Files
netbird/management/internals/modules/reverseproxy/proxy/manager/manager.go
T
mlsmaycon 1b6a9d86a0 [management] Refuse a proxy claiming a host another account's gateway pinned
The bootstrap check refused pinning onto a cluster another account runs,
but nothing held that decision afterwards: a proxy registration consulted
the proxies table alone, so the same host could be claimed by another
account a moment later, or a week later, and the pin it stranded was
immutable. Validating only at bootstrap meant the check was true when it
ran and not after.

Make the claim symmetric. IsClusterAddressAvailable now treats a gateway
pin as what it is — a claim on the host, served by whichever proxy
declares that address — and refuses a proxy from a different account,
since an account-scoped proxy never receives another account's mappings
and so cannot serve the pin it would displace. Both claims are checked in
one place so a caller cannot consult one and forget the other. An account
claiming the address its own gateway is pinned to is the documented
order, not a conflict: pin first, deploy the proxy after.

Shared (NetBird-operated) proxies register without an account and are
unaffected; an account-scoped proxy reaching for a shared cluster address
that accounts are pinned to was already refused by the proxy-row conflict
and now stays refused even if those rows are momentarily gone.

Whichever of the two lands first now wins and the second is refused,
which is what the bootstrap check could not do on its own. A genuinely
concurrent pair can still pass both checks — closing that needs an
invariant spanning the proxies and settings tables, not a wider read.
2026-09-12 12:05:13 +00:00

