Files
netbird/shared/relay/client/picker.go
Zoltán Papp 6639f7bf97 [client] Sweep connections on network loss via a shared netevents manager
Losing the last network only flipped the availability state: the dead
management, signal and relay sockets stayed silently connected until their
own timeouts, so the client kept reporting Connected with no network at all.

Introduce client/netevents with a Manager that ties the availability state,
the connection sweeper and the status recorder together, and move the
netstate and netsweep packages under it (netsweep renamed to sweep).
SetNetworkAvailable(false) now also sweeps the registered connections so
their owners redial and the listener reaches the NoNetwork state.

The Android and iOS bindings own a Manager instance and inject it through
the constructors; consumers hold the concrete *Manager whose nil zero value
reports always-online and never sweeps, with interfaces kept only as
parameter contracts. The relay guard settle wait moved into the Manager as
WaitSettled, removing the netevents import from the relay package.
2026-08-24 21:52:00 +02:00

122 lines
3.1 KiB
Go

package client
import (
"context"
"errors"
"fmt"
"sync/atomic"
"time"
log "github.com/sirupsen/logrus"
auth "github.com/netbirdio/netbird/shared/relay/auth/hmac"
)
const (
maxConcurrentServers = 7
defaultConnectionTimeout = 30 * time.Second
)
type connResult struct {
RelayClient *Client
Url string
Err error
}
type ServerPicker struct {
TokenStore *auth.TokenStore
ServerURLs atomic.Value
PeerID string
MTU uint16
ConnectionTimeout time.Duration
TransportFallback *transportFallback
NetEvents NetEvents
}
func (sp *ServerPicker) PickServer(parentCtx context.Context) (*Client, error) {
ctx, cancel := context.WithTimeout(parentCtx, sp.ConnectionTimeout)
defer cancel()
totalServers := len(sp.ServerURLs.Load().([]string))
connResultChan := make(chan connResult, totalServers)
successChan := make(chan connResult, 1)
errChan := make(chan error, 1)
concurrentLimiter := make(chan struct{}, maxConcurrentServers)
log.Debugf("pick server from list: %v", sp.ServerURLs.Load().([]string))
for _, url := range sp.ServerURLs.Load().([]string) {
// todo check if we have a successful connection so we do not need to connect to other servers
concurrentLimiter <- struct{}{}
go func(url string) {
defer func() {
<-concurrentLimiter
}()
sp.startConnection(parentCtx, connResultChan, url)
}(url)
}
go sp.processConnResults(connResultChan, successChan, errChan)
select {
case cr, ok := <-successChan:
if !ok {
return nil, <-errChan
}
log.Infof("chosen home Relay server: %s", cr.Url)
return cr.RelayClient, nil
case <-ctx.Done():
return nil, fmt.Errorf("connect to relay server: %w", ctx.Err())
}
}
func (sp *ServerPicker) startConnection(ctx context.Context, resultChan chan connResult, url string) {
log.Infof("try to connecting to relay server: %s", url)
relayClient := NewClient(url, sp.TokenStore, sp.PeerID, sp.MTU)
relayClient.SetTransportFallback(sp.TransportFallback)
relayClient.netEvents = sp.NetEvents
err := relayClient.Connect(ctx)
resultChan <- connResult{
RelayClient: relayClient,
Url: url,
Err: err,
}
}
func (sp *ServerPicker) processConnResults(resultChan chan connResult, successChan chan connResult, errChan chan error) {
var hasSuccess bool
var errs []error
for numOfResults := 0; numOfResults < cap(resultChan); numOfResults++ {
cr := <-resultChan
if cr.Err != nil {
log.Tracef("failed to connect to Relay server: %s: %v", cr.Url, cr.Err)
errs = append(errs, cr.Err)
continue
}
log.Infof("connected to Relay server: %s", cr.Url)
if hasSuccess {
log.Infof("closing unnecessary Relay connection to: %s", cr.Url)
if err := cr.RelayClient.Close(); err != nil {
log.Errorf("failed to close connection to %s: %v", cr.Url, err)
}
continue
}
hasSuccess = true
successChan <- cr
}
if !hasSuccess {
errChan <- pickErr(errs)
}
close(successChan)
}
// pickErr combines per-server connection failures into a single error.
func pickErr(errs []error) error {
if len(errs) == 0 {
return errors.New("no relay server available")
}
return errors.Join(errs...)
}