Changes to add state and retry logic to rsync

- Adds retry watcher and related functionality.
- Adds code to update, get the status from appcontext.
- Adds logic to handle state transition during terminate.

Issue-ID: MULTICLOUD-1005
Signed-off-by: Manjunath Ranganathaiah <manjunath.ranganathaiah@intel.com>
Change-Id: I2ed76efd9d8b6f40fec547bbe8b7d8a86f69ce07

1 files changed, 359 insertions, 22 deletions

diff --git a/src/rsync/pkg/context/context.go b/src/rsync/pkg/context/context.go
index f77482e6..4b886ec7 100644
--- a/src/rsync/pkg/context/context.go
+++ b/src/rsync/pkg/context/context.go
@@ -21,6 +21,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"sync"
 	"time"
 
 	"github.com/onap/multicloud-k8s/src/orchestrator/pkg/appcontext"
@@ -36,7 +37,9 @@ import (
 )
 
 type CompositeAppContext struct {
-	cid interface{}
+	cid   interface{}
+	chans []chan bool
+	mutex sync.Mutex
 }
 
 func getRes(ac appcontext.AppContext, name string, app string, cluster string) ([]byte, interface{}, error) {
@@ -144,26 +147,150 @@ func instantiateResource(ac appcontext.AppContext, c *kubeclient.Client, name st
 	return nil
 }
 
+func updateResourceStatus(ac appcontext.AppContext, resState resourcestatus.ResourceStatus, app string, cluster string, aov map[string][]string) error {
+
+	for _, res := range aov["resorder"] {
+
+		rh, err := ac.GetResourceHandle(app, cluster, res)
+		if err != nil {
+			return err
+		}
+		sh, err := ac.GetLevelHandle(rh, "status")
+		if err != nil {
+			return err
+		}
+
+		s, err := ac.GetValue(sh)
+		if err != nil {
+			return err
+		}
+		rStatus := resourcestatus.ResourceStatus{}
+		js, err := json.Marshal(s)
+		if err != nil {
+			return err
+		}
+		err = json.Unmarshal(js, &rStatus)
+		if err != nil {
+			return err
+		}
+		// no need to update a status that has reached a 'done' status
+		if rStatus.Status == resourcestatus.RsyncStatusEnum.Deleted ||
+			rStatus.Status == resourcestatus.RsyncStatusEnum.Applied ||
+			rStatus.Status == resourcestatus.RsyncStatusEnum.Failed {
+			continue
+		}
+
+		err = ac.UpdateStatusValue(sh, resState)
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+
+}
+
+// return true if all resources have reached a 'done' status - e.g. Applied, Deleted or Failed
+func allResourcesDone(ac appcontext.AppContext, app string, cluster string, aov map[string][]string) bool {
+
+	for _, res := range aov["resorder"] {
+
+		rh, err := ac.GetResourceHandle(app, cluster, res)
+		if err != nil {
+			return false
+		}
+		sh, err := ac.GetLevelHandle(rh, "status")
+		if err != nil {
+			return false
+		}
+
+		s, err := ac.GetValue(sh)
+		if err != nil {
+			return false
+		}
+		rStatus := resourcestatus.ResourceStatus{}
+		js, err := json.Marshal(s)
+		if err != nil {
+			return false
+		}
+		err = json.Unmarshal(js, &rStatus)
+		if err != nil {
+			return false
+		}
+		if rStatus.Status != resourcestatus.RsyncStatusEnum.Deleted &&
+			rStatus.Status != resourcestatus.RsyncStatusEnum.Applied &&
+			rStatus.Status != resourcestatus.RsyncStatusEnum.Failed {
+			return false
+		}
+	}
+
+	return true
+
+}
+
 // Wait for 2 secs
 const waitTime = 2
 
