🚀 EGS Quick Installation Guide

Overview

The EGS Quick Installer provides a one-command installation experience for EGS deployments. This guide is designed for users who want to get EGS up and running quickly without manual configuration. The installer supports both single-cluster and multi-cluster deployments, including the ability to install multiple worker clusters in a single command.

---

✨ Features

• 🎯 One-Command Installation: Install EGS with a single curl command
• 🔍 Auto-Detection: Automatically detects cluster capabilities (GPU nodes, cloud provider)
• 📝 Smart Defaults: Uses sensible defaults optimized for single-cluster and multi-cluster setups
• 🤖 Automated Setup: Handles all prerequisites automatically (PostgreSQL, Prometheus, GPU Operator)
• ⚡ Fast Deployment: Complete installation in 10-15 minutes
• 🔒 Conditional License: License only required when installing Controller (not for UI, Worker, or prerequisites)
• 🎛️ Flexible: Skip individual components as needed
• 🔄 Upgrade Support: Automatically detects existing installations and performs upgrades
• 🔗 Smart Dependencies: Validates component dependencies and checks for existing installations before blocking
• 🌐 Worker Registration: Register worker clusters with controller independently (--register-worker)
• 👥 Multiple Workers: Support for installing multiple worker clusters in a single command

---

🚦 Quick Start

Prerequisites

1. Kubernetes Cluster: Admin access to a Kubernetes cluster (v1.23.6+)
2. kubectl: Configured and connected to your cluster
3. EGS License: Valid license file (egs-license.yaml in current directory) - Only required when installing Controller. Not required for UI, Worker, or prerequisites (PostgreSQL, Prometheus, GPU Operator). In multi-cluster mode, the license is automatically applied to the controller cluster.
4. Required Tools: yq (v4.44.2+), helm (v3.15.0+), kubectl (v1.23.6+), jq (v1.6+), git
   • The installer checks these up front and aborts before downloading anything if git, kubectl, yq, or jq is missing.
   • Reliable yq install (avoids snap’s /tmp confinement): curl -fsSL https://github.com/mikefarah/yq/releases/download/v4.45.1/yq_linux_amd64 -o /usr/local/bin/yq && chmod +x /usr/local/bin/yq

📝 Registration Required

Complete the registration process at Avesha EGS Registration to receive:

• Access credentials
• Product license file (egs-license.yaml)

Simplest Installation

# Navigate to your installation directory
cd /path/to/your/directory

# Place your license file in the current directory
# (or specify path with --license-file)

# Run the installer
export KUBECONFIG=/path/to/your/kubeconfig.yaml
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash

That’s it! The script will:

1. ✅ Download EGS installer files internally
2. ✅ Auto-detect your cluster configuration
3. ✅ Generate egs-installer-config.yaml in your current directory
4. ✅ Apply the EGS license (only if installing Controller)
5. ✅ Install PostgreSQL, Prometheus, GPU Operator (unless explicitly skipped)
6. ✅ Install EGS Controller, UI, and Worker
7. ✅ Display access information and tokens

Preview the Generated Config First (Recommended)

Use --generate-config (alias --dry-run) to generate egs-installer-config.yaml and stop before touching the cluster. Review it, then re-run the same command without --generate-config to install.

# Generate and review the config WITHOUT installing anything
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file egs-license.yaml \
  --kubeconfig /path/to/kubeconfig \
  --cluster-name my-cluster \
  --generate-config

cat egs-installer-config.yaml     # review, then re-run WITHOUT --generate-config to install

---

📐 Topology-Based Installation Examples

This section provides copy-paste ready commands for different cluster topologies. Choose the topology that matches your setup.

---

🔹 Topology 1: Single Cluster (Everything in One Cluster)

Use case: PoC, development, or simple production setups where Controller, UI, and Worker all run on the same Kubernetes cluster.

1️⃣ Full Installation

📝 Note: Installs all components (PostgreSQL, Prometheus, GPU Operator, Controller, UI, Worker) on a single cluster.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"       # Path to your EGS license file
export KUBECONFIG_PATH="~/.kube/config"      # Path to your kubeconfig file
export CLUSTER_NAME="my-cluster"             # Name for your cluster

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --kubeconfig $KUBECONFIG_PATH \
  --cluster-name $CLUSTER_NAME

2️⃣ Skip Prerequisites

📝 Note: Use this when PostgreSQL, Prometheus, and GPU Operator are already installed on the cluster.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"       # Path to your EGS license file
export KUBECONFIG_PATH="~/.kube/config"      # Path to your kubeconfig file
export CLUSTER_NAME="my-cluster"             # Name for your cluster

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --kubeconfig $KUBECONFIG_PATH \
  --cluster-name $CLUSTER_NAME \
  --skip-postgresql --skip-prometheus --skip-gpu-operator

3️⃣ Install Only Worker

⚠️ Note: This installs Worker on the SAME cluster where Controller/UI are already running. Use this when you want to add Worker capability to an existing Controller cluster.

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="~/.kube/config"      # Path to your kubeconfig file

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-postgresql --skip-prometheus --skip-gpu-operator --skip-controller --skip-ui

---

🔹 Topology 2: 1 Controller Cluster + 1 Worker Cluster

Use case: Production setup with dedicated controller cluster and one worker cluster.

1️⃣ Full Installation

📝 Note: Installs Controller/UI on cluster-1 and Worker on cluster-2 with all prerequisites.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig
export WORKER_NAME="production-worker-1"                            # Name for the worker cluster

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --worker-name $WORKER_NAME

2️⃣ Skip All Prerequisites

📝 Note: Use this when PostgreSQL, Prometheus, and GPU Operator are already installed on both clusters.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig
export WORKER_NAME="production-worker-1"                            # Name for the worker cluster

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --worker-name $WORKER_NAME \
  --skip-postgresql \
  --skip-controller-prometheus --skip-controller-gpu-operator \
  --skip-worker-prometheus --skip-worker-gpu-operator

3️⃣ Skip Prerequisites on Controller Only

📝 Note: Skips prerequisites on controller cluster but installs them on worker cluster.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig
export WORKER_NAME="production-worker-1"                            # Name for the worker cluster

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --worker-name $WORKER_NAME \
  --skip-postgresql --skip-controller-prometheus --skip-controller-gpu-operator

---

🔹 Topology 3: 1 Controller Cluster + 2 Worker Clusters

Use case: Multi-region or multi-team setup with one controller managing two separate worker clusters.

1️⃣ Full Installation

📝 Note: Installs Controller/UI on cluster-1 and Workers on cluster-2 and cluster-3 with all prerequisites.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 cluster kubeconfig
export WORKER1_NAME="production-worker-1"                           # Name for worker 1
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 cluster kubeconfig
export WORKER2_NAME="production-worker-2"                           # Name for worker 2

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-name $WORKER2_NAME

2️⃣ With Custom Contexts

📝 Note: Use this when your kubeconfig files have multiple contexts and you need to specify which context to use.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export CONTROLLER_CONTEXT="controller-ctx"                          # Controller cluster context
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 cluster kubeconfig
export WORKER1_CONTEXT="worker1-ctx"                                # Worker 1 cluster context
export WORKER1_NAME="production-worker-1"                           # Name for worker 1
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 cluster kubeconfig
export WORKER2_CONTEXT="worker2-ctx"                                # Worker 2 cluster context
export WORKER2_NAME="production-worker-2"                           # Name for worker 2

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --controller-context $CONTROLLER_CONTEXT \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-context $WORKER1_CONTEXT \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-context $WORKER2_CONTEXT \
  --worker-name $WORKER2_NAME

3️⃣ Skip All Prerequisites

📝 Note: Use this when PostgreSQL, Prometheus, and GPU Operator are already installed on all clusters.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 cluster kubeconfig
export WORKER1_NAME="production-worker-1"                           # Name for worker 1
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 cluster kubeconfig
export WORKER2_NAME="production-worker-2"                           # Name for worker 2

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-name $WORKER2_NAME \
  --skip-postgresql \
  --skip-controller-prometheus --skip-controller-gpu-operator \
  --skip-worker-prometheus --skip-worker-gpu-operator

---

🔹 Topology 4: 1 Controller Cluster + 3 Worker Clusters

Use case: Large-scale production with one controller managing multiple worker clusters across different regions or cloud providers.

