Try to reload config when disconnected from the cluster

pkg/kclient/interface.go

@@ -76,6 +76,7 @@ type ClientInterface interface {
 	SetDiscoveryInterface(client discovery.DiscoveryInterface)
 	IsResourceSupported(apiGroup, apiVersion, resourceName string) (bool, error)
 	IsSSASupported() bool
+	Refresh() (newConfig bool, err error)
 
 	// namespace.go
 	GetCurrentNamespace() string

pkg/kclient/refresh.go

@@ -0,0 +1,67 @@
+package kclient
+
+import (
+	"fmt"
+	"reflect"
+
+	"k8s.io/client-go/tools/clientcmd"
+)
+
+// Refresh re-creates a new Kubernetes client and checks if the Config changes
+// If config changed, updates the Kubernetes client with the new configuration and returns true
+// If the namespace or cluster of the current context has changed since the last time
+// the config has been loaded, the function will not update the configuration
+func (c *Client) Refresh() (bool, error) {
+	newClient, err := New()
+	if err != nil {
+		return false, err
+	}
+
+	oldCluster, oldNs, err := getContext(c)
+	if err != nil {
+		return false, err
+	}
+	newCluster, newNs, err := getContext(newClient)
+	if err != nil {
+		return false, err
+	}
+
+	if oldCluster != newCluster {
+		return false, fmt.Errorf("cluster changed (%q -> %q), won't refresh the configuration", oldCluster, newCluster)
+	}
+	if oldNs != newNs {
+		return false, fmt.Errorf("namespace changed (%q -> %q), won't refresh the configuration", oldNs, newNs)
+	}
+
+	updated, err := isConfigUpdated(c.GetConfig(), newClient.GetConfig())
+	if err != nil {
+		return false, err
+	}
+
+	if updated {
+		*c = *newClient
+	}
+	return updated, nil
+}
+
+func getContext(c *Client) (cluster string, namespace string, err error) {
+	raw, err := c.GetConfig().RawConfig()
+	if err != nil {
+		return "", "", err
+	}
+
+	currentCtx := raw.Contexts[raw.CurrentContext]
+	return currentCtx.Cluster, currentCtx.Namespace, nil
+}
+
+func isConfigUpdated(oldC, newC clientcmd.ClientConfig) (bool, error) {
+	oldRaw, err := oldC.RawConfig()
+	if err != nil {
+		return false, err
+	}
+	newRaw, err := newC.RawConfig()
+	if err != nil {
+		return false, err
+	}
+	return !reflect.DeepEqual(oldRaw, newRaw), nil
+}

pkg/watch/watch.go

@@ -27,6 +27,7 @@ import (
 
 	appsv1 "k8s.io/api/apps/v1"
 	corev1 "k8s.io/api/core/v1"
+	kerrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/watch"
 	"k8s.io/klog"
@@ -175,7 +176,7 @@ func (o *WatchClient) WatchAndPush(out io.Writer, parameters WatchParameters, ct
 	}
 
 	o.keyWatcher = getKeyWatcher(ctx, out)
-	return o.eventWatcher(ctx, parameters, out, evaluateFileChanges, processEvents, componentStatus)
+	return o.eventWatcher(ctx, parameters, out, evaluateFileChanges, o.processEvents, componentStatus)
 }
 
 // eventWatcher loops till the context's Done channel indicates it to stop looping, at which point it performs cleanup.
@@ -424,7 +425,7 @@ func evaluateFileChanges(events []fsnotify.Event, path string, fileIgnores []str
 	return changedFiles, deletedPaths
 }
 
-func processEvents(
+func (o *WatchClient) processEvents(
 	changedFiles, deletedPaths []string,
 	parameters WatchParameters,
 	out io.Writer,
@@ -461,12 +462,23 @@ func processEvents(
 			return nil, err
 		}
 		klog.V(4).Infof("Error from Push: %v", err)
-		if parameters.WatchFiles {
-			// Log and output, but intentionally not exiting on error here.
-			// We don't want to break watch when push failed, it might be fixed with the next change.
-			fmt.Fprintf(out, "%s - %s\n\n", PushErrorString, err.Error())
+		// Log and output, but intentionally not exiting on error here.
+		// We don't want to break watch when push failed, it might be fixed with the next push.
+		if kerrors.IsUnauthorized(err) || kerrors.IsForbidden(err) {
+			fmt.Fprintf(out, "Error connecting to the cluster. Please log in again\n\n")
+			var refreshed bool
+			refreshed, err = o.kubeClient.Refresh()
+			if err != nil {
+				fmt.Fprintf(out, "Error updating Kubernetes config: %s\n", err)
+			} else if refreshed {
+				fmt.Fprintf(out, "Updated Kubernetes config\n")
+			}
 		} else {
-			return nil, err
+			if parameters.WatchFiles {
+				fmt.Fprintf(out, "%s - %s\n\n", PushErrorString, err.Error())
+			} else {
+				return nil, err
+			}
 		}
 		wait := backoff.Delay()
 		return &wait, nil
@@ -484,7 +496,11 @@ func (o *WatchClient) CleanupDevResources(devfileObj parser.DevfileObj, componen
 	fmt.Fprintln(out, "Cleaning resources, please wait")
 	isInnerLoopDeployed, resources, err := o.deleteClient.ListResourcesToDeleteFromDevfile(devfileObj, "app", componentName, labels.ComponentDevMode)
 	if err != nil {
-		fmt.Fprintf(out, "failed to delete inner loop resources: %v", err)
+		if kerrors.IsUnauthorized(err) || kerrors.IsForbidden(err) {
+			fmt.Fprintf(out, "Error connecting to the cluster, the resources were not cleaned up.\nPlease log in again and cleanup the resource with `odo delete component`\n\n")
+		} else {
+			fmt.Fprintf(out, "Failed to delete inner loop resources: %v\n", err)
+		}
 		return err
 	}
 	// if innerloop deployment resource is present, then execute preStop events