-func waitForClusterReady(c *kubeclient.Client, cluster string) error {
+func waitForClusterReady(instca *CompositeAppContext, ac appcontext.AppContext, c *kubeclient.Client, appname string, cluster string, aov map[string][]string) error {
+
+	forceDone := false
+	resStateUpdated := false
+	ch := addChan(instca)
+
+	rch := make(chan error, 1)
+	checkReachable := func() {
+		err := c.IsReachable()
+		rch <- err
+	}
+
+	go checkReachable()
+Loop:
 	for {
-		if err := c.IsReachable(); err != nil {
-			// TODO: Add more realistic error checking
-			// TODO: Add Incremental wait logic here
-			time.Sleep(waitTime * time.Second)
-		} else {
+		select {
+		case rerr := <-rch:
+			if rerr == nil {
+				break Loop
+			} else {
+				logutils.Info("Cluster is not reachable - keep trying::", logutils.Fields{"cluster": cluster})
+				go checkReachable()
+			}
+		case <-ch:
+			statusFailed := resourcestatus.ResourceStatus{
+				Status: resourcestatus.RsyncStatusEnum.Failed,
+			}
+			err := updateResourceStatus(ac, statusFailed, appname, cluster, aov)
+			if err != nil {
+				deleteChan(instca, ch)
+				return err
+			}
+			forceDone = true
+			break Loop
+		case <-time.After(waitTime * time.Second):
+			// on first timeout - cluster is apparently not reachable, update resources in
+			// this group to 'Retrying'
+			if !resStateUpdated {
+				statusRetrying := resourcestatus.ResourceStatus{
+					Status: resourcestatus.RsyncStatusEnum.Retrying,
+				}
+				err := updateResourceStatus(ac, statusRetrying, appname, cluster, aov)
+				if err != nil {
+					deleteChan(instca, ch)
+					return err
+				}
+				resStateUpdated = true
+			}
 			break
 		}
 	}
-	logutils.Info("Cluster is reachable::", logutils.Fields{
-		"cluster": cluster,
-	})
+
+	deleteChan(instca, ch)
+	if forceDone {
+		return pkgerrors.Errorf("Termination of rsync cluster retry: " + cluster)
+	}
 	return nil
 }
 
-// initializeResourceStatus sets the initial status of every resource appropriately based on the state of the AppContext
+// initializeAppContextStatus sets the initial status of every resource appropriately based on the state of the AppContext
 func initializeAppContextStatus(ac appcontext.AppContext, acStatus appcontext.AppContextStatus) error {
 	h, err := ac.GetCompositeAppHandle()
 	if err != nil {
@@ -320,12 +447,18 @@ func updateEndingAppContextStatus(ac appcontext.AppContext, handle interface{},
 	js, _ := json.Marshal(s)
 	json.Unmarshal(js, &acStatus)
 
-	if failure {
-		acStatus.Status = appcontext.AppContextStatusEnum.Failed
-	} else if acStatus.Status == appcontext.AppContextStatusEnum.Instantiating {
-		acStatus.Status = appcontext.AppContextStatusEnum.Instantiated
+	if acStatus.Status == appcontext.AppContextStatusEnum.Instantiating {
+		if failure {
+			acStatus.Status = appcontext.AppContextStatusEnum.InstantiateFailed
+		} else {
+			acStatus.Status = appcontext.AppContextStatusEnum.Instantiated
+		}
 	} else if acStatus.Status == appcontext.AppContextStatusEnum.Terminating {
-		acStatus.Status = appcontext.AppContextStatusEnum.Terminated
+		if failure {
+			acStatus.Status = appcontext.AppContextStatusEnum.TerminateFailed
+		} else {
+			acStatus.Status = appcontext.AppContextStatusEnum.Terminated
+		}
 	} else {
 		return pkgerrors.Errorf("Invalid AppContextStatus %v", acStatus)
 	}
@@ -337,20 +470,197 @@ func updateEndingAppContextStatus(ac appcontext.AppContext, handle interface{},
 	return nil
 }
 
+func getAppContextStatus(ac appcontext.AppContext) (*appcontext.AppContextStatus, error) {
+
+	h, err := ac.GetCompositeAppHandle()
+	if err != nil {
+		return nil, err
+	}
+	sh, err := ac.GetLevelHandle(h, "status")
+	if err != nil {
+		return nil, err
+	}
+	s, err := ac.GetValue(sh)
+	if err != nil {
+		return nil, err
+	}
+	acStatus := appcontext.AppContextStatus{}
+	js, _ := json.Marshal(s)
+	json.Unmarshal(js, &acStatus)
+
+	return &acStatus, nil
+
+}
+
 type fn func(ac appcontext.AppContext, client *kubeclient.Client, res string, app string, cluster string, label string) error
 
 type statusfn func(client *kubeclient.Client, app string, cluster string, label string) error
 
