Make plural handling more robust (#202)

Author: jthomperoo

CHANGELOG.md

@@ -5,17 +5,20 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
+### Fixed
+- Handling of plurals is more robust, it no longer simply adds a 's' to the end which failed for plurals such
+as 'logstash' -> 'logstashes', now queries using the discovery client and REST mapper.
 
 ## [v2.12.0] - 2025-02-12
 ### Added
-- Added support for Python 3.13 Docker image (`custompodautoscaler/python-3-13`)
+- Added support for Python 3.13 Docker image (`custompodautoscaler/python-3-13`).
 ### Changed
-- Upgraded to Go `v1.24`
-- Upgraded package dependencies
+- Upgraded to Go `v1.24`.
+- Upgraded package dependencies.
 - Updated `custompodautoscaler/python` to track Python 3.13.
 - Updated `custompodautoscaler/python-3-12` to track the latest Debian stable version.
 ### Removed
-- Dropped support for Python 3.8 Docker image (`custompodautoscaler/python-3-8`)
+- Dropped support for Python 3.8 Docker image (`custompodautoscaler/python-3-8`).
 
 ## [v2.11.0] - 2024-03-22
 ### Changed

README.md

@@ -44,6 +44,7 @@ or HTTP request.
 - Supports Kubernetes metrics that the Horizontal Pod Autoscaler uses, can be configured using a similar syntax and
 used in custom scaling decisions.
 - Supports [Argo Rollouts](https://argoproj.github.io/argo-rollouts/).
+- Supports targeting any custom resource that implements the scale subresource.
 
 ## Why would I use it?
 

docs/index.md

@@ -30,6 +30,7 @@ or HTTP request.
 - Supports Kubernetes metrics that the Horizontal Pod Autoscaler uses, can be configured using a similar syntax and
 used in custom scaling decisions.
 - Supports [Argo Rollouts](https://argoproj.github.io/argo-rollouts/).
+- Supports targeting any custom resource that implements the scale subresource.
 
 ## Why would I use it?
 

docs/reference/docker-images.md

@@ -2,9 +2,9 @@
 
 The Custom Pod Autoscaler is bundled with the following Docker base images:
 
-- `custompodautoscaler/python` tracks latest stable Python 3.x release (alpine based).
-- `custompodautoscaler/python-3-13` tracks latest stable Python 3.13.x release (alpine based).
-- `custompodautoscaler/python-3-12` tracks latest stable Python 3.12.x release (debian 12 bookworm based).
+- `custompodautoscaler/python` tracks latest stable Python 3.x release (debian based).
+- `custompodautoscaler/python-3-13` tracks latest stable Python 3.13.x release (debian based).
+- `custompodautoscaler/python-3-12` tracks latest stable Python 3.12.x release (debian based).
 - `custompodautoscaler/alpine` tracks latest stable Alpine 3.x release.
 
 ## Deprecated images

docs/reference/scaling-targets.md

@@ -13,6 +13,8 @@ scale subresource API. For example this means support for:
 * Argo Rollouts
 * Other third party resources.
 
+> To see how to target custom resources [see the Custom Resources page](../user-guide/custom-resources.md).
+
 ## Scale target reference
 
 To tell a Custom Pod Autoscaler which resource to target, provide a `scaleTargetRef` - a description of the resource to

docs/user-guide/custom-resources.md

@@ -0,0 +1,90 @@
+# Custom Resources
+
+The Custom Pod Autoscaler supports targeting custom resources; however this requires some additional configuration
+to make sure the autoscaler has the required permissions to be able to manage the targeted resources.
+
+By default the [Custom Pod Autoscaler
+Operator](https://github.com/jthomperoo/custom-pod-autoscaler-operator) will provision a role for your autoscaler
+which allows managing the built-in Kubernetes resources (deployments, replicasets, statefulsets,
+replicationcontrollers), but if you are targeting a custom resource you need to provide your own role with the
+correct permissions.
+
+You need to tell the operator not to provision a role for you, and provide your own with the permissions needed,
+for example to target Loadstash:
+
+```yaml
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: python-custom-autoscaler
+rules:
+- apiGroups:
+    - logstash.k8s.elastic.co
+  resources:
+    - logstashes
+    - logstashes/scale
+  verbs:
+    - '*'
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: python-custom-autoscaler
+---
+kind: RoleBinding
+apiVersion: rbac.authorization.k8s.io/v1
+metadata:
+  name: python-custom-autoscaler
+subjects:
+- kind: ServiceAccount
+  name: python-custom-autoscaler
+roleRef:
+  kind: Role
+  name: python-custom-autoscaler
+  apiGroup: rbac.authorization.k8s.io
+---
+apiVersion: custompodautoscaler.com/v1
+kind: CustomPodAutoscaler
+metadata:
+  name: python-custom-autoscaler
+spec:
+  template:
+    spec:
+      serviceAccountName: python-custom-autoscaler
+      containers:
+      - name: python-custom-autoscaler
+        image: python-custom-autoscaler:latest
+        imagePullPolicy: IfNotPresent
+  scaleTargetRef:
+    apiVersion: logstash.k8s.elastic.co/v1alpha1
+    kind: Logstash
+    name: quickstart
+  provisionRole: false
+  provisionRoleBinding: false
+  provisionServiceAccount: false
+  config:
+    - name: interval
+      value: "10000"
+```
+
+This takes over provisioning of the role, the role binding, and the service account from the operator.
+
+For any custom resource the CPO can support scaling it if the resource implements the scale subresource, and if so
+the permissions needed are generally:
+
+```yaml
+- apiGroups:
+    - my.api.group
+  resources:
+    - mycustomresource
+    - mycustomresource/scale
+  verbs:
+    - '*'
+```
+
+There is a special case for [Argo Rollouts](https://argoproj.github.io/rollouts/), which can simply provide the
+`roleRequiresArgoRollouts` field. [See more information
+here](https://github.com/jthomperoo/custom-pod-autoscaler-operator/blob/v1.4.2/USAGE.md#automatically-provisioning-a-role-that-supports-argo-rollouts).
+
+[See how to skip automatic role/any other resource provision
+here](https://github.com/jthomperoo/custom-pod-autoscaler-operator/blob/v1.4.2/USAGE.md#using-custom-resources).

