Add support for sampling rate in streamlog (#15919)

Signed-off-by: Tim Vaillancourt <[email protected]>

go/flags/endtoend/vtcombo.txt

@@ -272,6 +272,7 @@ Flags:
       --querylog-filter-tag string                                       string that must be present in the query for it to be logged; if using a value as the tag, you need to disable query normalization
       --querylog-format string                                           format for query logs ("text" or "json") (default "text")
       --querylog-row-threshold uint                                      Number of rows a query has to return or affect before being logged; not useful for streaming queries. 0 means all queries will be logged.
+      --querylog-sample-rate float                                       Sample rate for logging queries. Value must be between 0.0 (no logging) and 1.0 (all queries)
       --queryserver-config-acl-exempt-acl string                         an acl that exempt from table acl checking (this acl is free to access any vitess tables).
       --queryserver-config-annotate-queries                              prefix queries to MySQL backend with comment indicating vtgate principal (user) and target tablet type
       --queryserver-config-enable-table-acl-dry-run                      If this flag is enabled, tabletserver will emit monitoring metrics and let the request pass regardless of table acl check results

go/flags/endtoend/vtgate.txt

@@ -174,6 +174,7 @@ Flags:
       --querylog-filter-tag string                                       string that must be present in the query for it to be logged; if using a value as the tag, you need to disable query normalization
       --querylog-format string                                           format for query logs ("text" or "json") (default "text")
       --querylog-row-threshold uint                                      Number of rows a query has to return or affect before being logged; not useful for streaming queries. 0 means all queries will be logged.
+      --querylog-sample-rate float                                       Sample rate for logging queries. Value must be between 0.0 (no logging) and 1.0 (all queries)
       --redact-debug-ui-queries                                          redact full queries and bind variables from debug UI
       --remote_operation_timeout duration                                time to wait for a remote operation (default 15s)
       --retry-count int                                                  retry count (default 2)

go/flags/endtoend/vttablet.txt

@@ -263,6 +263,7 @@ Flags:
       --querylog-filter-tag string                                       string that must be present in the query for it to be logged; if using a value as the tag, you need to disable query normalization
       --querylog-format string                                           format for query logs ("text" or "json") (default "text")
       --querylog-row-threshold uint                                      Number of rows a query has to return or affect before being logged; not useful for streaming queries. 0 means all queries will be logged.
+      --querylog-sample-rate float                                       Sample rate for logging queries. Value must be between 0.0 (no logging) and 1.0 (all queries)
       --queryserver-config-acl-exempt-acl string                         an acl that exempt from table acl checking (this acl is free to access any vitess tables).
       --queryserver-config-annotate-queries                              prefix queries to MySQL backend with comment indicating vtgate principal (user) and target tablet type
       --queryserver-config-enable-table-acl-dry-run                      If this flag is enabled, tabletserver will emit monitoring metrics and let the request pass regardless of table acl check results

go/streamlog/streamlog.go

@@ -20,6 +20,7 @@ package streamlog
 import (
 	"fmt"
 	"io"
+	rand "math/rand/v2"
 	"net/http"
 	"net/url"
 	"os"
@@ -51,6 +52,7 @@ var (
 	queryLogFilterTag    string
 	queryLogRowThreshold uint64
 	queryLogFormat       = "text"
+	queryLogSampleRate   float64
 )
 
 func GetRedactDebugUIQueries() bool {
@@ -69,6 +71,10 @@ func SetQueryLogRowThreshold(newQueryLogRowThreshold uint64) {
 	queryLogRowThreshold = newQueryLogRowThreshold
 }
 
+func SetQueryLogSampleRate(sampleRate float64) {
+	queryLogSampleRate = sampleRate
+}
+
 func GetQueryLogFormat() string {
 	return queryLogFormat
 }
@@ -96,6 +102,8 @@ func registerStreamLogFlags(fs *pflag.FlagSet) {
 	// QueryLogRowThreshold only log queries returning or affecting this many rows
 	fs.Uint64Var(&queryLogRowThreshold, "querylog-row-threshold", queryLogRowThreshold, "Number of rows a query has to return or affect before being logged; not useful for streaming queries. 0 means all queries will be logged.")
 
+	// QueryLogSampleRate causes a sample of queries to be logged
+	fs.Float64Var(&queryLogSampleRate, "querylog-sample-rate", queryLogSampleRate, "Sample rate for logging queries. Value must be between 0.0 (no logging) and 1.0 (all queries)")
 }
 
 const (
@@ -249,9 +257,22 @@ func GetFormatter[T any](logger *StreamLogger[T]) LogFormatter {
 	}
 }
 
+// shouldSampleQuery returns true if a query should be sampled based on queryLogSampleRate
+func shouldSampleQuery() bool {
+	if queryLogSampleRate <= 0 {
+		return false
+	} else if queryLogSampleRate >= 1 {
+		return true
+	}
+	return rand.Float64() <= queryLogSampleRate
+}
+
 // ShouldEmitLog returns whether the log with the given SQL query
 // should be emitted or filtered
 func ShouldEmitLog(sql string, rowsAffected, rowsReturned uint64) bool {
+	if shouldSampleQuery() {
+		return true
+	}
 	if queryLogRowThreshold > max(rowsAffected, rowsReturned) && queryLogFilterTag == "" {
 		return false
 	}

go/streamlog/streamlog_test.go

@@ -30,6 +30,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
 	"vitess.io/vitess/go/vt/servenv"
@@ -264,18 +265,39 @@ func TestFile(t *testing.T) {
 	}
 }
 