-func applyFnComApp(cid interface{}, acStatus appcontext.AppContextStatus, f fn, sfn statusfn, breakonError bool) error {
-	con := connector.Init(cid)
+func addChan(instca *CompositeAppContext) chan bool {
+
+	instca.mutex.Lock()
+	c := make(chan bool)
+	instca.chans = append(instca.chans, c)
+	instca.mutex.Unlock()
+
+	return c
+}
+
+func deleteChan(instca *CompositeAppContext, c chan bool) error {
+
+	var i int
+	instca.mutex.Lock()
+	for i = 0; i < len(instca.chans); i++ {
+		if instca.chans[i] == c {
+			break
+		}
+	}
+
+	if i == len(instca.chans) {
+		instca.mutex.Unlock()
+		return pkgerrors.Errorf("Given channel was not found:")
+	}
+	instca.chans[i] = instca.chans[len(instca.chans)-1]
+	instca.chans = instca.chans[:len(instca.chans)-1]
+	instca.mutex.Unlock()
+
+	return nil
+}
+
+func waitForDone(ac appcontext.AppContext) {
+	count := 0
+	for {
+		time.Sleep(1 * time.Second)
+		count++
+		if count == 60*60 {
+			logutils.Info("Wait for done watcher running..", logutils.Fields{})
+			count = 0
+		}
+		acStatus, _ := getAppContextStatus(ac)
+		if acStatus.Status == appcontext.AppContextStatusEnum.Instantiated ||
+			acStatus.Status == appcontext.AppContextStatusEnum.InstantiateFailed {
+			return
+		}
+	}
+	return
+}
+
+func kickoffRetryWatcher(instca *CompositeAppContext, ac appcontext.AppContext, acStatus appcontext.AppContextStatus, wg *errgroup.Group) {
+
+	wg.Go(func() error {
+
+		var count int
+
+		count = 0
+		for {
+			time.Sleep(1 * time.Second)
+			count++
+			if count == 60*60 {
+				logutils.Info("Retry watcher running..", logutils.Fields{})
+				count = 0
+			}
+
+			cStatus, err := getAppContextStatus(ac)
+			if err != nil {
+				logutils.Error("Failed to get the app context status", logutils.Fields{
+					"error": err,
+				})
+				return err
+			}
+			flag, err := getAppContextFlag(ac)
+			if err != nil {
+				logutils.Error("Failed to get the stop flag", logutils.Fields{
+					"error": err,
+				})
+				return err
+			} else {
+				if flag == true {
+					instca.mutex.Lock()
+					for i := 0; i < len(instca.chans); i++ {
+						instca.chans[i] <- true
+						logutils.Info("kickoffRetryWatcher - send an exit message", logutils.Fields{})
+					}
+					instca.mutex.Unlock()
+					break
+				}
+			}
+			if acStatus.Status == appcontext.AppContextStatusEnum.Instantiating {
+				if cStatus.Status == appcontext.AppContextStatusEnum.Instantiated ||
+					cStatus.Status == appcontext.AppContextStatusEnum.InstantiateFailed {
+					break
+				}
+			} else {
+				if cStatus.Status == appcontext.AppContextStatusEnum.Terminated ||
+					cStatus.Status == appcontext.AppContextStatusEnum.TerminateFailed {
+					break
+				}
+			}
+
+		}
+		return nil
+	})
+
+}
+
+func getAppContextFlag(ac appcontext.AppContext) (bool, error) {
+	h, err := ac.GetCompositeAppHandle()
+	if err != nil {
+		return false, err
+	}
+	sh, err := ac.GetLevelHandle(h, "stopflag")
+	if sh == nil {
+		return false, err
+	} else {
+		v, err := ac.GetValue(sh)
+		if err != nil {
+			return false, err
+		} else {
+			return v.(bool), nil
+		}
+	}
+}
+
+func updateAppContextFlag(cid interface{}, sf bool) error {
+	ac := appcontext.AppContext{}
+	_, err := ac.LoadAppContext(cid)
+	if err != nil {
+		return err
+	}
+	hc, err := ac.GetCompositeAppHandle()
+	if err != nil {
+		return err
+	}
+	sh, err := ac.GetLevelHandle(hc, "stopflag")
+	if sh == nil {
+		_, err = ac.AddLevelValue(hc, "stopflag", sf)
+	} else {
+		err = ac.UpdateValue(sh, sf)
+	}
+	if err != nil {
+		return err
+	}
+	return nil
+}
+
+func applyFnComApp(instca *CompositeAppContext, acStatus appcontext.AppContextStatus, f fn, sfn statusfn, breakonError bool) error {
+	con := connector.Init(instca.cid)
 	//Cleanup
 	defer con.RemoveClient()
 	ac := appcontext.AppContext{}
-	h, err := ac.LoadAppContext(cid)
+	h, err := ac.LoadAppContext(instca.cid)
 	if err != nil {
 		return err
 	}
 
