kubectl-kedify plugin

Simple TUI based shell script for working with Kedify.

(pauseable demo)

Commands

Core Commands

• status, s - Prints the status of Kedify agent
• logs, l - Prints the logs of Kedify agent
• autoscale, a - Runs the interactive mode for creating HTTPScaledObject

Debug & Analysis Commands

• debug, dbg - Provides low-level information regarding Kedify components

  • so/scaledobject - Inspect ScaledObject metrics through its generated HPA or Kedify Pod Autoscaler (KPA)
  • httpaddon - Verify HTTP Addon setup and current configuration

• insights, ins - Analyzes ScaledObjects for potential configuration issues

  • Checks for polling interval effectiveness when minReplicaCount > 0
  • Identifies low polling interval values that might overload services
  • Detects missing fallback configuration for supported scalers

• dump, dmp - Collects comprehensive debug information from Kedify/KEDA components

  • Gathers cluster-wide information (nodes, autoscaler data, etc.)
  • Collects namespace-specific data (events, scaling resources, pod logs)
  • Supports output to directory or compressed archive format
  • Use -c false or --collect-cluster-data=false to skip cluster-wide data collection for faster execution
  • When KPA is installed, collects its CRD, objects, events, controller resources, logs, endpoint state, and Prometheus metrics

KPA diagnostics

Kedify KEDA stores the generated pod autoscaler's name in ScaledObject.status.hpaName for both autoscaler classes. kubectl kedify debug scaledobject resolves that name as a Kubernetes HPA and as an optional autoscaling.kedify.io/v1alpha1 KedifyPodAutoscaler. An explicit autoscaling.kedify.io/class requires that exact kind to be readable, and any dual-kind coexistence is reported as a conflict instead of showing potentially stale metrics. KPA metric values come directly from its Resource, ContainerResource, and External status entries, without depending on the Kubernetes external metrics adapter. This makes the command safe during HPA-to-KPA class transitions without requiring the private KPA Go API.

kubectl kedify dump uses the stable app.kubernetes.io/part-of=kedify-pod-autoscaler label to find KPA controller resources across tenant namespaces. Missing KPA CRDs and denied optional KPA reads are reported as unavailable and do not stop the rest of the diagnostic collection.

Quick start

Having krew installed, just run:

kubectl krew install --manifest-url=https://github.com/kedify/kubectl-kedify/raw/main/.krew.yaml

# output:
Installing plugin: kedify
Installed plugin: kedify
\
 | Use this plugin:
 | 	kubectl kedify
 | Documentation:
 | 	https://github.com/kedify/kubectl-kedify
/

Usage

k kedify --version

Update

kubectl kedify -v
kubectl krew uninstall kedify && kubectl krew install --manifest-url=https://github.com/kedify/kubectl-kedify/raw/main/.krew.yaml
kubectl kedify -v

Requirements

This plugin requires couple of binaries to work properly. kubecolor is optional, but recommended.

Mac:

brew install bat curl figlet fzf kubecolor yq jq

Linux:

yum install bat curl figlet fzf jq

or

apt-get install bat curl figlet fzf jq

and for yq consult the readme.

Running Dump Script Standalone

The dump command can also be run as a standalone script directly from GitHub without installing the kubectl plugin. This is useful for quick diagnostics or in environments where kubectl plugins cannot be installed.

Usage

Download and run the script directly:

bash <(curl -s https://raw.githubusercontent.com/kedify/kubectl-kedify/refs/heads/main/dump.sh) -A --archive

Download and inspect (and run) the script

curl -s https://raw.githubusercontent.com/kedify/kubectl-kedify/refs/heads/main/dump.sh > dump.sh

chmod +x dump.sh
./dump.sh -A --archive

Standalone Options

All the same options available in the kubectl plugin are supported:

• -o, --output DIR - Output directory or archive file path
• -n, --namespace NS - Specific namespace (default: current namespace)
• -A, --all-namespaces - Collect from all namespaces
• -q, --quiet - Quiet mode - suppress all status output for cleaner automation
• --archive - Create tar.gz archive
• -c, --collect-cluster-data=BOOL - Collect cluster-wide data (default: true)

Requirements for Standalone Usage

The standalone script requires the same dependencies as the plugin:

• kubectl (configured with cluster access)
• curl, jq, yq
• tar and gzip (for archive mode)

Development and Testing

Local Testing

This project includes a cross-platform test script that automatically detects and validates all bash scripts.

# Run default tests (full on Mac/Linux, smoke on others)
./test-cross-platform.sh

# Run smoke tests only (basic syntax, function loading - works on all platforms)
./test-cross-platform.sh smoke

# Run full tests (includes shellcheck, platform-specific features - Mac/Linux only)
./test-cross-platform.sh full

Running as kubectl plugin vs locally

The script automatically detects whether it's running as a kubectl plugin (via krew) or locally in development:

• As kubectl plugin: Uses scripts from the krew store (~/.krew/store/kedify/)
• Locally: Uses scripts from the project directory

This allows for seamless development and testing of unreleased features.