212 lines
8.0 KiB
Go

package manager
import (
"context"
"time"
log "github.com/sirupsen/logrus"
"go.opentelemetry.io/otel/metric"
"github.com/netbirdio/netbird/management/internals/modules/reverseproxy/proxy"
)
// store defines the interface for proxy persistence operations
type store interface {
SaveProxy(ctx context.Context, p *proxy.Proxy) error
DisconnectProxy(ctx context.Context, proxyID, sessionID string) error
UpdateProxyHeartbeat(ctx context.Context, p *proxy.Proxy) error
GetActiveProxyClusterAddresses(ctx context.Context) ([]string, error)
GetActiveProxyClusterAddressesForAccount(ctx context.Context, accountID string) ([]string, error)
GetProxyClusters(ctx context.Context, accountID string) ([]proxy.Cluster, error)
GetClusterSupportsCustomPorts(ctx context.Context, clusterAddr string) *bool
GetClusterRequireSubdomain(ctx context.Context, clusterAddr string) *bool
GetClusterSupportsCrowdSec(ctx context.Context, clusterAddr string) *bool
GetClusterSupportsPrivate(ctx context.Context, clusterAddr string) *bool
CleanupStaleProxies(ctx context.Context, inactivityDuration time.Duration) error
GetProxyByAccountID(ctx context.Context, accountID string) (*proxy.Proxy, error)
CountProxiesByAccountID(ctx context.Context, accountID string) (int64, error)
IsClusterAddressConflicting(ctx context.Context, clusterAddress, accountID string) (bool, error)
HasGatewayPinnedByOtherAccount(ctx context.Context, host, accountID string) (bool, error)
DeleteAccountCluster(ctx context.Context, clusterAddress, accountID string) error
}
// Manager handles all proxy operations
type Manager struct {
store store
metrics *metrics
}
// NewManager creates a new proxy Manager
func NewManager(store store, meter metric.Meter) (*Manager, error) {
m, err := newMetrics(meter)
if err != nil {
return nil, err
}
return &Manager{
store: store,
metrics: m,
}, nil
}
// Connect registers a new proxy connection in the database.
// capabilities may be nil for old proxies that do not report them.
func (m *Manager) Connect(ctx context.Context, proxyID, sessionID, clusterAddress, ipAddress string, accountID *string, capabilities *proxy.Capabilities) (*proxy.Proxy, error) {
now := time.Now()
var caps proxy.Capabilities
if capabilities != nil {
caps = *capabilities
}
p := &proxy.Proxy{
ID: proxyID,
SessionID: sessionID,
ClusterAddress: clusterAddress,
IPAddress: ipAddress,
AccountID: accountID,
LastSeen: now,
ConnectedAt: &now,
Status: proxy.StatusConnected,
Capabilities: caps,
}
if err := m.store.SaveProxy(ctx, p); err != nil {
log.WithContext(ctx).Errorf("failed to register proxy %s: %v", proxyID, err)
return nil, err
}
log.WithContext(ctx).WithFields(log.Fields{
"proxyID": proxyID,
"sessionID": sessionID,
"clusterAddress": clusterAddress,
"ipAddress": ipAddress,
}).Info("proxy connected")
return p, nil
}
// Disconnect marks a proxy as disconnected in the database.
func (m *Manager) Disconnect(ctx context.Context, proxyID, sessionID string) error {
if err := m.store.DisconnectProxy(ctx, proxyID, sessionID); err != nil {
log.WithContext(ctx).Errorf("failed to disconnect proxy %s session %s: %v", proxyID, sessionID, err)
return err
}
log.WithContext(ctx).WithFields(log.Fields{
"proxyID": proxyID,
"sessionID": sessionID,
}).Info("proxy disconnected")
return nil
}
// Heartbeat updates the proxy's last seen timestamp.
func (m *Manager) Heartbeat(ctx context.Context, p *proxy.Proxy) error {
if err := m.store.UpdateProxyHeartbeat(ctx, p); err != nil {
log.WithContext(ctx).Debugf("failed to update proxy %s heartbeat: %v", p.ID, err)
return err
}
log.WithContext(ctx).Tracef("updated heartbeat for proxy %s session %s", p.ID, p.SessionID)
m.metrics.IncrementProxyHeartbeatCount()
return nil
}
// GetActiveClusterAddresses returns all unique cluster addresses for active proxies
func (m *Manager) GetActiveClusterAddresses(ctx context.Context) ([]string, error) {
addresses, err := m.store.GetActiveProxyClusterAddresses(ctx)
if err != nil {
log.WithContext(ctx).Errorf("failed to get active proxy cluster addresses: %v", err)
return nil, err
}
return addresses, nil
}
// ClusterSupportsCustomPorts returns whether any active proxy in the cluster
// supports custom ports. Returns nil when no proxy has reported capabilities.
func (m Manager) ClusterSupportsCustomPorts(ctx context.Context, clusterAddr string) *bool {
return m.store.GetClusterSupportsCustomPorts(ctx, clusterAddr)
}
// ClusterRequireSubdomain returns whether any active proxy in the cluster
// requires a subdomain. Returns nil when no proxy has reported capabilities.
func (m Manager) ClusterRequireSubdomain(ctx context.Context, clusterAddr string) *bool {
return m.store.GetClusterRequireSubdomain(ctx, clusterAddr)
}
// ClusterSupportsCrowdSec returns whether all active proxies in the cluster
// have CrowdSec configured (unanimous). Returns nil when no proxy has reported capabilities.
func (m Manager) ClusterSupportsCrowdSec(ctx context.Context, clusterAddr string) *bool {
return m.store.GetClusterSupportsCrowdSec(ctx, clusterAddr)
}
// ClusterSupportsPrivate reports whether any active proxy claims the private capability (nil = unreported).
func (m Manager) ClusterSupportsPrivate(ctx context.Context, clusterAddr string) *bool {
return m.store.GetClusterSupportsPrivate(ctx, clusterAddr)
}
// CleanupStale removes proxies that haven't sent heartbeat in the specified duration
func (m *Manager) CleanupStale(ctx context.Context, inactivityDuration time.Duration) error {
if err := m.store.CleanupStaleProxies(ctx, inactivityDuration); err != nil {
log.WithContext(ctx).Errorf("failed to cleanup stale proxies: %v", err)
return err
}
return nil
}
func (m *Manager) GetActiveClusterAddressesForAccount(ctx context.Context, accountID string) ([]string, error) {
addresses, err := m.store.GetActiveProxyClusterAddressesForAccount(ctx, accountID)
if err != nil {
log.WithContext(ctx).Errorf("failed to get active proxy cluster addresses for account %s: %v", accountID, err)
return nil, err
}
return addresses, nil
}
func (m *Manager) GetAccountProxy(ctx context.Context, accountID string) (*proxy.Proxy, error) {
return m.store.GetProxyByAccountID(ctx, accountID)
}
func (m *Manager) CountAccountProxies(ctx context.Context, accountID string) (int64, error) {
return m.store.CountProxiesByAccountID(ctx, accountID)
}
// IsClusterAddressAvailable reports whether the account may claim this cluster
// address.
//
// Two kinds of claim make an address unavailable, and both are checked here so
// that no caller can consult one and forget the other. A proxy row is the
// obvious one. An agent network gateway pinned to the address by another
// account is the second: that pin is immutable and is served by whichever
// proxy declares the address, so letting a proxy from a different account take
// it strands the pin — an account-scoped proxy never receives another
// account's mappings. An account claiming an address its own gateway is pinned
// to is the intended order, not a conflict: pin first, deploy the proxy after.
func (m *Manager) IsClusterAddressAvailable(ctx context.Context, clusterAddress, accountID string) (bool, error) {
conflicting, err := m.store.IsClusterAddressConflicting(ctx, clusterAddress, accountID)
if err != nil {
return false, err
}
if conflicting {
return false, nil
}
pinned, err := m.store.HasGatewayPinnedByOtherAccount(ctx, clusterAddress, accountID)
if err != nil {
return false, err
}
if pinned {
log.WithContext(ctx).Infof("cluster address %s is pinned as another account's agent network gateway, refusing claim by account %s", clusterAddress, accountID)
return false, nil
}
return true, nil
}
func (m *Manager) DeleteAccountCluster(ctx context.Context, clusterAddress, accountID string) error {
if err := m.store.DeleteAccountCluster(ctx, clusterAddress, accountID); err != nil {
log.WithContext(ctx).Errorf("failed to delete cluster %s for account %s: %v", clusterAddress, accountID, err)
return err
}
return nil
}