1️⃣ Full Installation

📝 Note: Installs Controller/UI on cluster-1 and Workers on cluster-2, cluster-3, and cluster-4 with all prerequisites.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 (US East) kubeconfig
export WORKER1_NAME="us-east-worker"                                # Name for worker 1
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 (US West) kubeconfig
export WORKER2_NAME="us-west-worker"                                # Name for worker 2
export WORKER3_KUBECONFIG="/path/to/worker3-kubeconfig.yaml"        # Worker 3 (EU West) kubeconfig
export WORKER3_NAME="eu-west-worker"                                # Name for worker 3

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-name $WORKER2_NAME \
  --worker-kubeconfig $WORKER3_KUBECONFIG \
  --worker-name $WORKER3_NAME

2️⃣ With Cloud Provider and Region

📝 Note: --cloud-provider and --cloud-region are global in a single full-install command — they are applied to every worker’s cluster_registration entry. To give each worker a distinct provider/region, register the workers separately with --register-worker (which takes per-worker --cloud-provider/--cloud-region), as shown in Adding a New Worker → With Telemetry Endpoint.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 kubeconfig
export WORKER1_NAME="us-east-worker"                                # Name for worker 1
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 kubeconfig
export WORKER2_NAME="us-west-worker"                                # Name for worker 2
export WORKER3_KUBECONFIG="/path/to/worker3-kubeconfig.yaml"        # Worker 3 kubeconfig
export WORKER3_NAME="eu-west-worker"                                # Name for worker 3
export CLOUD_PROVIDER="GCP"                                         # Cloud provider, applied to ALL workers
export CLOUD_REGION="us-west1"                                      # Cloud region, applied to ALL workers

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-name $WORKER2_NAME \
  --worker-kubeconfig $WORKER3_KUBECONFIG \
  --worker-name $WORKER3_NAME \
  --cloud-provider $CLOUD_PROVIDER \
  --cloud-region $CLOUD_REGION

3️⃣ Skip Prerequisites on Controller Only

📝 Note: Skips prerequisites on controller but installs them on all worker clusters.

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="egs-license.yaml"                           # Path to your EGS license file
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 (US East) kubeconfig
export WORKER1_NAME="us-east-worker"                                # Name for worker 1
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 (US West) kubeconfig
export WORKER2_NAME="us-west-worker"                                # Name for worker 2
export WORKER3_KUBECONFIG="/path/to/worker3-kubeconfig.yaml"        # Worker 3 (EU West) kubeconfig
export WORKER3_NAME="eu-west-worker"                                # Name for worker 3

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-name $WORKER2_NAME \
  --worker-kubeconfig $WORKER3_KUBECONFIG \
  --worker-name $WORKER3_NAME \
  --skip-postgresql --skip-controller-prometheus --skip-controller-gpu-operator

💡 Tip: This pattern scales to any number of workers. Simply add additional --worker-kubeconfig and --worker-name pairs for each worker cluster.

---

🔹 Adding a New Worker to Existing Setup

Use case: You already have a Controller + Workers deployed and want to add a new worker cluster.

1️⃣ Register AND Install (Recommended)

📝 Note: Registers the worker with the controller AND installs EGS Worker in one command.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/new-worker-kubeconfig.yaml"      # New worker cluster kubeconfig
export CLUSTER_NAME="new-worker-1"                                  # Name for the new worker
export PROJECT_NAME="avesha"                                        # Project name (default: avesha)

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME

2️⃣ With Telemetry Endpoint

📝 Note: Use this when the worker has an external Prometheus endpoint that the controller needs to access.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/new-worker-kubeconfig.yaml"      # New worker cluster kubeconfig
export CLUSTER_NAME="new-worker-1"                                  # Name for the new worker
export PROJECT_NAME="avesha"                                        # Project name (default: avesha)
export TELEMETRY_ENDPOINT="http://prometheus.new-worker.example.com:9090"  # External Prometheus endpoint
export CLOUD_PROVIDER="GCP"                                         # Cloud provider (GCP, AWS, Azure, etc.)
export CLOUD_REGION="us-west1"                                      # Cloud region

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME \
  --telemetry-endpoint $TELEMETRY_ENDPOINT \
  --cloud-provider $CLOUD_PROVIDER \
  --cloud-region $CLOUD_REGION

3️⃣ Register Only (No Installation)

📝 Note: Only registers the worker with the controller. Use this when you want to install the worker separately later.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export CLUSTER_NAME="new-worker-1"                                  # Name for the new worker
export PROJECT_NAME="avesha"                                        # Project name (default: avesha)

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME

---

🔹 Telemetry Endpoint Explained

The --telemetry-endpoint parameter specifies the Prometheus endpoint for the worker cluster. This is used by the controller to collect metrics and telemetry data from the worker cluster.

Scenario	Telemetry Endpoint Value
Single-cluster	http://prometheus-kube-prometheus-prometheus.egs-monitoring.svc.cluster.local:9090 (auto-configured)
Multi-cluster (same network)	Worker’s internal Prometheus endpoint (accessible from controller)
Multi-cluster (different networks)	Worker’s LoadBalancer/NodePort Prometheus endpoint (must be externally accessible)

Examples:

# Internal endpoint (same network)
--telemetry-endpoint http://prometheus-kube-prometheus-prometheus.egs-monitoring.svc.cluster.local:9090

# LoadBalancer endpoint (different networks)
--telemetry-endpoint http://prometheus-lb.worker-cluster.example.com:9090

# NodePort endpoint
--telemetry-endpoint http://worker-node-ip:30090

⚠️ Important: For multi-cluster setups where controller and workers are in different networks, the telemetry endpoint must be externally accessible from the controller cluster.

---

🌐 Multi-Cluster Network Requirements

In a multi-cluster setup the worker’s EGS operator connects back to the controller’s Kubernetes API to sync its Cluster and WorkerSlice* resources. Two independent reachability paths must therefore exist:

Path	Direction	Why it’s needed
Worker → Controller API	worker pods → https://<controller-ip>:6443	The worker operator reads/writes its resources in the controller’s project namespace (kubeslice-<project>).
Controller → Worker telemetry	controller → worker Prometheus	KubeTally / metrics collection (see Telemetry Endpoint Explained above).

