sql: add sql.txn.contended.count metrics

Expose transaction contention metric for cockroachdb#73566. Release note (sql change): new sql.txn.contended.count metric exposes total number of transactions that experienced contentions.
Azhng · May 4, 2022 · 9fbaefc · 9fbaefc
1 parent 609825e
commit 9fbaefc
Show file tree

Hide file tree

Showing 6 changed files with 30 additions and 0 deletions.
diff --git a/pkg/server/status_test.go b/pkg/server/status_test.go
@@ -3138,6 +3138,9 @@ func TestStatusAPIContentionEvents(t *testing.T) {
 	server1Conn := sqlutils.MakeSQLRunner(testCluster.ServerConn(0))
 	server2Conn := sqlutils.MakeSQLRunner(testCluster.ServerConn(1))
 
+	contentionCountBefore := testCluster.Server(1).SQLServer().(*sql.Server).
+		Metrics.EngineMetrics.SQLContendedTxns.Count()
+
 	sqlutils.CreateTable(
 		t,
 		testCluster.ServerConn(0),
@@ -3214,6 +3217,13 @@ SET TRACING=off;
     (statistics -> 'execution_statistics' -> 'contentionTime' ->> 'mean')::FLOAT > 0
     AND app_name = 'contentionTest'
 `, [][]string{{"1"}})
+
+	contentionCountNow := testCluster.Server(1).SQLServer().(*sql.Server).
+		Metrics.EngineMetrics.SQLContendedTxns.Count()
+
+	require.Greaterf(t, contentionCountNow, contentionCountBefore,
+		"expected txn contention count to be more than %d, but it is %d",
+		contentionCountBefore, contentionCountNow)
 }
 
 func TestStatusCancelSessionGatewayMetadataPropagation(t *testing.T) {

diff --git a/pkg/sql/conn_executor.go b/pkg/sql/conn_executor.go
@@ -429,6 +429,7 @@ func makeMetrics(internal bool) Metrics {
 				6*metricsSampleInterval),
 			SQLTxnsOpen:         metric.NewGauge(getMetricMeta(MetaSQLTxnsOpen, internal)),
 			SQLActiveStatements: metric.NewGauge(getMetricMeta(MetaSQLActiveQueries, internal)),
+			SQLContendedTxns:    metric.NewCounter(getMetricMeta(MetaSQLTxnContended, internal)),
 
 			TxnAbortCount:                     metric.NewCounter(getMetricMeta(MetaTxnAbort, internal)),
 			FailureCount:                      metric.NewCounter(getMetricMeta(MetaFailure, internal)),

diff --git a/pkg/sql/conn_executor_exec.go b/pkg/sql/conn_executor_exec.go
@@ -2216,6 +2216,10 @@ func (ex *connExecutor) recordTransactionFinish(
 	}
 	ex.metrics.EngineMetrics.SQLTxnLatency.RecordValue(txnTime.Nanoseconds())
 
+	if contentionDuration := ex.extraTxnState.accumulatedStats.ContentionTime.Nanoseconds(); contentionDuration > 0 {
+		ex.metrics.EngineMetrics.SQLContendedTxns.Inc(1)
+	}
+
 	ex.txnIDCacheWriter.Record(contentionpb.ResolvedTxnID{
 		TxnID:            ev.txnID,
 		TxnFingerprintID: transactionFingerprintID,

diff --git a/pkg/sql/exec_util.go b/pkg/sql/exec_util.go
@@ -850,6 +850,12 @@ var (
 		Measurement: "SQL Statements",
 		Unit:        metric.Unit_COUNT,
 	}
+	MetaSQLTxnContended = metric.Metadata{
+		Name:        "sql.txn.contended.count",
+		Help:        "Number of SQL transactions experienced contention",
+		Measurement: "Contention",
+		Unit:        metric.Unit_COUNT,
+	}
 	MetaSelectStarted = metric.Metadata{
 		Name:        "sql.select.started.count",
 		Help:        "Number of SQL SELECT statements started",

diff --git a/pkg/sql/executor_statement_metrics.go b/pkg/sql/executor_statement_metrics.go
@@ -40,6 +40,7 @@ type EngineMetrics struct {
 	SQLTxnLatency         *metric.Histogram
 	SQLTxnsOpen           *metric.Gauge
 	SQLActiveStatements   *metric.Gauge
+	SQLContendedTxns      *metric.Counter
 
 	// TxnAbortCount counts transactions that were aborted, either due
 	// to non-retriable errors, or retriable errors when the client-side

diff --git a/pkg/ts/catalog/chart_catalog.go b/pkg/ts/catalog/chart_catalog.go
@@ -2363,6 +2363,14 @@ var charts = []sectionDescription{
 				},
 				AxisLabel: "Count",
 			},
+			{
+				Title: "SQL Transaction Contention",
+				Metrics: []string{
+					"sql.txn.contended.count",
+					"sql.txn.contended.count.internal",
+				},
+				AxisLabel: "Transactions",
+			},
 		},
 	},
 	{