 				r.durationReported = true
 			}
 			return nil

diff --git a/distsql/stream.go b/distsql/stream.go
@@ -72,7 +72,7 @@ func (r *streamResult) readDataFromResponse(ctx context.Context, resp kv.Respons
 	resultSubset, err := resp.Next(ctx)
 	// TODO: Add a label to distinguish between success or failure.
 	// https://github.com/pingcap/tidb/issues/11397
-	metrics.DistSQLQueryHistgram.WithLabelValues(r.label, r.sqlType).Observe(time.Since(startTime).Seconds())
+	metrics.DistSQLQueryHistogram.WithLabelValues(r.label, r.sqlType).Observe(time.Since(startTime).Seconds())
 	if err != nil {
 		return false, err
 	}

diff --git a/metrics/ddl.go b/metrics/ddl.go
@@ -31,7 +31,7 @@ var (
 			Subsystem: "ddl",
 			Name:      "handle_job_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of handle jobs",
-			Buckets:   prometheus.ExponentialBuckets(0.01, 2, 22), // 10ms ~ 12hours
+			Buckets:   prometheus.ExponentialBuckets(0.01, 2, 22), // 10ms ~ 6hours
 		}, []string{LblType, LblResult})
 
 	BatchAddIdxHistogram = prometheus.NewHistogramVec(
@@ -40,7 +40,7 @@ var (
 			Subsystem: "ddl",
 			Name:      "batch_add_idx_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of batch handle data",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 1hours
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 0.5hours
 		}, []string{LblType})
 
 	SyncerInit            = "init"
@@ -53,7 +53,7 @@ var (
 			Subsystem: "ddl",
 			Name:      "deploy_syncer_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of deploy syncer",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 1024s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 524s
 		}, []string{LblType, LblResult})
 
 	UpdateSelfVersionHistogram = prometheus.NewHistogramVec(
@@ -62,7 +62,7 @@ var (
 			Subsystem: "ddl",
 			Name:      "update_self_ver_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of update self version",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 1024s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 524s
 		}, []string{LblResult})
 
 	OwnerUpdateGlobalVersion    = "update_global_version"
@@ -77,7 +77,7 @@ var (
 			Subsystem: "ddl",
 			Name:      "owner_handle_syncer_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of handle syncer",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 1024s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 524s
 		}, []string{LblType, LblResult})
 
 	// Metrics for ddl_worker.go.
@@ -91,7 +91,7 @@ var (
 			Subsystem: "ddl",
 			Name:      "worker_operation_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of ddl worker operations",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 4096s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 2048s
 		}, []string{LblType, LblAction, LblResult})
 
 	CreateDDLInstance = "create_ddl_instance"

diff --git a/metrics/distsql.go b/metrics/distsql.go
@@ -19,13 +19,13 @@ import (
 
 // distsql metrics.
 var (
-	DistSQLQueryHistgram = prometheus.NewHistogramVec(
+	DistSQLQueryHistogram = prometheus.NewHistogramVec(
 		prometheus.HistogramOpts{
 			Namespace: "tidb",
 			Subsystem: "distsql",
 			Name:      "handle_query_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of handled queries.",
-			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 18), // 0.5ms ~ 128s
+			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 18), // 0.5ms ~ 64s
 		}, []string{LblType, LblSQLType})
 
 	DistSQLScanKeysPartialHistogram = prometheus.NewHistogram(

diff --git a/metrics/domain.go b/metrics/domain.go
@@ -35,7 +35,7 @@ var (
 			Subsystem: "domain",
 			Name:      "load_schema_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) in load schema.",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 4000s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 524s
 		})
 
 	// LoadPrivilegeCounter records the counter of load privilege.

diff --git a/metrics/gc_worker.go b/metrics/gc_worker.go
@@ -33,7 +33,7 @@ var (
 			Subsystem: "tikvclient",
 			Name:      "gc_seconds",
 			Help:      "Bucketed histogram of gc duration.",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 20), // 1s ~ 12days
+			Buckets:   prometheus.ExponentialBuckets(1, 2, 20), // 1s ~ 6days
 		}, []string{"stage"})
 