Controller endpoint embedded in the worker onboarding secret. When a worker is registered, the controller mints a secret (kubeslice-rbac-worker-<name>) whose controllerEndpoint is taken from the controller’s advertised API URL. On single-node K3s/kubeadm the controller kubeconfig server is often https://127.0.0.1:6443, which a remote worker cannot reach. Override it so the worker receives a routable address:

curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --worker-name $WORKER_NAME \
  --controller-endpoint https://<routable-controller-ip>:6443

🩺 Symptom of an unreachable controller endpoint: the worker chart deploys and the kubeslice-operator pod can even report Ready (its readiness probe is a local health check), but its manager container logs repeat the following and reconciles never succeed:

failed to get server groups: Get "https://<controller-ip>:6443/api":
dial tcp <controller-ip>:6443: connect: no route to host

The worker chart itself installed correctly — it simply cannot reach the controller API. Fix L3 reachability (routing / firewall / NAT) and/or set --controller-endpoint to an address the worker can route to. The same applies in reverse for the telemetry path.

---

📋 Command Options

curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash [OPTIONS]

Available Options

Option	Description	Default	Required
--license-file PATH	Path to EGS license file (relative to current directory)	egs-license.yaml	No
--kubeconfig PATH	Path to kubeconfig file	Auto-detect	No
--context NAME	Kubernetes context to use	Current context	No
--cluster-name NAME	Cluster name for registration	worker-1	No
--project-name NAME	Project name used for cluster registration on full installs	avesha	No
--help, -h	Show help message	-	No

Behavior & Safety Flags

These flags make the installer safer to run repeatedly and easier to review before mutating a cluster.

Option	Description	Default
--generate-config (alias --dry-run)	Generate egs-installer-config.yaml and exit before any cluster-mutating action (no license apply, no installs). Use it to review the config first.	Off
--preserve-config	Reuse an existing egs-installer-config.yaml as-is instead of regenerating it, so manual edits survive a re-run.	Off
--skip-dependency-check	Bypass the helm/deployment-based PostgreSQL/Controller/UI prerequisite detection (for dependencies running under non-standard release names).	Off
--local-repo PATH	Use a local egs-installation checkout instead of cloning from GitHub (air-gapped, pinned-version, or pre-merge release-* branch testing). See Using a Local Checkout for examples.	Clone from GitHub

✅ Safe by default: The installer performs a single up-front tool check and aborts before any download if git, kubectl, yq, or jq is missing. On any early exit a cleanup trap removes temporary files and restores your original kubectl context, so the cluster is never left half-switched. When a config is regenerated, the previous one is backed up to egs-installer-config.yaml.bak.<timestamp>.

Using a Local Checkout (--local-repo)

By default the installer clones the main branch of egs-installation from GitHub at runtime and uses its charts/, egs-installer.sh, and egs-installer-config.yaml. --local-repo PATH makes it use an existing local checkout instead of cloning — essential for:

• Testing a specific branch/tag (e.g. a release-* branch) before it merges to main.
• Air-gapped / offline installs where the cluster host cannot reach GitHub.
• Pinned, reproducible installs against an audited copy of the repo.

What to pass: the absolute path to the root of an egs-installation checkout (a directory, not a file). The installer validates that the path exists and contains egs-installer.sh; it also needs charts/ and egs-installer-config.yaml inside it.

/root/egs-rel/                    ← pass THIS path to --local-repo
├── egs-installer.sh              ✅ required (presence is validated)
├── egs-installer-config.yaml     ✅ required (used as the config template)
├── charts/                       ✅ required (these charts are what get deployed)
│   ├── kubeslice-controller-egs/
│   ├── kubeslice-ui-egs/
│   └── kubeslice-worker-egs/
├── egs-install-prerequisites.sh
└── egs-uninstall.sh

When --local-repo is used, the installer prints ✅ Using local EGS installer checkout: <PATH> and skips the GitHub clone entirely — so the chart version that gets deployed is whatever that checkout contains, not main.

⚠️ Pass the repo root, not a sub-path. --local-repo /root/egs-rel/charts (too deep) or --local-repo /root/egs-rel/install-egs.sh (a file) will fail with --local-repo path is not a valid EGS installer checkout.

Example A — Test a release branch before it merges to main

📝 Note: Use this to validate release-* charts through the Quick Installer while main still ships the previous version. The chart version deployed comes from the checked-out branch.

# ============ CUSTOMIZE THESE VALUES ============
export RELEASE_BRANCH="release-v1.17.2"          # Branch/tag to test
export LOCAL_REPO="/root/egs-rel"                # Where to clone it
export KUBECONFIG_PATH="/etc/rancher/k3s/k3s.yaml"
export LICENSE_FILE="egs-license.yaml"
export CLUSTER_NAME="my-cluster"

# ============ STEP 1: Clone the release branch locally ============
git clone --depth 1 --branch "$RELEASE_BRANCH" \
  https://github.com/kubeslice-ent/egs-installation.git "$LOCAL_REPO"

# ============ STEP 2: Preview the config using the release charts (no install) ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --local-repo "$LOCAL_REPO" \
  --kubeconfig "$KUBECONFIG_PATH" \
  --license-file "$LICENSE_FILE" \
  --cluster-name "$CLUSTER_NAME" \
  --generate-config

# ============ STEP 3: Install for real (re-run WITHOUT --generate-config) ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --local-repo "$LOCAL_REPO" \
  --kubeconfig "$KUBECONFIG_PATH" \
  --license-file "$LICENSE_FILE" \
  --cluster-name "$CLUSTER_NAME"

# ============ VERIFY: charts deployed at the release version ============
helm list -A | grep -E "egs-controller|egs-ui|egs-worker"

Example B — Air-gapped / offline install (no GitHub at runtime)

📝 Note: Stage the repo on the host beforehand; --local-repo then needs no network for the installer artifacts.

# ============ ON A MACHINE WITH INTERNET: package the repo ============
git clone --depth 1 --branch release-v1.17.2 \
  https://github.com/kubeslice-ent/egs-installation.git egs-installation
tar -czf egs-installation.tgz egs-installation
# ...transfer egs-installation.tgz AND install-egs.sh to the air-gapped host...

# ============ ON THE AIR-GAPPED HOST: extract and install ============
tar -xzf egs-installation.tgz -C /opt/          # creates /opt/egs-installation
bash install-egs.sh \
  --local-repo /opt/egs-installation \
  --kubeconfig /etc/rancher/k3s/k3s.yaml \
  --license-file egs-license.yaml \
  --cluster-name my-cluster

Example C — Reuse an existing local checkout

# You already have the repo checked out (any branch you want to install from)
cd /home/user/egs-installation
git checkout release-v1.17.2

curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --local-repo /home/user/egs-installation \
  --kubeconfig ~/.kube/config \
  --generate-config

Common Skip Flags (works for both single & multi-cluster)

Option	Description	Default
--skip-postgresql	Skip PostgreSQL installation (controller only - PostgreSQL is never on workers)	Install
--skip-controller	Skip EGS Controller installation	Install
--skip-ui	Skip EGS UI installation	Install
--skip-worker	Skip EGS Worker installation	Install

Single-Cluster Mode Skip Flags (use ONLY in single-cluster mode)

Option	Description	Default
--skip-prometheus	Skip Prometheus installation	Install
--skip-gpu-operator	Skip GPU Operator installation	Install

Multi-Cluster Mode Skip Flags (use in multi-cluster mode)

Controller Cluster:

Option	Description	Default
--skip-controller-prometheus	Skip Prometheus on controller cluster	Install
--skip-controller-gpu-operator	Skip GPU Operator on controller cluster	Install

Worker Cluster(s):