+func TestShouldSampleQuery(t *testing.T) {
+	queryLogSampleRate = -1
+	assert.False(t, shouldSampleQuery())
+
+	queryLogSampleRate = 0
+	assert.False(t, shouldSampleQuery())
+
+	// for test coverage, can't test a random result
+	queryLogSampleRate = 0.5
+	shouldSampleQuery()
+
+	queryLogSampleRate = 1.0
+	assert.True(t, shouldSampleQuery())
+
+	queryLogSampleRate = 100.0
+	assert.True(t, shouldSampleQuery())
+}
+
 func TestShouldEmitLog(t *testing.T) {
 	origQueryLogFilterTag := queryLogFilterTag
 	origQueryLogRowThreshold := queryLogRowThreshold
+	origQueryLogSampleRate := queryLogSampleRate
 	defer func() {
 		SetQueryLogFilterTag(origQueryLogFilterTag)
 		SetQueryLogRowThreshold(origQueryLogRowThreshold)
+		SetQueryLogSampleRate(origQueryLogSampleRate)
 	}()
 
 	tests := []struct {
 		sql              string
 		qLogFilterTag    string
 		qLogRowThreshold uint64
+		qLogSampleRate   float64
 		rowsAffected     uint64
 		rowsReturned     uint64
 		ok               bool
@@ -284,6 +306,7 @@ func TestShouldEmitLog(t *testing.T) {
 			sql:              "queryLogThreshold smaller than affected and returned",
 			qLogFilterTag:    "",
 			qLogRowThreshold: 2,
+			qLogSampleRate:   0.0,
 			rowsAffected:     7,
 			rowsReturned:     7,
 			ok:               true,
@@ -292,6 +315,7 @@ func TestShouldEmitLog(t *testing.T) {
 			sql:              "queryLogThreshold greater than affected and returned",
 			qLogFilterTag:    "",
 			qLogRowThreshold: 27,
+			qLogSampleRate:   0.0,
 			rowsAffected:     7,
 			rowsReturned:     17,
 			ok:               false,
@@ -300,6 +324,7 @@ func TestShouldEmitLog(t *testing.T) {
 			sql:              "this doesn't contains queryFilterTag: TAG",
 			qLogFilterTag:    "special tag",
 			qLogRowThreshold: 10,
+			qLogSampleRate:   0.0,
 			rowsAffected:     7,
 			rowsReturned:     17,
 			ok:               false,
@@ -308,6 +333,25 @@ func TestShouldEmitLog(t *testing.T) {
 			sql:              "this contains queryFilterTag: TAG",
 			qLogFilterTag:    "TAG",
 			qLogRowThreshold: 0,
+			qLogSampleRate:   0.0,
+			rowsAffected:     7,
+			rowsReturned:     17,
+			ok:               true,
+		},
+		{
+			sql:              "this contains querySampleRate: 1.0",
+			qLogFilterTag:    "",
+			qLogRowThreshold: 0,
+			qLogSampleRate:   1.0,
+			rowsAffected:     7,
+			rowsReturned:     17,
+			ok:               true,
+		},
+		{
+			sql:              "this contains querySampleRate: 1.0 without expected queryFilterTag",
+			qLogFilterTag:    "TAG",
+			qLogRowThreshold: 0,
+			qLogSampleRate:   1.0,
 			rowsAffected:     7,
 			rowsReturned:     17,
 			ok:               true,
@@ -318,12 +362,37 @@ func TestShouldEmitLog(t *testing.T) {
 		t.Run(tt.sql, func(t *testing.T) {
 			SetQueryLogFilterTag(tt.qLogFilterTag)
 			SetQueryLogRowThreshold(tt.qLogRowThreshold)
+			SetQueryLogSampleRate(tt.qLogSampleRate)
 
 			require.Equal(t, tt.ok, ShouldEmitLog(tt.sql, tt.rowsAffected, tt.rowsReturned))
 		})
 	}
 }
 
+func BenchmarkShouldEmitLog(b *testing.B) {
+	b.Run("default", func(b *testing.B) {
+		SetQueryLogSampleRate(0.0)
+		for i := 0; i < b.N; i++ {
+			ShouldEmitLog("select * from test where user='someone'", 0, 123)
+		}
+	})
+	b.Run("filter_tag", func(b *testing.B) {
+		SetQueryLogSampleRate(0.0)
+		SetQueryLogFilterTag("LOG_QUERY")
+		defer SetQueryLogFilterTag("")
+		for i := 0; i < b.N; i++ {
+			ShouldEmitLog("select /* LOG_QUERY=1 */ * from test where user='someone'", 0, 123)
+		}
+	})
+	b.Run("50%_sample_rate", func(b *testing.B) {
+		SetQueryLogSampleRate(0.5)
+		defer SetQueryLogSampleRate(0.0)
+		for i := 0; i < b.N; i++ {
+			ShouldEmitLog("select * from test where user='someone'", 0, 123)
+		}
+	})
+}
+
 func TestGetFormatter(t *testing.T) {
 	tests := []struct {
 		name           string