Merge #155656

155656: pgwire: make max repeated error count configurable via cluster setting r=rafiss a=rafiss

Previously, the maximum number of repeated network read errors before aborting a connection was a hardcoded constant set to 256 (1 << 8). This change makes the value configurable via a non-public cluster setting `sql.pgwire.max_repeated_error_count`.

This allows operators to tune the threshold for aborting connections experiencing repeated network errors, and allows us to backport this change along with 5f562ad.

Epic: None
Release note: None

Co-authored-by: Rafi Shamim <rafi@cockroachlabs.com>

Author: craig[bot]

pkg/sql/pgwire/conn_test.go

@@ -1449,8 +1449,11 @@ func TestConnServerAbortsOnRepeatedErrors(t *testing.T) {
 	conn, err := db.Conn(ctx)
 	require.NoError(t, err)
 
+	// Get the current value of the cluster setting.
+	maxErrors := int(maxRepeatedErrorCount.Get(&srv.ClusterSettings().SV))
+
 	atomic.StoreUint32(&shouldError, 1)
-	for i := 0; i < maxRepeatedErrorCount+100; i++ {
+	for i := 0; i < maxErrors+100; i++ {
 		var s int
 		err := conn.QueryRowContext(ctx, "SELECT 1").Scan(&s)
 		if err != nil {
@@ -1459,9 +1462,9 @@ func TestConnServerAbortsOnRepeatedErrors(t *testing.T) {
 			}
 			if errors.Is(err, driver.ErrBadConn) {
 				// The server closed the connection, which is what we want!
-				require.GreaterOrEqualf(t, i, maxRepeatedErrorCount,
+				require.GreaterOrEqualf(t, i, maxErrors,
 					"the server should have aborted after seeing %d errors",
-					maxRepeatedErrorCount,
+					maxErrors,
 				)
 				return
 			}

pkg/sql/pgwire/server.go

@@ -94,6 +94,15 @@ var logVerboseSessionAuth = settings.RegisterBoolSetting(
 	false,
 	settings.WithPublic)
 
+var maxRepeatedErrorCount = settings.RegisterIntSetting(
+	settings.ApplicationLevel,
+	"sql.pgwire.max_repeated_error_count",
+	"the maximum number of times an error can be received while reading from a "+
+		"network connection before the server aborts the connection",
+	1<<8, // 256
+	settings.PositiveInt,
+)
+
 const (
 	// ErrSSLRequired is returned when a client attempts to connect to a
 	// secure server in cleartext.
@@ -1079,11 +1088,6 @@ func (s *Server) newConn(
 	return c
 }
 
-// maxRepeatedErrorCount is the number of times an error can be received
-// while reading from the network connection before the server decides to give
-// up and abort the connection.
-const maxRepeatedErrorCount = 1 << 8
-
 // serveImpl continuously reads from the network connection and pushes execution
 // instructions into a sql.StmtBuf, from where they'll be processed by a command
 // "processor" goroutine (a connExecutor).
@@ -1455,7 +1459,7 @@ func (s *Server) serveImpl(
 			// 3. we reached an arbitrary threshold of repeated errors.
 			if netutil.IsClosedConnection(err) ||
 				errors.Is(err, context.Canceled) ||
-				repeatedErrorCount > maxRepeatedErrorCount {
+				repeatedErrorCount > int(maxRepeatedErrorCount.Get(&s.execCfg.Settings.SV)) {
 				break
 			}
 		} else {