Option	Description	Default
--skip-worker-prometheus	Skip Prometheus on worker cluster(s)	Install
--skip-worker-gpu-operator	Skip GPU Operator on worker cluster(s)	Install

Important:

• --skip-postgresql works the same in both modes (PostgreSQL is only on controller)
• In single-cluster mode: Use --skip-prometheus, --skip-gpu-operator
• In multi-cluster mode: Use --skip-controller-prometheus, --skip-worker-prometheus, etc.
• EGS component flags (--skip-controller, --skip-ui, --skip-worker) work the same in both modes

Multi-Cluster Mode Options

Option	Description	Default	Required
--controller-kubeconfig PATH	Path to controller cluster kubeconfig	-	Yes (multi-cluster mode)
--controller-context NAME	Controller cluster context	Auto-detected	No
--worker-kubeconfig PATH	Path to worker cluster kubeconfig (can be specified multiple times)	-	Yes (multi-cluster mode)
--worker-context NAME	Worker cluster context (can be specified multiple times, matches order of –worker-kubeconfig)	Auto-detected	No
--worker-name NAME	Worker cluster name (can be specified multiple times, defaults to worker-1, worker-2, etc.)	Auto-generated	No

Note: Multi-cluster mode is automatically detected when both --controller-kubeconfig and at least one --worker-kubeconfig are provided. You can specify multiple --worker-kubeconfig flags to install multiple worker clusters.

Worker Registration Options

Option	Description	Default	Required
--register-worker	Register a worker cluster with controller (standalone mode)	-	No
--controller-kubeconfig PATH	Path to controller cluster kubeconfig	-	Yes (if --register-worker)
--controller-context NAME	Controller cluster context	Current context	No
--worker-kubeconfig PATH	Path to worker cluster kubeconfig (for validation)	-	No
--worker-context NAME	Worker cluster context	Current context	No
--register-cluster-name NAME	Cluster name to register	-	Yes (if --register-worker)
--register-project-name NAME	Project name	avesha	No
--telemetry-endpoint URL	Prometheus endpoint URL	Auto-detected	No
--telemetry-provider NAME	Telemetry provider	prometheus	No
--cloud-provider NAME	Cloud provider name (overrides auto-detection)	Auto-detected	No
--cloud-region NAME	Cloud region	-	No
--controller-namespace NAME	Controller namespace	kubeslice-controller	No

Advanced Override Options

Option	Description	Default	Required
--controller-endpoint URL	Override the controller API endpoint that gets embedded in each worker onboarding secret (controllerEndpoint). Required when the controller kubeconfig server is 127.0.0.1/localhost (single-node K3s/kubeadm) and workers run in another network; also useful for Rancher or custom API server URLs. See Multi-Cluster Network Requirements.	Auto-detected from kubeconfig	No
--worker-endpoint URL	Override the auto-detected worker cluster API endpoint (can be specified multiple times, matches order of --worker-kubeconfig)	Auto-detected from kubeconfig	No
--ui-service-type TYPE	Set UI proxy service type: LoadBalancer, NodePort, or ClusterIP. The generated config ships NodePort by default (verified on K3s/bare-metal); pass this flag to override (e.g. LoadBalancer on clusters with a cloud LB controller). Invalid values are rejected.	NodePort (from config template)	No

---

📝 Usage Examples

Example 1: Basic Installation (License in Current Directory)

# ============ CUSTOMIZE THESE VALUES ============
export INSTALL_DIR="/home/user/egs-install"  # Directory for installation
export KUBECONFIG_PATH="/home/user/.kube/config"  # Path to your kubeconfig

# ============ RUN THE INSTALLER ============
cd $INSTALL_DIR
# Place egs-license.yaml in this directory
export KUBECONFIG=$KUBECONFIG_PATH
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash

Example 2: Specify License File Path

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="/home/user/.kube/config"  # Path to your kubeconfig
export LICENSE_FILE="/path/to/my-license.yaml"   # Path to your license file

# ============ RUN THE INSTALLER ============
export KUBECONFIG=$KUBECONFIG_PATH
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE

Example 3: Custom Cluster Name

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="/home/user/.kube/config"  # Path to your kubeconfig
export CLUSTER_NAME="production-cluster"          # Custom cluster name

# ============ RUN THE INSTALLER ============
export KUBECONFIG=$KUBECONFIG_PATH
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --cluster-name $CLUSTER_NAME

Example 4: Skip Specific Components

Skip PostgreSQL and GPU Operator

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="~/.kube/config"          # Path to your kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-postgresql --skip-gpu-operator

Install only Controller and UI (skip prerequisites and worker)

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="~/.kube/config"          # Path to your kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-postgresql --skip-prometheus --skip-gpu-operator --skip-worker

Example 5: Install Only Prerequisites (No License Required)

Install only Prometheus and GPU Operator

📝 Note: No license needed for prerequisites only.

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="~/.kube/config"          # Path to your kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-postgresql --skip-controller --skip-ui --skip-worker

Install only PostgreSQL

📝 Note: No license needed for prerequisites only.

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="~/.kube/config"          # Path to your kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-prometheus --skip-gpu-operator --skip-controller --skip-ui --skip-worker

Example 6: Install UI or Worker Without License (Controller Already Installed)

Install only UI

📝 Note: No license needed if Controller is already installed.

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="~/.kube/config"          # Path to your kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-postgresql --skip-prometheus --skip-gpu-operator --skip-controller --skip-worker

Install only Worker on Controller Cluster

⚠️ Note: This installs Worker on the SAME cluster where Controller/UI are already running. Use this for single-cluster setups.

# ============ CUSTOMIZE THESE VALUES ============
export KUBECONFIG_PATH="~/.kube/config"          # Path to your kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-postgresql --skip-prometheus --skip-gpu-operator --skip-controller --skip-ui

Example 7: Multi-Cluster Installation (Controller/UI in One Cluster, Worker in Another)

⚠️ Important: In multi-cluster mode, each cluster needs its own prerequisites (Prometheus, GPU Operator). The installer automatically configures:

• Controller cluster: PostgreSQL, Prometheus, GPU Operator
• Worker cluster(s): Prometheus, GPU Operator (no PostgreSQL needed)

Full Multi-Cluster Installation

📝 Note: Installs all prerequisites on all clusters.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG

Skip PostgreSQL

📝 Note: PostgreSQL is only on controller. Same flag for single & multi-cluster.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-postgresql

Skip Prerequisites on Controller Only

📝 Note: Prerequisites already installed on controller cluster, install on workers.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-postgresql --skip-controller-prometheus --skip-controller-gpu-operator

Skip All Prerequisites on ALL Clusters

📝 Note: Prerequisites already installed on all clusters.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-postgresql \
  --skip-controller-prometheus --skip-controller-gpu-operator \
  --skip-worker-prometheus --skip-worker-gpu-operator

Multiple Worker Clusters (Default Names)

📝 Note: Workers will be named worker-1, worker-2 automatically.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 kubeconfig
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-kubeconfig $WORKER2_KUBECONFIG

Multiple Worker Clusters with Custom Names

📝 Note: Use --worker-name after each --worker-kubeconfig to set custom names.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 kubeconfig
export WORKER1_NAME="production-worker-1"                           # Worker 1 name
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 kubeconfig
export WORKER2_NAME="production-worker-2"                           # Worker 2 name

