Honor Kafka admin connector timeouts with full coverage

- Update pkg/admin/brokerclient.go to expose a requestTimeout helper that returns the configured connector timeout when set (defaulting to defaultRequestTimeout), and use it for alter-reassignment and leader-election requests so user overrides take effect.
- Teach pkg/admin/connector.go to propagate ConnTimeout to whichever dialer path we take, while keeping the 10s default when nothing is supplied.
- Harden scripts/set_up_net_alias.sh with set -euo pipefail, better Linux tooling detection (ifconfig vs ip), idempotent alias creation, and clearer error handling

Author: Lars Grahmann

pkg/admin/brokerclient.go

@@ -15,7 +15,7 @@ import (
 )
 
 const (
-	defaultTimeout = 5 * time.Second
+	defaultRequestTimeout = 5 * time.Second
 
 	// Used for filtering out default configs
 	configSourceUnknown                    int8 = 0
@@ -138,6 +138,14 @@ func NewBrokerAdminClient(
 	return adminClient, nil
 }
 
+func (c *BrokerAdminClient) requestTimeout() time.Duration {
+	if c.config.ConnTimeout > 0 {
+		return c.config.ConnTimeout
+	}
+
+	return defaultRequestTimeout
+}
+
 // GetClusterID gets the ID of the cluster.
 func (c *BrokerAdminClient) GetClusterID(ctx context.Context) (string, error) {
 	resp, err := c.getMetadata(ctx, nil)
@@ -597,7 +605,7 @@ func (c *BrokerAdminClient) AssignPartitions(
 	req := kafka.AlterPartitionReassignmentsRequest{
 		Topic:       topic,
 		Assignments: apiAssignments,
-		Timeout:     defaultTimeout,
+		Timeout:     c.requestTimeout(),
 	}
 	log.Debugf("AlterPartitionReassignments request: %+v", req)
 
@@ -708,7 +716,7 @@ func (c *BrokerAdminClient) RunLeaderElection(
 	req := kafka.ElectLeadersRequest{
 		Topic:      topic,
 		Partitions: partitions,
-		Timeout:    defaultTimeout,
+		Timeout:    c.requestTimeout(),
 	}
 	log.Debugf("ElectLeaders request: %+v", req)
 

pkg/admin/brokerclient_test.go

@@ -44,6 +44,25 @@ func TestBrokerClientControllerID(t *testing.T) {
 	}, fmt.Sprintf("Received %d, Broker Controller ID should be between 1 and 6.", controllerID))
 }
 
+func TestBrokerAdminClientRequestTimeoutDefault(t *testing.T) {
+	client := &BrokerAdminClient{}
+
+	assert.Equal(t, defaultRequestTimeout, client.requestTimeout())
+}
+
+func TestBrokerAdminClientRequestTimeoutOverride(t *testing.T) {
+	customTimeout := 42 * time.Second
+	client := &BrokerAdminClient{
+		config: BrokerAdminClientConfig{
+			ConnectorConfig: ConnectorConfig{
+				ConnTimeout: customTimeout,
+			},
+		},
+	}
+
+	assert.Equal(t, customTimeout, client.requestTimeout())
+}
+
 func TestBrokerClientGetClusterID(t *testing.T) {
 	if !util.CanTestBrokerAdmin() {
 		t.Skip("Skipping because KAFKA_TOPICS_TEST_BROKER_ADMIN is not set")

pkg/admin/connector.go

@@ -73,6 +73,11 @@ func NewConnector(config ConnectorConfig) (*Connector, error) {
 		Config: config,
 	}
 
+	timeout := config.ConnTimeout
+	if timeout == 0 {
+		timeout = 10 * time.Second
+	}
+
 	var mechanismClient sasl.Mechanism
 	var tlsConfig *tls.Config
 	var err error
@@ -143,6 +148,7 @@ func NewConnector(config ConnectorConfig) (*Connector, error) {
 
 	if !config.TLS.Enabled {
 		connector.Dialer = kafka.DefaultDialer
+		connector.Dialer.Timeout = timeout
 		connector.Dialer.SASLMechanism = mechanismClient
 	} else {
 		var certs []tls.Certificate
@@ -184,7 +190,7 @@ func NewConnector(config ConnectorConfig) (*Connector, error) {
 		}
 		connector.Dialer = &kafka.Dialer{
 			SASLMechanism: mechanismClient,
-			Timeout:       10 * time.Second,
+			Timeout:       timeout,
 			TLS:           tlsConfig,
 		}
 	}