 	GCConfigGauge = prometheus.NewGaugeVec(

diff --git a/metrics/meta.go b/metrics/meta.go
@@ -28,7 +28,7 @@ var (
 			Subsystem: "autoid",
 			Name:      "operation_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of handled autoid.",
-			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 500us ~ 2097s
+			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 0.5ms ~ 1048s
 		}, []string{LblType, LblResult})
 
 	GetSchemaDiff    = "get_schema_diff"
@@ -43,6 +43,6 @@ var (
 			Subsystem: "meta",
 			Name:      "operation_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of tidb meta data operations.",
-			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 500us ~ 2097s
+			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 0.5ms ~ 1048s
 		}, []string{LblType, LblResult})
 )
diff --git a/metrics/metrics.go b/metrics/metrics.go
@@ -76,7 +76,7 @@ func RegisterMetrics() {
 	prometheus.MustRegister(DDLWorkerHistogram)
 	prometheus.MustRegister(DeploySyncerHistogram)
 	prometheus.MustRegister(DistSQLPartialCountHistogram)
-	prometheus.MustRegister(DistSQLQueryHistgram)
+	prometheus.MustRegister(DistSQLQueryHistogram)
 	prometheus.MustRegister(DistSQLScanKeysHistogram)
 	prometheus.MustRegister(DistSQLScanKeysPartialHistogram)
 	prometheus.MustRegister(DumpFeedbackCounter)

diff --git a/metrics/oracles.go b/metrics/oracles.go
@@ -25,6 +25,6 @@ var (
 			Subsystem: "pdclient",
 			Name:      "ts_future_wait_seconds",
 			Help:      "Bucketed histogram of seconds cost for waiting timestamp future.",
-			Buckets:   prometheus.ExponentialBuckets(0.000005, 2, 20), // 5us ~ 5s
+			Buckets:   prometheus.ExponentialBuckets(0.000005, 2, 24), // 5us ~ 40s
 		})
 )
diff --git a/metrics/owner.go b/metrics/owner.go
@@ -25,7 +25,7 @@ var (
 			Subsystem: "owner",
 			Name:      "new_session_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of new session.",
-			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 500us ~ 2097s
+			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 0.5ms ~ 1048s
 		}, []string{LblType, LblResult})
 
 	WatcherClosed     = "watcher_closed"

diff --git a/metrics/server.go b/metrics/server.go
@@ -34,7 +34,7 @@ var (
 			Subsystem: "server",
 			Name:      "handle_query_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of handled queries.",
-			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 500us ~ 2097s
+			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 22), // 0.5ms ~ 1048s
 		}, []string{LblSQLType})
 
 	QueryTotalCounter = prometheus.NewCounterVec(
@@ -128,7 +128,7 @@ var (
 			Subsystem: "server",
 			Name:      "get_token_duration_seconds",
 			Help:      "Duration (us) for getting token, it should be small until concurrency limit is reached.",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 26), // 1us ~ 67s
+			Buckets:   prometheus.ExponentialBuckets(1, 2, 26), // 1us ~ 33s
 		})
 
 	TotalQueryProcHistogram = prometheus.NewHistogram(
@@ -137,23 +137,23 @@ var (
 			Subsystem: "server",
 			Name:      "slow_query_process_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of of slow queries.",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 4096s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 2048s
 		})
 	TotalCopProcHistogram = prometheus.NewHistogram(
 		prometheus.HistogramOpts{
 			Namespace: "tidb",
 			Subsystem: "server",
 			Name:      "slow_query_cop_duration_seconds",
 			Help:      "Bucketed histogram of all cop processing time (s) of of slow queries.",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 4096s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 2048s
 		})
 	TotalCopWaitHistogram = prometheus.NewHistogram(
 		prometheus.HistogramOpts{
 			Namespace: "tidb",
 			Subsystem: "server",
 			Name:      "slow_query_wait_duration_seconds",
 			Help:      "Bucketed histogram of all cop waiting time (s) of of slow queries.",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 4096s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 22), // 1ms ~ 2048s
 		})
 )
 