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-name $WORKER2_NAME

Multiple Workers with Custom Contexts

📝 Note: Use when kubeconfig files have multiple contexts.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export CONTROLLER_CONTEXT="controller-ctx"                          # Controller cluster context
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 kubeconfig
export WORKER1_CONTEXT="worker1-ctx"                                # Worker 1 context
export WORKER1_NAME="worker-1"                                      # Worker 1 name
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 kubeconfig
export WORKER2_CONTEXT="worker2-ctx"                                # Worker 2 context
export WORKER2_NAME="worker-2"                                      # Worker 2 name

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --controller-context $CONTROLLER_CONTEXT \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-context $WORKER1_CONTEXT \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-context $WORKER2_CONTEXT \
  --worker-name $WORKER2_NAME

Example 8: Register Worker Cluster with Controller

Basic Worker Registration

📝 Note: Registers the worker with the controller. Add --worker-kubeconfig to also install the worker.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export CLUSTER_NAME="worker-2"                                      # Name for the worker cluster
export PROJECT_NAME="avesha"                                        # Project name (default: avesha)

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME

Register with Telemetry Endpoint and Cloud Provider

📝 Note: Use this for external Prometheus endpoints and geo-location tracking.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig
export CLUSTER_NAME="worker-3"                                      # Name for the worker cluster
export PROJECT_NAME="avesha"                                        # Project name (default: avesha)
export TELEMETRY_ENDPOINT="http://prometheus.example.com:9090"      # External Prometheus endpoint
export CLOUD_PROVIDER="GCP"                                         # Cloud provider (GCP, AWS, Azure)
export CLOUD_REGION="us-west1"                                      # Cloud region

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME \
  --telemetry-endpoint $TELEMETRY_ENDPOINT \
  --cloud-provider $CLOUD_PROVIDER \
  --cloud-region $CLOUD_REGION

Register Linode Worker

📝 Note: For Linode clusters, cloud provider/region are automatically left empty.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"      # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/linode-worker-kubeconfig.yaml"       # Linode worker kubeconfig
export CLUSTER_NAME="worker-linode-1"                                   # Name for the Linode worker
export PROJECT_NAME="avesha"                                            # Project name (default: avesha)

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME

---

🎯 What Gets Installed

Single-Cluster Installation Order

1. 📜 EGS License (Applied to kubeslice-controller namespace) - Only applied when installing Controller. Not required for UI, Worker, or prerequisites only.
2. 🗄️ PostgreSQL (Namespace: kt-postgresql) - Can be skipped
3. 📊 Prometheus Stack (Namespace: egs-monitoring) - Can be skipped
4. 🎮 GPU Operator (Namespace: egs-gpu-operator) - Can be manually skipped with --skip-gpu-operator
5. 🎛️ EGS Controller (Namespace: kubeslice-controller) - Can be skipped
6. 🌐 EGS UI (Namespace: kubeslice-controller) - Can be skipped
7. ⚙️ EGS Worker (Namespace: kubeslice-system) - Can be skipped, supports multiple workers in multi-cluster mode

Multi-Cluster Installation Order

In multi-cluster mode, prerequisites are installed on EACH cluster:

Controller Cluster:

1. 📜 EGS License (Applied to kubeslice-controller namespace)
2. 🗄️ PostgreSQL (Namespace: kt-postgresql) - Can be skipped with --skip-postgresql
3. 📊 Prometheus Stack (Namespace: egs-monitoring) - Can be skipped with --skip-controller-prometheus
4. 🎮 GPU Operator (Namespace: egs-gpu-operator) - Can be skipped with --skip-controller-gpu-operator
5. 🎛️ EGS Controller (Namespace: kubeslice-controller)
6. 🌐 EGS UI (Namespace: kubeslice-controller)

Worker Cluster(s):

1. 📊 Prometheus Stack (Namespace: egs-monitoring) - Can be skipped with --skip-worker-prometheus
2. 🎮 GPU Operator (Namespace: egs-gpu-operator) - Can be skipped with --skip-worker-gpu-operator
3. 📋 GPU Operator Quota (Namespace: egs-gpu-operator) - ResourceQuota for GPU pods
4. 🖥️ NVIDIA Driver Installer (Namespace: kube-system) - DaemonSet for GPU drivers
5. ⚙️ EGS Worker (Namespace: kubeslice-system)

Note: The worker requires Prometheus CRDs (PodMonitor) to be installed. If you skip Prometheus on the worker cluster, you may encounter errors like no matches for kind "PodMonitor".

Service Types

The generated egs-installer-config.yaml provisions these service types (verified on a K3s single-cluster install):

• Grafana: NodePort
• Prometheus: NodePort
• UI Proxy: NodePort on K3s / bare-metal (no cloud LoadBalancer); reachable at https://<node-ip>:<nodePort> — e.g. kubectl get svc kubeslice-ui-proxy -n kubeslice-controller shows 443:<nodePort>/TCP.

💡 On clusters that have a cloud LoadBalancer controller, pass --ui-service-type LoadBalancer to expose the UI via an external IP instead of a NodePort.

---

🔍 Auto-Detection Features

The script automatically detects and configures:

GPU Nodes Detection

# Script checks for GPU nodes
GPU_NODES=$(kubectl get nodes -o json | jq -r '.items[] | select(.status.capacity["nvidia.com/gpu"] != null) | .metadata.name')

Behavior:

• GPU nodes found: Sets enable_custom_apps: true, installs GPU Operator (unless --skip-gpu-operator)
• No GPU nodes (CPU-only): Sets enable_custom_apps: false

Cloud Provider Detection

# Auto-detects cloud provider from node providerID
kubectl get nodes -o jsonpath='{.items[0].spec.providerID}'

Special Handling:

• User Override: If --cloud-provider is provided, it takes precedence over auto-detection
• Linode: If auto-detected as Linode, cloudProvider field is left empty (Linode-specific requirement)
• Other providers: Sets cloudProvider to detected value (e.g., gcp, aws, azure)
• Cloud Region: Use --cloud-region to set the region (e.g., us-west1, us-east-1)

Node Labeling

The script automatically labels nodes with kubeslice.io/node-type=gateway if no such nodes exist, ensuring kubeslice-dns pod can be scheduled.

---

📁 Generated Files

After running the installer, you’ll find the following files in your current directory:

current-directory/
├── egs-installer-config.yaml    # Generated configuration (from repo template)
├── egs-installer.sh             # Main installer script
├── egs-install-prerequisites.sh # Prerequisites installer
├── egs-uninstall.sh             # Uninstaller script
├── fetch_egs_slice_token.sh     # Helper to fetch UI/admin login tokens
├── charts/                      # Helm charts directory
└── egs-license.yaml             # Your license file (if placed here)

Note: The installer clones the repository internally and uses egs-installer-config.yaml from the repository as the source of truth. It then updates this file with your specific configuration.

⚠️ Important: The Quick Installer always uses the same egs-installer-config.yaml file in your working directory. When you run the installer multiple times, it updates this existing file with your new configuration - it does not create a separate config file. In --register-worker mode, existing workers are preserved and the new worker is appended.

---

🎛️ Skip Parameters

You can skip individual components during installation. The installer intelligently handles dependencies and upgrade scenarios.

