Adds an overwrite flag to force annotation updates

docs/content/reference/pgo_backup.md

@@ -7,9 +7,10 @@ Backup cluster
 
 ### Synopsis
 
-Backup allows you to take a backup of a PostgreSQL cluster, either using
+Backup allows you to backup a PostgreSQL cluster either by using
 the current "spec.backups.pgbackrest.manual" settings on the PostgreSQL cluster
-or by overwriting those settings using the flags
+or by using flags to write your settings. Overwriting those settings may require
+the --force-conflicts flag.
 
 ### RBAC Requirements
     Resources                                           Verbs
@@ -33,6 +34,9 @@ pgo backup hippo
 # on the 'hippo' postgrescluster and trigger a backup
 pgo backup hippo --repoName="repo1" --options="--type=full"
 
+# Resolve ownership conflict
+pgo backup hippo --force-conflicts
+
 ```
 ### Example output
 ```
@@ -42,6 +46,7 @@ postgresclusters/hippo backup initiated
 ### Options
 
 ```
+      --force-conflicts       take ownership and overwrite the backup settings
   -h, --help                  help for backup
       --options stringArray   options for taking a backup; can be used multiple times
       --repoName string       repoName to backup to

docs/content/reference/pgo_restore.md

@@ -37,11 +37,16 @@ WARNING: This action is destructive and PostgreSQL will be unavailable while its
 
 Do you want to continue? (yes/no): yes
 postgresclusters/hippo patched
+
+# Resolve ownership conflict
+pgo restore hippo --force-conflicts
+
 ```
 