diff --git a/metrics/session.go b/metrics/session.go
@@ -23,7 +23,7 @@ var (
 			Subsystem: "session",
 			Name:      "parse_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) in parse SQL.",
-			Buckets:   prometheus.ExponentialBuckets(0.00004, 2, 22), // 40us ~ 168s
+			Buckets:   prometheus.ExponentialBuckets(0.00004, 2, 22), // 40us ~ 84s
 		}, []string{LblSQLType})
 	SessionExecuteCompileDuration = prometheus.NewHistogramVec(
 		prometheus.HistogramOpts{
@@ -32,15 +32,15 @@ var (
 			Name:      "compile_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) in query optimize.",
 			// Build plan may execute the statement, or allocate table ID, so it might take a long time.
-			Buckets: prometheus.ExponentialBuckets(0.00004, 2, 22), // 40us ~ 168s
+			Buckets: prometheus.ExponentialBuckets(0.00004, 2, 22), // 40us ~ 84s
 		}, []string{LblSQLType})
 	SessionExecuteRunDuration = prometheus.NewHistogramVec(
 		prometheus.HistogramOpts{
 			Namespace: "tidb",
 			Subsystem: "session",
 			Name:      "execute_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) in running executor.",
-			Buckets:   prometheus.ExponentialBuckets(0.0001, 2, 22), // 100us ~ 419s
+			Buckets:   prometheus.ExponentialBuckets(0.0001, 2, 22), // 100us ~ 209s
 		}, []string{LblSQLType})
 	SchemaLeaseErrorCounter = prometheus.NewCounterVec(
 		prometheus.CounterOpts{
@@ -55,7 +55,7 @@ var (
 			Subsystem: "session",
 			Name:      "retry_num",
 			Help:      "Bucketed histogram of session retry count.",
-			Buckets:   prometheus.LinearBuckets(0, 1, 20), // 0 ~ 20
+			Buckets:   prometheus.LinearBuckets(0, 1, 21), // 0 ~ 20
 		})
 	SessionRetryErrorCounter = prometheus.NewCounterVec(
 		prometheus.CounterOpts{
@@ -79,7 +79,7 @@ var (
 			Subsystem: "session",
 			Name:      "transaction_statement_num",
 			Help:      "Bucketed histogram of statements count in each transaction.",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 16), // 1 ~ 65536
+			Buckets:   prometheus.ExponentialBuckets(1, 2, 16), // 1 ~ 32768
 		}, []string{LblSQLType, LblType})
 
 	TransactionDuration = prometheus.NewHistogramVec(
@@ -88,7 +88,7 @@ var (
 			Subsystem: "session",
 			Name:      "transaction_duration_seconds",
 			Help:      "Bucketed histogram of a transaction execution duration, including retry.",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 1049s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 524s
 		}, []string{LblSQLType, LblType})
 
 	StatementDeadlockDetectDuration = prometheus.NewHistogram(
@@ -97,7 +97,7 @@ var (
 			Subsystem: "session",
 			Name:      "statement_deadlock_detect_duration_seconds",
 			Help:      "Bucketed histogram of a statement deadlock detect duration.",
-			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 1049s
+			Buckets:   prometheus.ExponentialBuckets(0.001, 2, 20), // 1ms ~ 524s
 		},
 	)
 
@@ -107,7 +107,7 @@ var (
 			Subsystem: "session",
 			Name:      "statement_pessimistic_retry_count",
 			Help:      "Bucketed histogram of statement pessimistic retry count",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 16), // 1 ~ 65536
+			Buckets:   prometheus.ExponentialBuckets(1, 2, 16), // 1 ~ 32768
 		})
 
 	StatementLockKeysCount = prometheus.NewHistogram(
@@ -116,7 +116,7 @@ var (
 			Subsystem: "session",
 			Name:      "statement_lock_keys_count",
 			Help:      "Keys locking for a single statement",
-			Buckets:   prometheus.ExponentialBuckets(1, 2, 21), // 1 ~ 2097152
+			Buckets:   prometheus.ExponentialBuckets(1, 2, 21), // 1 ~ 1048576
 		})
 )
 

diff --git a/metrics/stats.go b/metrics/stats.go
@@ -25,7 +25,7 @@ var (
 			Subsystem: "statistics",
 			Name:      "auto_analyze_duration_seconds",
 			Help:      "Bucketed histogram of processing time (s) of auto analyze.",
-			Buckets:   prometheus.ExponentialBuckets(0.01, 2, 20), // 10ms ~ 3hours
+			Buckets:   prometheus.ExponentialBuckets(0.01, 2, 20), // 10ms ~ 1.5hours
 		})
 
 	AutoAnalyzeCounter = prometheus.NewCounterVec(