Single-Cluster Skip Flags

These flags apply to all components in a single-cluster setup, or to ALL clusters in multi-cluster mode (if no multi-cluster flags are used):

Skip Prerequisites:

• --skip-postgresql: Skip PostgreSQL installation (useful if using existing PostgreSQL)
• --skip-prometheus: Skip Prometheus installation (useful if using existing Prometheus)
• --skip-gpu-operator: Skip GPU Operator installation (useful for CPU-only clusters or existing GPU setup)

Skip EGS Components:

• --skip-controller: Skip EGS Controller installation
• --skip-ui: Skip EGS UI installation
• --skip-worker: Skip EGS Worker installation (applies to all workers when multiple workers are configured)

Multi-Cluster Skip Flags

These flags provide fine-grained control over prerequisites in multi-cluster mode:

Controller Cluster Prerequisites:

• --skip-controller-prometheus: Skip Prometheus on controller cluster only
• --skip-controller-gpu-operator: Skip GPU Operator on controller cluster only

Worker Cluster Prerequisites:

• --skip-worker-prometheus: Skip Prometheus on ALL worker clusters
• --skip-worker-gpu-operator: Skip GPU Operator on ALL worker clusters

Note: For PostgreSQL, use --skip-postgresql (same flag for both modes since PostgreSQL is only on controller)

How Skip Flags Work in Multi-Cluster Mode:

What you use	Controller Cluster	Worker Cluster(s)
--skip-postgresql	❌ PostgreSQL Skipped	N/A (not installed)
--skip-controller-prometheus	❌ Prometheus Skipped	✅ Prometheus Installed
--skip-worker-prometheus	✅ Prometheus Installed	❌ Prometheus Skipped
--skip-controller-prometheus --skip-worker-prometheus	❌ Prometheus Skipped	❌ Prometheus Skipped
--skip-controller-gpu-operator	❌ GPU Op Skipped	✅ GPU Op Installed
--skip-worker-gpu-operator	✅ GPU Op Installed	❌ GPU Op Skipped

Rules:

1. --skip-postgresql → Only affects controller (PostgreSQL is never on workers)
2. In multi-cluster mode, use --skip-controller-* and --skip-worker-* for Prometheus/GPU Operator
3. Using --skip-prometheus or --skip-gpu-operator in multi-cluster mode will show a warning
4. EGS component flags (--skip-controller, --skip-ui, --skip-worker) always work the same way

Examples:

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker.yaml"          # Worker cluster kubeconfig

# ============ EXAMPLE 1: Skip PostgreSQL ============
# (controller only - same flag for single & multi-cluster)
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-postgresql

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker.yaml"          # Worker cluster kubeconfig

# ============ EXAMPLE 2: Skip Prometheus on Controller Only ============
# (workers still get Prometheus)
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-controller-prometheus

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker.yaml"          # Worker cluster kubeconfig

# ============ EXAMPLE 3: Skip Prometheus on Workers Only ============
# (controller still gets Prometheus)
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-worker-prometheus

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker.yaml"          # Worker cluster kubeconfig

# ============ EXAMPLE 4: Skip All Prerequisites on Controller ============
# (install prerequisites on workers only)
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-postgresql --skip-controller-prometheus --skip-controller-gpu-operator

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker.yaml"          # Worker cluster kubeconfig

# ============ EXAMPLE 5: Skip Prometheus & GPU Operator Everywhere ============
# (all clusters)
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --skip-controller-prometheus --skip-controller-gpu-operator \
  --skip-worker-prometheus --skip-worker-gpu-operator

🔗 Dependency Management & Upgrade Support

The installer automatically validates component dependencies and supports upgrades:

Controller Dependencies:

• Requires PostgreSQL: If you use --skip-postgresql, the installer checks if PostgreSQL is already installed
  • ✅ If PostgreSQL exists: Controller installation/upgrade proceeds automatically
  • ❌ If PostgreSQL missing: Installation fails with clear error message

Worker Dependencies:

• Single-Cluster Mode: Requires both Controller and UI in the same cluster
  • ✅ If both exist: Worker installation/upgrade proceeds automatically
  • ❌ If either missing: Installation fails with clear error message
• Multi-Cluster Mode: Dependency checks are relaxed (Controller/UI may be in a different cluster)
  • ⚠️ Warning issued: If Controller/UI are not found in the worker cluster, a warning is shown but installation continues
  • ℹ️ Assumes multi-cluster setup: The installer assumes Controller/UI are in the controller cluster

Upgrade Scenarios:

• If a component is already installed, the installer automatically performs an upgrade instead of a fresh installation
• You can skip dependencies if they’re already installed (e.g., --skip-controller --skip-ui to upgrade only Worker)
• In multi-cluster mode, you can install workers independently of Controller/UI location

---

👥 Multiple Workers Support

The Quick Installer supports installing multiple worker clusters in a single command. This is particularly useful for multi-cluster deployments where you have multiple worker clusters that need to be managed by a single controller.

How It Works

• Multiple --worker-kubeconfig Flags: Specify --worker-kubeconfig multiple times, once for each worker cluster
• Worker Names: Use --worker-name to assign custom names to each worker (defaults to worker-1, worker-2, etc.)
• Worker Contexts: Use --worker-context to specify contexts for each worker (auto-detected if not provided)
• Order Matters: The order of --worker-kubeconfig, --worker-context, and --worker-name flags should match

Examples

Two Workers with Default Names

📝 Note: Workers will be named worker-1, worker-2 automatically.

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 kubeconfig
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 kubeconfig

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-kubeconfig $WORKER2_KUBECONFIG

Two Workers with Custom Names

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 kubeconfig
export WORKER1_NAME="production-worker-1"                           # Worker 1 name
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 kubeconfig
export WORKER2_NAME="production-worker-2"                           # Worker 2 name

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-name $WORKER2_NAME

Three Workers (Mix of Named and Default)

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER1_KUBECONFIG="/path/to/worker1-kubeconfig.yaml"        # Worker 1 kubeconfig
export WORKER1_NAME="worker-1"                                      # Worker 1 name
export WORKER2_KUBECONFIG="/path/to/worker2-kubeconfig.yaml"        # Worker 2 kubeconfig (auto-named)
export WORKER3_KUBECONFIG="/path/to/worker3-kubeconfig.yaml"        # Worker 3 kubeconfig
export WORKER3_NAME="worker-3"                                      # Worker 3 name

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER1_KUBECONFIG \
  --worker-name $WORKER1_NAME \
  --worker-kubeconfig $WORKER2_KUBECONFIG \
  --worker-kubeconfig $WORKER3_KUBECONFIG \
  --worker-name $WORKER3_NAME

Configuration

All workers are automatically added to the kubeslice_worker_egs array in the generated egs-installer-config.yaml file. Each worker gets:

• A unique name (custom or auto-generated)
• Its own kubeconfig file (copied to the working directory)
• Its own context (auto-detected or specified)
• Shared configuration (image registry, etc.)