+	// if terminating, wait for all retrying instantiate threads to exit
+	if acStatus.Status == appcontext.AppContextStatusEnum.Terminating {
+		waitForDone(ac)
+		err := updateAppContextFlag(instca.cid, false)
+		if err != nil {
+			return err
+		}
+	}
+
 	// initialize appcontext status
 	err = initializeAppContextStatus(ac, acStatus)
 	if err != nil {
@@ -375,6 +685,8 @@ func applyFnComApp(cid interface{}, acStatus appcontext.AppContextStatus, f fn,
 	})
 	id, _ := ac.GetCompositeAppHandle()
 	g, _ := errgroup.WithContext(context.Background())
+	wg, _ := errgroup.WithContext(context.Background())
+	kickoffRetryWatcher(instca, ac, acStatus, wg)
 	// Iterate over all the subapps
 	for _, app := range appList["apporder"] {
 		appName := app
@@ -414,8 +726,13 @@ func applyFnComApp(cid interface{}, acStatus appcontext.AppContextStatus, f fn,
 					json.Unmarshal([]byte(resorder.(string)), &aov)
 					// Keep retrying for reachability
 					for {
+						done := allResourcesDone(ac, appName, cluster, aov)
+						if done {
+							break
+						}
+
 						// Wait for cluster to be reachable
-						err = waitForClusterReady(c, cluster)
+						err := waitForClusterReady(instca, ac, c, appName, cluster, aov)
 						if err != nil {
 							// TODO: Add error handling
 							return err
@@ -479,19 +796,39 @@ func applyFnComApp(cid interface{}, acStatus appcontext.AppContextStatus, f fn,
 		logutils.Error("Encountered error updating AppContext status", logutils.Fields{"error": err})
 		return err
 	}
+	if err := wg.Wait(); err != nil {
+		logutils.Error("Encountered error in watcher thread", logutils.Fields{"error": err})
+		return err
+	}
 	return nil
 }
 
 // InstantiateComApp Instantiate Apps in Composite App
 func (instca *CompositeAppContext) InstantiateComApp(cid interface{}) error {
-	go applyFnComApp(cid, appcontext.AppContextStatus{Status: appcontext.AppContextStatusEnum.Instantiating},
+	instca.cid = cid
+	instca.chans = []chan bool{}
+	instca.mutex = sync.Mutex{}
+	err := updateAppContextFlag(cid, false)
+	if err != nil {
+		logutils.Error("Encountered error updating AppContext flag", logutils.Fields{"error": err})
+		return err
+	}
+	go applyFnComApp(instca, appcontext.AppContextStatus{Status: appcontext.AppContextStatusEnum.Instantiating},
 		instantiateResource, addStatusTracker, true)
 	return nil
 }
 
 // TerminateComApp Terminates Apps in Composite App
 func (instca *CompositeAppContext) TerminateComApp(cid interface{}) error {
-	go applyFnComApp(cid, appcontext.AppContextStatus{Status: appcontext.AppContextStatusEnum.Terminating},
+	instca.cid = cid
+	instca.chans = []chan bool{}
+	instca.mutex = sync.Mutex{}
+	err := updateAppContextFlag(cid, true)
+	if err != nil {
+		logutils.Error("Encountered error updating AppContext flag", logutils.Fields{"error": err})
+		return err
+	}
+	go applyFnComApp(instca, appcontext.AppContextStatus{Status: appcontext.AppContextStatusEnum.Terminating},
 		terminateResource, deleteStatusTracker, false)
 	return nil
 }