pkg/admin/connector_test.go

@@ -0,0 +1,110 @@
+package admin
+
+import (
+	"context"
+	"errors"
+	"net"
+	"testing"
+	"time"
+
+	"github.com/segmentio/kafka-go"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestNewConnectorDefaultTimeout(t *testing.T) {
+	originalTimeout := kafka.DefaultDialer.Timeout
+	t.Cleanup(func() { kafka.DefaultDialer.Timeout = originalTimeout })
+
+	connector, err := NewConnector(
+		ConnectorConfig{
+			BrokerAddr: "localhost:9092",
+		},
+	)
+	require.NoError(t, err)
+
+	assert.Equal(t, 10*time.Second, connector.Dialer.Timeout)
+}
+
+func TestNewConnectorCustomTimeout(t *testing.T) {
+	customTimeout := 3 * time.Second
+
+	connector, err := NewConnector(
+		ConnectorConfig{
+			BrokerAddr:  "localhost:9092",
+			ConnTimeout: customTimeout,
+			TLS: TLSConfig{
+				Enabled:    true,
+				SkipVerify: true,
+			},
+		},
+	)
+	require.NoError(t, err)
+
+	assert.Equal(t, customTimeout, connector.Dialer.Timeout)
+	assert.NotNil(t, connector.Dialer.TLS)
+}
+
+func TestConnectorDialerTimeoutHappyPath(t *testing.T) {
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = listener.Close() })
+
+	acceptErrCh := make(chan error, 1)
+	go func() {
+		conn, err := listener.Accept()
+		if err != nil {
+			acceptErrCh <- err
+			return
+		}
+		acceptErrCh <- nil
+		_ = conn.Close()
+	}()
+
+	connector, err := NewConnector(
+		ConnectorConfig{
+			BrokerAddr:  listener.Addr().String(),
+			ConnTimeout: 100 * time.Millisecond,
+		},
+	)
+	require.NoError(t, err)
+
+	ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+	defer cancel()
+
+	conn, err := connector.Dialer.DialContext(ctx, "tcp", listener.Addr().String())
+	require.NoError(t, err)
+	require.NoError(t, conn.Close())
+
+	select {
+	case err := <-acceptErrCh:
+		require.NoError(t, err)
+	case <-time.After(time.Second):
+		t.Fatal("timed out waiting for listener accept")
+	}
+}
+
+func TestConnectorDialerTimeoutUnhappyPath(t *testing.T) {
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = listener.Close() })
+
+	connector, err := NewConnector(
+		ConnectorConfig{
+			BrokerAddr:  listener.Addr().String(),
+			ConnTimeout: time.Nanosecond,
+		},
+	)
+	require.NoError(t, err)
+
+	_, err = connector.Dialer.DialContext(context.Background(), "tcp", listener.Addr().String())
+	require.Error(t, err)
+
+	var netErr net.Error
+	if errors.As(err, &netErr) {
+		require.True(t, netErr.Timeout(), "expected timeout error, got: %v", err)
+		return
+	}
+
+	require.True(t, errors.Is(err, context.DeadlineExceeded), "expected deadline exceeded, got: %v", err)
+}

scripts/set_up_net_alias.sh

@@ -1,17 +1,39 @@
 #!/bin/bash
 
+set -euo pipefail
+
 ADDR=169.254.123.123
+NETMASK=255.255.255.0
+CIDR=24
 echo "Aliasing $ADDR to localhost..."
 
-UNAME=$(uname -a)
-case "$UNAME" in
-    Linux*) sudo ifconfig lo:0 $ADDR netmask 255.255.255.0 up;;
-    Darwin*) sudo ifconfig lo0 alias $ADDR;;
-    *) exit
+alias_exists_with_ip() {
+    ip addr show dev lo | grep -q "$ADDR"
+}
+
+OS=$(uname -s)
+case "$OS" in
+    Linux*)
+        if command -v ifconfig >/dev/null 2>&1; then
+            sudo ifconfig lo:0 "$ADDR" netmask "$NETMASK" up
+        elif command -v ip >/dev/null 2>&1; then
+            if alias_exists_with_ip; then
+                echo "Alias already present on loopback interface."
+            else
+                sudo ip addr add "$ADDR/$CIDR" dev lo
+            fi
+        else
+            >&2 echo "Neither ifconfig nor ip is available; cannot create alias."
+            exit 1
+        fi
+        ;;
+    Darwin*)
+        sudo ifconfig lo0 alias "$ADDR"
+        ;;
+    *)
+        >&2 echo "Unsupported platform: $OS"
+        exit 1
+        ;;
 esac
 
-if [[ $? != 0 ]]
-then
-    >&2 echo "Unable to create alias"
-    exit 1
-fi
+echo "Alias configured."