Duplicate Worker Handling

When using --register-worker to add a new worker cluster:

• Automatic Duplicate Removal: If a worker with the same name already exists in the configuration, all duplicate entries are automatically removed before adding the new worker
• Preserved Workers: Existing workers (not being re-registered) automatically have skip_installation=true to prevent reinstallation
• New Worker: The newly registered worker has skip_installation=false (unless --skip-worker flag is provided) so it will be installed
• No Manual Cleanup Required: The script handles all duplicate detection and removal automatically

Example: If you register test-worker-2 twice:

1. First registration: Creates test-worker-2 with skip_installation=false
2. Second registration: Removes old test-worker-2 entry, creates new one with skip_installation=false
3. Other workers remain unchanged with skip_installation=true

Backward Compatibility

Single worker installations continue to work as before:

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig

# ============ SINGLE WORKER (STILL SUPPORTED) ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG

---

🔗 Worker Cluster Registration

The --register-worker feature allows you to register a worker cluster with an existing controller cluster. It supports two modes:

1. Registration Only: Register the cluster without installing the worker
2. Registration + Installation: Register AND install the worker in one command (when --worker-kubeconfig is provided)

When to Use

• Multi-cluster setups: Register worker clusters in different Kubernetes clusters
• Add new workers: Add new worker clusters to an existing EGS deployment
• Separate registration: Register workers independently, then install later
• One-command registration + installation: Register and install worker simultaneously

How It Works

1. Connects to Controller: Uses --controller-kubeconfig to connect to the controller cluster
2. Registers Cluster: Creates cluster CRD in the controller’s project namespace
3. Validates Worker (optional): If --worker-kubeconfig is provided, validates worker cluster connectivity
4. Detects Cloud Provider: Automatically detects Linode clusters and leaves cloud provider/region empty
5. Handles Duplicates: Automatically removes any existing worker entries with the same name
6. Preserves Existing Workers: Sets skip_installation=true for all existing workers
7. Installs Worker (if --worker-kubeconfig provided): Automatically proceeds with worker installation on the new cluster
8. Verifies Registration: Confirms the cluster was successfully registered

Required Parameters

• --register-worker: Enables registration mode
• --controller-kubeconfig PATH: Path to controller cluster kubeconfig file
• --register-cluster-name NAME: Unique name for the worker cluster

Optional Parameters

• --controller-context NAME: Controller cluster context (if not using default)
• --worker-kubeconfig PATH: Worker cluster kubeconfig (Important: If provided, worker will be installed automatically after registration)
• --worker-context NAME: Worker cluster context
• --register-project-name NAME: Project name (default: avesha)
• --telemetry-endpoint URL: Prometheus endpoint URL
• --telemetry-provider NAME: Telemetry provider (default: prometheus)
• --cloud-provider NAME: Cloud provider name (overrides auto-detection)
• --cloud-region NAME: Cloud region (e.g., us-west1, us-east-1)
• --controller-namespace NAME: Controller namespace (default: kubeslice-controller)
• --controller-endpoint URL: Override the controller API endpoint embedded in the worker onboarding secret (controllerEndpoint) — set this to a routable address when the controller kubeconfig uses 127.0.0.1/localhost and the worker is in another network; also useful for Rancher. Note: this flag is only applied when the Controller is installed/upgraded (it writes kubeslice.controller.endpoint into the controller’s Helm values). In --register-worker‑only mode the controller already exists, so this flag has no effect — the controllerEndpoint minted into new worker secrets comes from the controller’s existing advertised endpoint. To change it, re-run the installer against the controller with --controller-endpoint.
• --worker-endpoint URL: Override the auto-detected worker API endpoint (can be specified multiple times)
• --ui-service-type TYPE: Set UI proxy service type (LoadBalancer, NodePort, or ClusterIP)
• --skip-worker: Skip worker installation (even if --worker-kubeconfig is provided)

Linode Cluster Handling

When a Linode cluster is detected (via --worker-kubeconfig), the installer automatically:

• Sets cloudProvider to empty string
• Sets cloudRegion to empty string
• Ignores any user-provided cloud provider/region values

This is a Linode-specific requirement and is handled automatically.

Example Workflows

Option 1: Register AND Install Worker (One Command - Recommended)

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig
export CLUSTER_NAME="worker-2"                                      # Worker cluster name
export PROJECT_NAME="avesha"                                        # Project name

# ============ REGISTER AND INSTALL IN ONE COMMAND ============
# The script will:
# 1. Register worker-2 with the controller
# 2. Automatically install the worker on worker-2 cluster
# 3. Preserve existing workers (skip_installation=true)
# 4. Set new worker: skip_installation=false (will install)
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME

Option 2: Register Only (Without Installation)

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export CLUSTER_NAME="worker-2"                                      # Worker cluster name
export PROJECT_NAME="avesha"                                        # Project name

# ============ REGISTER ONLY (NO INSTALLATION) ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME

# ============ LATER: INSTALL THE WORKER SEPARATELY ============
export KUBECONFIG_PATH="~/.kube/config"  # Path to worker kubeconfig

curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --kubeconfig $KUBECONFIG_PATH \
  --skip-postgresql --skip-prometheus --skip-gpu-operator \
  --skip-controller --skip-ui

Option 3: Register Without Installing (Even With Kubeconfig)

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"          # Worker cluster kubeconfig
export CLUSTER_NAME="worker-2"                                      # Worker cluster name
export PROJECT_NAME="avesha"                                        # Project name

# ============ REGISTER BUT SKIP INSTALLATION ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --register-worker \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG \
  --register-cluster-name $CLUSTER_NAME \
  --register-project-name $PROJECT_NAME \
  --skip-worker

Verification

After registration, verify the cluster status:

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"  # Controller cluster kubeconfig
export PROJECT_NAMESPACE="kubeslice-avesha"                         # Project namespace

# ============ VERIFY CLUSTER REGISTRATION ============
kubectl --kubeconfig $CONTROLLER_KUBECONFIG \
  get cluster.controller.kubeslice.io -n $PROJECT_NAMESPACE

Error Handling

The registration process validates:

• ✅ Controller kubeconfig file exists and is accessible (fatal if it fails)
• ✅ Controller cluster connectivity (fatal if it fails)
• ⚠️ Worker cluster connectivity (if kubeconfig provided) — non-fatal: if the worker is unreachable the installer prints ⚠️ Cannot connect to worker cluster (non-fatal, continuing...), still registers the Cluster CR, and proceeds to worker installation. (Worker install itself will then fail later if the worker truly is unreachable.)
• ✅ Project namespace exists in controller cluster (fatal if missing)
• ✅ Required parameters are provided — e.g. --register-cluster-name is required; omitting it exits early with ❌ ERROR: --register-cluster-name is required for --register-worker

Fatal validations display a clear error message and exit early (with temporary-file cleanup). Worker-connectivity is intentionally non-fatal so registration can complete before the worker is reachable.

---

🔐 License File

Important: The EGS license file is only required when installing the Controller. It is not required for:

• Installing prerequisites (PostgreSQL, Prometheus, GPU Operator)
• Installing UI (if Controller is already installed)
• Installing Worker (if Controller and UI are already installed)
• Registering worker clusters (--register-worker)

Default Behavior

When installing Controller, the installer expects egs-license.yaml in the current directory by default:

# ============ CUSTOMIZE THESE VALUES ============
export INSTALL_DIR="/my/install/dir"  # Installation directory

# ============ INSTALL WITH DEFAULT LICENSE LOCATION ============
cd $INSTALL_DIR
# Place egs-license.yaml here
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash

Custom License Path

You can specify a custom path (relative to current directory):

# ============ CUSTOMIZE THESE VALUES ============
export LICENSE_FILE="/path/to/egs-license.yaml"  # Custom license file path

# ============ RUN THE INSTALLER ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE

License Not Found

Note: This only applies when installing Controller. If the license file is not found during Controller installation, the installer will:

1. Display an error message
2. Show steps to generate the license file
3. Exit with instructions

To generate your license:

1. Navigate to https://avesha.io/egs-registration
2. Complete the registration form
3. Generate cluster fingerprint: kubectl get namespace kube-system -o=jsonpath='{.metadata.creationTimestamp}{.metadata.uid}{"\n"}'
4. Receive license file via email
5. Save as egs-license.yaml in your installation directory

---

🔍 Accessing Your Installation

After successful installation, access the EGS UI.

