qemu: make saveSandbox wait for migration completion

Then we can remove the arbitrary sleep waiting for migration completion when creating a tempalte vm. Fixes: #728 Signed-off-by: Peng Tao <[email protected]>
kata-containers · Oct 9, 2018 · eb77a41 · eb77a41
1 parent c3cfe82
commit eb77a41
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 5 deletions.
diff --git a/virtcontainers/factory/template/template.go b/virtcontainers/factory/template/template.go
@@ -25,7 +25,6 @@ type template struct {
 
 var templateProxyType = vc.KataBuiltInProxyType
 var templateWaitForAgent = 2 * time.Second
-var templateWaitForMigration = 1 * time.Second
 
 // Fetch finds and returns a pre-built template factory.
 // TODO: save template metadata and fetch from storage.
@@ -145,9 +144,6 @@ func (t *template) createTemplateVM(ctx context.Context) error {
 		return err
 	}
 
-	// qemu QMP does not wait for migration to finish...
-	time.Sleep(templateWaitForMigration)
-
 	return nil
 }
 

diff --git a/virtcontainers/factory/template/template_test.go b/virtcontainers/factory/template/template_test.go
@@ -20,7 +20,6 @@ import (
 func TestTemplateFactory(t *testing.T) {
 	assert := assert.New(t)
 
-	templateWaitForMigration = 1 * time.Microsecond
 	templateWaitForAgent = 1 * time.Microsecond
 
 	testDir, _ := ioutil.TempDir("", "vmfactory-tmp-")

diff --git a/virtcontainers/qemu.go b/virtcontainers/qemu.go
@@ -76,6 +76,7 @@ const (
 	qmpCapErrMsg                      = "Failed to negoatiate QMP capabilities"
 	qmpCapMigrationBypassSharedMemory = "bypass-shared-memory"
 	qmpExecCatCmd                     = "exec:cat"
+	qmpMigrationWaitTimeout           = 5 * time.Second
 
 	scsiControllerID = "scsi0"
 	rngID            = "rng0"
@@ -1181,6 +1182,29 @@ func (q *qemu) saveSandbox() error {
 		return err
 	}
 
+	t := time.NewTimer(qmpMigrationWaitTimeout)
+	defer t.Stop()
+	for {
+		status, err := q.qmpMonitorCh.qmp.ExecuteQueryMigration(q.qmpMonitorCh.ctx)
+		if err != nil {
+			q.Logger().WithError(err).Error("failed to query migration status")
+			return err
+		}
+		if status.Status == "completed" {
+			break
+		}
+
+		select {
+		case <-t.C:
+			q.Logger().WithField("migration-status", status).Error("timeout waiting for qemu migration")
+			return fmt.Errorf("timed out after %d seconds waiting for qemu migration", qmpMigrationWaitTimeout)
+		default:
+			// migration in progress
+			q.Logger().WithField("migration-status", status).Debug("migration in progress")
+			time.Sleep(100 * time.Millisecond)
+		}
+	}
+
 	return nil
 }