From a3952f4c2b3a724033480a3de8b939fb550385d4 Mon Sep 17 00:00:00 2001
From: Srdjan Petrovic <spetrovic@google.com>
Date: Thu, 7 Mar 2024 12:25:54 -0800
Subject: [PATCH] Fix the app and version names in the MultidimPodAutoscaler
 resources. (#126)

These names have to be valid label names. Moreover, they have to match
the resources we're deleting.

Without these changes, the multidimpodautoscaler resources will end up
hanging.
---
 internal/gke/gke.go   | 10 +++++-----
 internal/gke/nanny.go |  5 +++--
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/internal/gke/gke.go b/internal/gke/gke.go
index 4f2350c..2cabe3f 100644
--- a/internal/gke/gke.go
+++ b/internal/gke/gke.go
@@ -450,7 +450,7 @@ var autoJSONTmpl = template.Must(template.New("auto").Parse(`{
 // configured for the given Kubernetes ReplicaSet.
 func ensureReplicaSetAutoscaler(ctx context.Context, cluster *ClusterInfo, logger *slog.Logger, cfg *config.GKEConfig, replicaSet string) error {
 	dep := cfg.Deployment
-	name := name{dep.App.Name, replicaSet, dep.Id[:8]}.DNSLabel()
+	aName := name{dep.App.Name, replicaSet, dep.Id[:8]}.DNSLabel()
 	var b strings.Builder
 	if err := autoJSONTmpl.Execute(&b, struct {
 		Name             string
@@ -462,14 +462,14 @@ func ensureReplicaSetAutoscaler(ctx context.Context, cluster *ClusterInfo, logge
 		TargetKind       string
 		TargetName       string
 	}{
-		Name:             name,
+		Name:             aName,
 		Namespace:        namespaceName,
-		AppName:          dep.App.Name,
-		DeploymentID:     dep.Id,
+		AppName:          name{dep.App.Name}.DNSLabel(),
+		DeploymentID:     name{dep.Id}.DNSLabel(),
 		AppContainerName: appContainerName,
 		MinMemory:        memoryUnit.String(),
 		TargetKind:       "Deployment",
-		TargetName:       name,
+		TargetName:       aName,
 	}); err != nil {
 		return err
 	}
diff --git a/internal/gke/nanny.go b/internal/gke/nanny.go
index 9058847..538c86c 100644
--- a/internal/gke/nanny.go
+++ b/internal/gke/nanny.go
@@ -372,13 +372,14 @@ func RunManager(ctx context.Context, port int) error {
 		},
 		// deleteAppVersions
 		func(_ context.Context, app string, versions []*config.GKEConfig) error {
+			var errs []error
 			for _, version := range versions {
 				id := version.Deployment.Id
 				if err := kill(ctx, cluster, logger, app, id); err != nil {
-					return fmt.Errorf("kill %q: %w", version, err)
+					errs = append(errs, fmt.Errorf("kill %q: %w", version, err))
 				}
 			}
-			return nil
+			return errors.Join(errs...)
 		},
 	)