Browse Source

fix: canonical replica address resolution (CP13-2)

ReplicaReceiver.DataAddr()/CtrlAddr() now return canonical ip:port
instead of raw listener addresses that may be wildcard (:port,
0.0.0.0:port, [::]:port).

New canonicalizeListenerAddr() resolves wildcard IPs using the
provided advertised host (from VS listen address). Falls back to
outbound-IP detection when no advertised host is available.

NewReplicaReceiver accepts optional advertisedHost parameter for
multi-NIC correctness. In production, the assignment path already
provides canonical addresses; this fix ensures test patterns with
:0 bind also produce routable addresses.

7 new tests. TestBug3_ReplicaAddr_MustBeIPPort_WildcardBind flips
from FAIL to PASS.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feature/sw-block
Ping Qiu 6 days ago
parent
commit
4f3edffb0a
  1. 53
      weed/storage/blockvol/net_util.go
  2. 73
      weed/storage/blockvol/net_util_test.go
  3. 21
      weed/storage/blockvol/replica_apply.go

53
weed/storage/blockvol/net_util.go

@ -0,0 +1,53 @@
package blockvol
import (
"net"
"strconv"
)
// canonicalizeListenerAddr resolves wildcard listener addresses to a routable
// ip:port string using the provided advertised host as the preferred IP.
//
// If the listener bound to a wildcard (":0", "0.0.0.0:port", "[::]:port"),
// the returned address uses advertisedHost instead of the wildcard.
//
// If advertisedHost is empty, falls back to preferredOutboundIP() as a
// best-effort guess. On multi-NIC hosts, the advertised host should always
// be provided to avoid selecting the wrong network interface.
func canonicalizeListenerAddr(addr net.Addr, advertisedHost string) string {
tcpAddr, ok := addr.(*net.TCPAddr)
if !ok {
return addr.String()
}
ip := tcpAddr.IP
if ip != nil && !ip.IsUnspecified() {
// Already bound to a specific IP — return as-is.
return net.JoinHostPort(ip.String(), strconv.Itoa(tcpAddr.Port))
}
// Wildcard bind — use advertised host or fallback.
host := advertisedHost
if host == "" {
host = preferredOutboundIP()
}
if host == "" {
// Last resort: return raw address (will be ":port").
return addr.String()
}
return net.JoinHostPort(host, strconv.Itoa(tcpAddr.Port))
}
// preferredOutboundIP returns the machine's preferred outbound IP as a string.
// Uses the standard Go pattern: dial a UDP socket (no data sent), read the
// local address. Returns "" if discovery fails.
//
// This is a fallback — callers should prefer an explicitly configured
// advertised host when available.
func preferredOutboundIP() string {
conn, err := net.Dial("udp", "8.8.8.8:80")
if err != nil {
return ""
}
defer conn.Close()
localAddr := conn.LocalAddr().(*net.UDPAddr)
return localAddr.IP.String()
}

