replication: handle context cancellation errors as GlobalError

replication/fsrep/fsfsm.go

@@ -82,6 +82,7 @@ const (
 type Error interface {
 	error
 	Temporary() bool
+	ContextErr() bool
 	LocalToFS() bool
 }
 
@@ -187,6 +188,8 @@ func (e *ReplicationConflictError) Error() string { return fmt.Sprintf("permanen
 
 func (e *ReplicationConflictError) LocalToFS() bool { return true }
 
+func (e *ReplicationConflictError) ContextErr() bool { return false }
+
 func NewReplicationConflictError(fs string, err error) *Replication {
 	return &Replication{
 		state: Completed,
@@ -332,6 +335,16 @@ func (e StepError) LocalToFS() bool {
 	return true // conservative approximation: we'd like to check for specific errors returned over RPC here...
 }
 
+func (e StepError) ContextErr() bool {
+	switch e.err {
+	case context.Canceled:
+		return true
+	case context.DeadlineExceeded:
+		return true
+	}
+	return false
+}
+
 func (fsr *Replication) Report() *Report {
 	fsr.lock.Lock()
 	defer fsr.lock.Unlock()

replication/mainfsm.go

@@ -459,7 +459,14 @@ func stateWorking(ctx context.Context, ka *watchdog.KeepAlive, sender Sender, re
 	}).rsf()
 
 	if err != nil {
-		if err.LocalToFS() {
+		if err.ContextErr() && ctx.Err() != nil {
+			getLogger(ctx).WithError(err).
+				Info("filesystem replication was cancelled")
+			u(func(r*Replication) {
+				r.err = GlobalError{Err: err, Temporary: false}
+				r.state = PermanentError
+			})
+		} else if err.LocalToFS() {
 			getLogger(ctx).WithError(err).
 				Error("filesystem replication encountered a filesystem-specific error")
 			// we stay in this state and let the queuing logic above de-prioritize this failing FS