 ### Options
 
 ```
+      --force-conflicts       take ownership and overwrite the restore settings
   -h, --help                  help for restore
       --options stringArray   options to pass to the "pgbackrest restore" command; can be used multiple times
       --repoName string       repository to restore from

internal/cmd/backup.go

@@ -35,9 +35,10 @@ func newBackupCommand(config *internal.Config) *cobra.Command {
 	cmdBackup := &cobra.Command{
 		Use:   "backup CLUSTER_NAME",
 		Short: "Backup cluster",
-		Long: `Backup allows you to take a backup of a PostgreSQL cluster, either using
+		Long: `Backup allows you to backup a PostgreSQL cluster either by using
 the current "spec.backups.pgbackrest.manual" settings on the PostgreSQL cluster
-or by overwriting those settings using the flags
+or by using flags to write your settings. Overwriting those settings may require
+the --force-conflicts flag.
 
 ### RBAC Requirements
     Resources                                           Verbs
@@ -55,16 +56,20 @@ pgo backup hippo
 # on the 'hippo' postgrescluster and trigger a backup
 pgo backup hippo --repoName="repo1" --options="--type=full"
 
+# Resolve ownership conflict
+pgo backup hippo --force-conflicts
+
 ### Example output
 postgresclusters/hippo backup initiated`)
 
 	// Limit the number of args, that is, only one cluster name
 	cmdBackup.Args = cobra.ExactArgs(1)
 
-	// `backup` command accepts `repoName` and `options` flags;
+	// `backup` command accepts `repoName`, `force-conflicts` and `options` flags;
 	// multiple options flags can be used, with each becoming a new line
 	// in the options array on the spec
 	backup := pgBackRestBackup{}
+	cmdBackup.Flags().BoolVar(&backup.ForceConflicts, "force-conflicts", false, "take ownership and overwrite the backup settings")
 	cmdBackup.Flags().StringVar(&backup.RepoName, "repoName", "", "repoName to backup to")
 	cmdBackup.Flags().StringArrayVar(&backup.Options, "options", []string{},
 		"options for taking a backup; can be used multiple times")
@@ -110,12 +115,16 @@ postgresclusters/hippo backup initiated`)
 
 		// Update the spec/annotate
 		// TODO(benjaminjb): Would we want to allow a dry-run option here?
-		// TODO(benjaminjb): Would we want to allow a force option here?
+		patchOptions := metav1.PatchOptions{}
+		if backup.ForceConflicts {
+			b := true
+			patchOptions.Force = &b
+		}
 		_, err = client.Namespace(configNamespace).Patch(ctx,
 			args[0], // the name of the cluster object, limited to one name through `ExactArgs(1)`
 			types.ApplyPatchType,
 			patch,
-			config.Patch.PatchOptions(metav1.PatchOptions{}),
+			config.Patch.PatchOptions(patchOptions),
 		)
 
 		if err != nil {
@@ -134,8 +143,9 @@ postgresclusters/hippo backup initiated`)
 }
 
 type pgBackRestBackup struct {
-	Options  []string
-	RepoName string
+	Options        []string
+	RepoName       string
+	ForceConflicts bool
 }
 
 func (config pgBackRestBackup) modifyIntent(

internal/cmd/restore.go

@@ -54,7 +54,11 @@ WARNING: You are about to restore from pgBackRest with {options:[] repoName:repo
 WARNING: This action is destructive and PostgreSQL will be unavailable while its data is restored.
 
 Do you want to continue? (yes/no): yes
-postgresclusters/hippo patched`)
+postgresclusters/hippo patched
+
+# Resolve ownership conflict
+pgo restore hippo --force-conflicts
+`)
 
 	restore := pgBackRestRestore{Config: config}
 
@@ -64,6 +68,8 @@ postgresclusters/hippo patched`)
 	cmd.Flags().StringVar(&restore.RepoName, "repoName", "",
 		"repository to restore from")
 
+	cmd.Flags().BoolVar(&restore.ForceConflicts, "force-conflicts", false, "take ownership and overwrite the restore settings")
+
 	// Only one positional argument: the PostgresCluster name.
 	cmd.Args = cobra.ExactArgs(1)
 
@@ -119,8 +125,9 @@ postgresclusters/hippo patched`)
 type pgBackRestRestore struct {
 	*internal.Config
 
-	Options  []string
-	RepoName string
+	Options        []string
+	RepoName       string
+	ForceConflicts bool
 
 	PostgresCluster string
 }
@@ -168,14 +175,20 @@ func (config pgBackRestRestore) Run(ctx context.Context) error {
 	if err != nil {
 		return err
 	}
+	patchOptions := metav1.PatchOptions{
+		DryRun: []string{metav1.DryRunAll},
+	}
+
+	if config.ForceConflicts {
+		b := true
+		patchOptions.Force = &b
+	}
 
 	// Perform a dry-run patch to understand what settings will be used should
 	// the restore proceed.
 	cluster, err = client.Namespace(namespace).Patch(ctx,
 		config.PostgresCluster, types.ApplyPatchType, patch,
-		config.Patch.PatchOptions(metav1.PatchOptions{
-			DryRun: []string{metav1.DryRunAll},
-		}))
+		config.Patch.PatchOptions(patchOptions))
 	if err != nil {
 		return err
 	}
@@ -191,10 +204,16 @@ func (config pgBackRestRestore) Run(ctx context.Context) error {
 		return nil
 	}
 
+	patchOptions = metav1.PatchOptions{}
+	if config.ForceConflicts {
+		b := true
+		patchOptions.Force = &b
+	}
+
 	// They agreed to continue. Send the patch again without dry-run.
 	_, err = client.Namespace(namespace).Patch(ctx,
 		config.PostgresCluster, types.ApplyPatchType, patch,
-		config.Patch.PatchOptions(metav1.PatchOptions{}))
+		config.Patch.PatchOptions(patchOptions))
 
 	if err == nil {
 		fmt.Fprintf(config.Out, "%s/%s patched\n",

testing/kuttl/e2e/backup/12--backup-with-force-conflicts.yaml

@@ -0,0 +1,29 @@
+---
+apiVersion: kuttl.dev/v1beta1
+kind: TestStep
+commands:
+- script: |
+    pgbackrest_backup_annotation() {
+      kubectl get --namespace "${NAMESPACE}" postgrescluster/backup-cluster \
+        --output 'go-template={{ index .metadata.annotations "postgres-operator.crunchydata.com/pgbackrest-backup" }}'
+    }
+
+    kubectl --namespace "${NAMESPACE}" annotate postgrescluster/backup-cluster \
+      postgres-operator.crunchydata.com/pgbackrest-backup="$(date)" --overwrite || exit
+
+    PRIOR=$(pgbackrest_backup_annotation)
+    RESULT=$(kubectl-pgo --namespace "${NAMESPACE}" backup backup-cluster --force-conflicts)
+    CURRENT=$(pgbackrest_backup_annotation)
+
+    if [ "${CURRENT}" = "${PRIOR}" ]; then
+      printf 'Expected annotation to change, got %q' "${CURRENT}"
+      exit 1
+    fi
+
+    echo "RESULT from taking backup: ${RESULT}"
+
+    if [[ -n $RESULT && "$RESULT" == "postgresclusters/backup-cluster backup initiated" ]]; then
+        exit 0
+    fi
+
+    exit 1

testing/kuttl/e2e/restore/31--force-conflicts.yaml

@@ -0,0 +1,27 @@
+---
+apiVersion: kuttl.dev/v1beta1
+kind: TestStep
+commands:
+- script: |
+    pgbackrest_restore_annotation() {
+      kubectl --namespace "${NAMESPACE}" get postgrescluster/restore-cluster \
+        --output "jsonpath-as-json={.metadata.annotations['postgres-operator\.crunchydata\.com/pgbackrest-restore']}"
+    }
+
+    kubectl --namespace "${NAMESPACE}" annotate postgrescluster/restore-cluster \
+      postgres-operator.crunchydata.com/pgbackrest-restore="$(date)" --overwrite || exit
+
+
+    PRIOR=$(pgbackrest_restore_annotation)
+    # Running restore will update the annotation.
+    echo yes | kubectl-pgo --namespace="${NAMESPACE}" restore restore-cluster --options="--db-include=restore-cluster" --repoName="repo2" --force-conflicts
+    CURRENT=$(pgbackrest_restore_annotation)
+
+    if [ "${CURRENT}" != "${PRIOR}" ]; then
+      exit 0
+    fi
+
+    printf 'Expected annotation to change, got PRIOR %q CURRENT %q' "${PRIOR}" "${CURRENT}"
+    echo "RESULT from taking restore: ${RESULT}"
+
+    exit 1