docs/user-guide/troubleshooting.md

@@ -0,0 +1,18 @@
+# Troubleshooting
+
+This page explains some of the common issues people may experience.
+
+## Autoscaler is Forbidden From Managing Custom Resources
+
+You may encounter an error like this:
+
+`
+E0214 11:30:09.556332       1 main.go:266] Error while autoscaling: failed to get managed resource: logstashes.logstash.k8s.elastic.co "quickstart" is forbidden: User "system:serviceaccount:default:python-custom-autoscaler" cannot get resource "logstashes" in API group "logstash.k8s.elastic.co" in the namespace "default"
+`
+
+And see that your autoscaler is not scaling the target resource if you are targeting a custom resource.
+
+This is because your autoscaler does not have the correct permissions to manage the resource you are
+targeting.
+
+See [the Custom Resources page for details on how to resolve this](./custom-resources.md).

example/simple-pod-metrics-python/config.yaml

@@ -1,15 +1,16 @@
 evaluate:
   type: "shell"
   timeout: 2500
-  shell: 
+  shell:
     entrypoint: "python"
-    command: 
+    command:
       - "/evaluate.py"
 metric:
   type: "shell"
   timeout: 2500
-  shell: 
+  shell:
     entrypoint: "python"
-    command: 
+    command:
       - "/metric.py"
-runMode: "per-pod"
+runMode: "per-pod"
+logVerbosity: 3

internal/resourceclient/resourceclient.go

@@ -1,5 +1,5 @@
 /*
-Copyright 2021 The Custom Pod Autoscaler Authors.
+Copyright 2025 The Custom Pod Autoscaler Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -18,10 +18,8 @@ package resourceclient
 
 import (
 	"context"
-	"fmt"
-
-	"strings"
 
+	meta "k8s.io/apimachinery/pkg/api/meta"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
 	"k8s.io/apimachinery/pkg/runtime/schema"
@@ -37,31 +35,21 @@ type Client interface {
 // UnstructuredClient is an implementation of the arbitrary resource client that uses a dynamic Kubernetes interface,
 // retrieving unstructured k8s objects and converting them to metav1.Object
 type UnstructuredClient struct {
-	Dynamic dynamic.Interface
+	Dynamic    dynamic.Interface
+	RESTMapper meta.RESTMapper
 }
 
 // Get takes descriptors of a Kubernetes object (api version, kind, name, namespace) and fetches the matching object,
 // returning it as an unstructured Kubernetes resource
 func (u *UnstructuredClient) Get(apiVersion string, kind string, name string, namespace string) (*unstructured.Unstructured, error) {
-	// TODO: update this to be less hacky
-	// Convert to plural and lowercase
-	kindPlural := fmt.Sprintf("%ss", strings.ToLower(kind))
-
-	// Parse group version
-	resourceGV, err := schema.ParseGroupVersion(apiVersion)
+	resourceGK := schema.FromAPIVersionAndKind(apiVersion, kind)
+	mapping, err := u.RESTMapper.RESTMapping(resourceGK.GroupKind(), resourceGK.Version)
 	if err != nil {
 		return nil, err
 	}
 
-	// Build GVR
-	resourceGVR := schema.GroupVersionResource{
-		Group:    resourceGV.Group,
-		Version:  resourceGV.Version,
-		Resource: kindPlural,
-	}
-
 	// Get resource
-	resource, err := u.Dynamic.Resource(resourceGVR).Namespace(namespace).Get(context.Background(), name, metav1.GetOptions{})
+	resource, err := u.Dynamic.Resource(mapping.Resource).Namespace(namespace).Get(context.Background(), name, metav1.GetOptions{})
 	if err != nil {
 		return nil, err
 	}