Automated cherry pick of #15475: fix(host): convert migrate set downtime value to float64 (#15476)

* fix(host): convert migrate set downtime value to float64

* fix(host): live migration optimize

Qemu will send `STOP` event on start last time migration, but this
moment migration not completed. So we need wait magration complete.
Configurable Auto converge max cpu throttle.

pkg/hostman/guestman/guesttasks.go

@@ -1130,6 +1130,7 @@ func (s *SGuestLiveMigrateTask) startMigrateStatusCheck(res string) {
 func (s *SGuestLiveMigrateTask) onGetMigrateStats(stats *monitor.MigrationInfo, err error) {
 	if err != nil {
 		log.Errorf("%s get migrate stats failed %s", s.GetName(), err)
+		s.migrateFailed(fmt.Sprintf("%s get migrate stats failed %s", s.GetName(), err))
 		return
 	}
 	s.onGetMigrateStatus(stats)
@@ -1177,8 +1178,7 @@ func (s *SGuestLiveMigrateTask) onGetMigrateStatus(stats *monitor.MigrationInfo)
 				return
 			}
 
-			if *stats.CPUThrottlePercentage < 99 {
-				// TODO: configureable tolerate cpu throttle percentage
+			if *stats.CPUThrottlePercentage < options.HostOptions.LiveMigrateCpuThrottleMax {
 				return
 			}
 
@@ -1201,7 +1201,7 @@ func (s *SGuestLiveMigrateTask) onGetMigrateStatus(stats *monitor.MigrationInfo)
 						log.Errorf("failed set migrate downtime %s", res)
 					}
 				}
-				s.Monitor.MigrateSetDowntime(float32(*stats.ExpectedDowntime/1000.0), cb)
+				s.Monitor.MigrateSetDowntime(float64(*stats.ExpectedDowntime)/1000.0, cb)
 			}
 		}
 	}
@@ -1216,6 +1216,25 @@ func (s *SGuestLiveMigrateTask) onMigrateStartPostcopy(res string) {
 	}
 }
 
+func (s *SGuestLiveMigrateTask) onMigrateReceivedStopEvent() {
+	s.Monitor.GetMigrateStats(func(stats *monitor.MigrationInfo, err error) {
+		if err != nil {
+			log.Errorf("%s get migrate stats failed %s", s.GetName(), err)
+			return
+		}
+
+		switch *stats.Status {
+		case "completed":
+			s.migrateComplete(jsonutils.Marshal(stats))
+		case "failed", "cancelled":
+			s.migrateFailed(fmt.Sprintf("Query migrate got status: %s", *stats.Status))
+		case "active":
+			time.Sleep(10 * time.Millisecond)
+			s.onMigrateReceivedStopEvent()
+		}
+	})
+}
+
 func (s *SGuestLiveMigrateTask) migrateComplete(stats jsonutils.JSONObject) {
 	s.MigrateTask = nil
 	if s.c != nil {
@@ -1229,6 +1248,7 @@ func (s *SGuestLiveMigrateTask) migrateComplete(stats jsonutils.JSONObject) {
 		res.Set("migration_info", stats)
 	}
 	hostutils.TaskComplete(s.ctx, res)
+	hostutils.UpdateServerProgress(context.Background(), s.Id, 0.0, 0)
 }
 
 func (s *SGuestLiveMigrateTask) migrateFailed(msg string) {

pkg/hostman/guestman/qemu-kvm.go

@@ -756,10 +756,8 @@ func (s *SKVMGuestInstance) onReceiveQMPEvent(event *monitor.Event) {
 		s.eventGuestPaniced(event)
 	case event.Event == `"STOP"`:
 		if s.MigrateTask != nil {
-			// migrating complete
-			s.MigrateTask.migrateComplete(nil)
+			s.MigrateTask.onMigrateReceivedStopEvent()
 		}
-		hostutils.UpdateServerProgress(context.Background(), s.Id, 0.0, 0)
 	}
 }
 

pkg/hostman/monitor/hmp.go

@@ -293,7 +293,7 @@ func (m *HmpMonitor) DeviceAdd(dev string, params map[string]string, callback St
 	m.Query(fmt.Sprintf("device_add %s,%s", dev, strings.Join(paramsKvs, ",")), callback)
 }
 
-func (m *HmpMonitor) MigrateSetDowntime(dtSec float32, callback StringCallback) {
+func (m *HmpMonitor) MigrateSetDowntime(dtSec float64, callback StringCallback) {
 	m.Query(fmt.Sprintf("migrate_set_downtime %f", dtSec), callback)
 }
 

pkg/hostman/monitor/monitor.go

@@ -224,7 +224,7 @@ type Monitor interface {
 	BlockReopenImage(drive, newImagePath, format string, cb StringCallback)
 	SnapshotBlkdev(drive, newImagePath, format string, reuse bool, cb StringCallback)
 
-	MigrateSetDowntime(dtSec float32, callback StringCallback)
+	MigrateSetDowntime(dtSec float64, callback StringCallback)
 	MigrateSetCapability(capability, state string, callback StringCallback)
 	MigrateSetParameter(key string, val interface{}, callback StringCallback)
 	MigrateIncoming(address string, callback StringCallback)

pkg/hostman/monitor/qmp.go

@@ -611,7 +611,7 @@ func (m *QmpMonitor) DeviceAdd(dev string, params map[string]string, callback St
 	// m.Query(cmd, cb)
 }
 
-func (m *QmpMonitor) MigrateSetDowntime(dtSec float32, callback StringCallback) {
+func (m *QmpMonitor) MigrateSetDowntime(dtSec float64, callback StringCallback) {
 	m.HumanMonitorCommand(fmt.Sprintf("migrate_set_downtime %f", dtSec), callback)
 }
 

pkg/hostman/options/options.go

@@ -28,7 +28,8 @@ type SHostBaseOptions struct {
 
 	DisableSecurityGroup bool `help:"disable security group" default:"false"`
 
-	HostCpuPassthrough bool `default:"true" help:"if it is true, set qemu cpu type as -cpu host, otherwise, qemu64. default is true"`
+	HostCpuPassthrough        bool  `default:"true" help:"if it is true, set qemu cpu type as -cpu host, otherwise, qemu64. default is true"`
+	LiveMigrateCpuThrottleMax int64 `default:"99" help:"live migrate auto converge cpu throttle max"`
 
 	DefaultQemuVersion string `help:"Default qemu version" default:"4.2.0"`