73
weed/storage/blockvol/net_util_test.go

@ -0,0 +1,73 @@
package blockvol
import (
"net"
"strings"
"testing"
)
func TestCanonicalizeAddr_WildcardIPv4_UsesAdvertised(t *testing.T) {
addr := &net.TCPAddr{IP: net.IPv4zero, Port: 5099}
result := canonicalizeListenerAddr(addr, "192.168.1.184")
if result != "192.168.1.184:5099" {
t.Fatalf("expected 192.168.1.184:5099, got %q", result)
}
}
func TestCanonicalizeAddr_WildcardIPv6_UsesAdvertised(t *testing.T) {
addr := &net.TCPAddr{IP: net.IPv6zero, Port: 5099}
result := canonicalizeListenerAddr(addr, "10.0.0.3")
if result != "10.0.0.3:5099" {
t.Fatalf("expected 10.0.0.3:5099, got %q", result)
}
}
func TestCanonicalizeAddr_NilIP_UsesAdvertised(t *testing.T) {
addr := &net.TCPAddr{IP: nil, Port: 5099}
result := canonicalizeListenerAddr(addr, "192.168.1.184")
if result != "192.168.1.184:5099" {
t.Fatalf("expected 192.168.1.184:5099, got %q", result)
}
}
func TestCanonicalizeAddr_AlreadyCanonical_Unchanged(t *testing.T) {
addr := &net.TCPAddr{IP: net.ParseIP("192.168.1.5"), Port: 5099}
result := canonicalizeListenerAddr(addr, "10.0.0.1")
if result != "192.168.1.5:5099" {
t.Fatalf("expected 192.168.1.5:5099 (unchanged), got %q", result)
}
}
func TestCanonicalizeAddr_Loopback_Unchanged(t *testing.T) {
addr := &net.TCPAddr{IP: net.ParseIP("127.0.0.1"), Port: 3260}
result := canonicalizeListenerAddr(addr, "192.168.1.184")
if result != "127.0.0.1:3260" {
t.Fatalf("expected 127.0.0.1:3260 (loopback intentional), got %q", result)
}
}
func TestCanonicalizeAddr_NoAdvertised_FallsBackToOutbound(t *testing.T) {
addr := &net.TCPAddr{IP: net.IPv4zero, Port: 5099}
result := canonicalizeListenerAddr(addr, "")
// Should not be wildcard.
if strings.HasPrefix(result, "0.0.0.0:") || strings.HasPrefix(result, "[::]:") || strings.HasPrefix(result, ":") {
t.Fatalf("fallback should produce routable addr, got %q", result)
}
if !strings.Contains(result, ":5099") {
t.Fatalf("port should be preserved, got %q", result)
}
}
func TestPreferredOutboundIP_NotEmpty(t *testing.T) {
ip := preferredOutboundIP()
if ip == "" {
t.Skip("no network interface available")
}
parsed := net.ParseIP(ip)
if parsed == nil {
t.Fatalf("not a valid IP: %q", ip)
}
if parsed.IsUnspecified() {
t.Fatalf("returned unspecified IP: %q", ip)
}
}

21
weed/storage/blockvol/replica_apply.go

@ -20,6 +20,7 @@ var (
type ReplicaReceiver struct {
vol *BlockVol
barrierTimeout time.Duration
advertisedHost string // canonical IP for address reporting; empty = auto-detect
mu sync.Mutex
receivedLSN uint64
@ -38,7 +39,10 @@ type ReplicaReceiver struct {
const defaultBarrierTimeout = 5 * time.Second
// NewReplicaReceiver creates and starts listening on the data and control ports.
func NewReplicaReceiver(vol *BlockVol, dataAddr, ctrlAddr string) (*ReplicaReceiver, error) {
// advertisedHost is the canonical IP for this server (from VS listen addr or
// heartbeat identity). If empty, DataAddr()/CtrlAddr() will fall back to
// outbound-IP detection. On multi-NIC hosts, always provide advertisedHost.
func NewReplicaReceiver(vol *BlockVol, dataAddr, ctrlAddr string, advertisedHost ...string) (*ReplicaReceiver, error) {
dataLn, err := net.Listen("tcp", dataAddr)
if err != nil {
return nil, fmt.Errorf("replica: listen data %s: %w", dataAddr, err)
@ -49,9 +53,14 @@ func NewReplicaReceiver(vol *BlockVol, dataAddr, ctrlAddr string) (*ReplicaRecei
return nil, fmt.Errorf("replica: listen ctrl %s: %w", ctrlAddr, err)
}
var advHost string
if len(advertisedHost) > 0 {
advHost = advertisedHost[0]
}
r := &ReplicaReceiver{
vol: vol,
barrierTimeout: defaultBarrierTimeout,
advertisedHost: advHost,
dataListener: dataLn,
ctrlListener: ctrlLn,
stopCh: make(chan struct{}),
@ -293,12 +302,14 @@ func (r *ReplicaReceiver) ReceivedLSN() uint64 {
return r.receivedLSN
}
// DataAddr returns the data listener's address (useful for tests with :0 ports).
// DataAddr returns the data listener's canonical address (ip:port).
// Wildcard listener addresses are resolved using the advertised host
// or outbound-IP fallback.
func (r *ReplicaReceiver) DataAddr() string {
return r.dataListener.Addr().String()
return canonicalizeListenerAddr(r.dataListener.Addr(), r.advertisedHost)
}
// CtrlAddr returns the control listener's address.
// CtrlAddr returns the control listener's canonical address (ip:port).
func (r *ReplicaReceiver) CtrlAddr() string {
return r.ctrlListener.Addr().String()
return canonicalizeListenerAddr(r.ctrlListener.Addr(), r.advertisedHost)
}
Loading…
Cancel
Save