📝 Note: The installer script output will display the UI URL and access token automatically at the end of installation. The steps below are for manual access if you need to retrieve these details later.

Script Output Example

At the end of installation, the script displays access information like this:

┌─────────────────────────────────────────────────────────────────────────────────────┐
│                         🌐 KUBESLICE UI ACCESS INFORMATION                          │
├─────────────────────────────────────────────────────────────────────────────────────┤
│ Service Type: ⚖️  LoadBalancer                                                      │
│ Access URL  : 🔗 https://<EXTERNAL-IP>                                              │
│ Status      : ✅ Ready for external access via LoadBalancer                         │
└─────────────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────────────┐
│                        🔐 KUBESLICE PROJECT ACCESS TOKENS                           │
├─────────────────────────────────────────────────────────────────────────────────────┤
│ 🔑 TOKEN: ✅ Available                                                              │
│                                                                                     │
├─────────────────────────────────────────────────────────────────────────────────────┤
eyJhbGciOiJSUzI1NiIsImtpZCI6....<TOKEN>....                                           │
├─────────────────────────────────────────────────────────────────────────────────────┤
│ 💡 USAGE: 📋 COPY THE ABOVE TOKEN AND PASTE IT ON PLACE OF ENTER SERVICE            │
│              ACCOUNT TOKEN IN BROWSER                                               │
└─────────────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────────────┐
│                          🏁 INSTALLATION SUMMARY COMPLETE                           │
├─────────────────────────────────────────────────────────────────────────────────────┤
│ ✅ All configured components have been processed.                                   │
│ 📋 Access information displayed above for quick reference.                          │
│ 🔧 For troubleshooting, check logs in file egs-installer-output.log                 │
│ 📚 Refer to documentation https://docs.avesha.io/documentation/enterprise-egs       │
└─────────────────────────────────────────────────────────────────────────────────────┘

Simply copy the Access URL and Token from the script output to access the UI.

---

Manual Access (If Needed)

If you need to retrieve the UI access details manually (e.g., after the script has finished), follow these steps:

1. Get the UI URL

# Get the UI service
kubectl get svc kubeslice-ui-proxy -n kubeslice-controller

# Example output (cloud LoadBalancer):
# NAME                 TYPE           CLUSTER-IP     EXTERNAL-IP      PORT(S)
# kubeslice-ui-proxy   LoadBalancer   10.x.x.x       <EXTERNAL-IP>    443:xxxxx/TCP

# Example output (K3s / bare-metal NodePort, verified):
# NAME                 TYPE       CLUSTER-IP     EXTERNAL-IP   PORT(S)
# kubeslice-ui-proxy   NodePort   10.43.33.240   <none>        443:32568/TCP

Access the UI at:

• LoadBalancer: https://<EXTERNAL-IP>
• NodePort: https://<node-ip>:<nodePort> (e.g. https://192.168.122.180:32568). A curl -sk -o /dev/null -w '%{http_code}' https://<node-ip>:<nodePort> returns 302 (redirect to the login page), confirming the UI is serving.

2. Get the Admin Token (recommended)

The most reliable method — works with only kubectl and the controller kubeconfig:

# Direct token retrieval using kubectl (project namespace is kubeslice-<project>, default: kubeslice-avesha)
kubectl get secret kubeslice-rbac-rw-admin -n kubeslice-avesha -o jsonpath='{.data.token}' | base64 -d

Copy the output and paste it into the UI login screen.

3. Helper Script (Alternative)

# The Quick Installer copies fetch_egs_slice_token.sh into your working directory
./fetch_egs_slice_token.sh -k /path/to/kubeconfig -p avesha -a -u admin

📝 Note: fetch_egs_slice_token.sh is copied into your working directory by the Quick Installer (alongside egs-installer.sh, egs-install-prerequisites.sh, and egs-uninstall.sh), so it is ready to run after a Quick Install.

Parameters:

• -k /path/to/kubeconfig: Absolute path to your kubeconfig file
• -p avesha: Project name (default: avesha)
• -a: Fetch admin token
• -u admin: Username for the admin token

📖 For detailed token retrieval options: See Slice & Admin Token Guide

---

🐛 Troubleshooting

License File Not Found

Error: ❌ ERROR: License file not found

Solution:

# Ensure license file exists in current directory
ls -la egs-license.yaml

# Or specify custom path
export LICENSE_FILE="/path/to/egs-license.yaml"
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --license-file $LICENSE_FILE

Kubeconfig Not Accessible

Error: No active Kubernetes context found!

Solution:

# ============ CUSTOMIZE THIS VALUE ============
export KUBECONFIG="/path/to/your/kubeconfig.yaml"

# ============ VERIFY CONNECTION ============
kubectl get nodes

Installation Timeout

If installation times out during Helm operations, check:

# Check pods status
kubectl get pods -A | grep -E "kubeslice|egs-|kt-postgresql"

# Check helm releases
helm list -A

# Check node resources
kubectl top nodes

PodMonitor CRD Not Found (Multi-Cluster)

Error: resource mapping not found for name: "gateway-pods-podmonitor" namespace: "egs-monitoring" from "": no matches for kind "PodMonitor" in version "monitoring.coreos.com/v1"

Cause: The worker cluster doesn’t have Prometheus CRDs installed. In multi-cluster mode, each worker cluster needs Prometheus installed to provide the PodMonitor CRD.

Solution:

# ============ CUSTOMIZE THESE VALUES ============
export CONTROLLER_KUBECONFIG="/path/to/controller.yaml"  # Controller kubeconfig
export WORKER_KUBECONFIG="/path/to/worker.yaml"          # Worker kubeconfig

# ============ OPTION 1: INSTALL WITH PROMETHEUS (RECOMMENDED) ============
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
  --controller-kubeconfig $CONTROLLER_KUBECONFIG \
  --worker-kubeconfig $WORKER_KUBECONFIG

# ============ CUSTOMIZE THIS VALUE ============
export WORKER_KUBECONFIG="/path/to/worker.yaml"  # Worker kubeconfig

# ============ OPTION 2: INSTALL PROMETHEUS MANUALLY ============
helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
helm install prometheus prometheus-community/kube-prometheus-stack \
  --namespace egs-monitoring --create-namespace \
  --kubeconfig $WORKER_KUBECONFIG

Dependency Errors

Error: ❌ ERROR: Controller installation requires PostgreSQL to be installed.

Solution:

• Install PostgreSQL first, or
• If PostgreSQL is already installed, ensure it’s detected by the installer:

  # Verify PostgreSQL is installed
  helm list -A | grep postgresql
    
  # If installed, the installer should detect it automatically
  # If not detected, check the release name matches (postgresql or kt-postgresql)
  

Error: ❌ ERROR: Worker installation requires Controller to be installed.

Solution:

• Single-Cluster Mode: Install Controller first, or ensure it’s detected:

  # Verify Controller is installed
  helm list -A | grep egs-controller
    
  # If installed, the installer should detect it automatically
  

• Multi-Cluster Mode: Use --controller-kubeconfig to specify the controller cluster:

  # ============ CUSTOMIZE THESE VALUES ============
  export CONTROLLER_KUBECONFIG="/path/to/controller-kubeconfig.yaml"
  export WORKER_KUBECONFIG="/path/to/worker-kubeconfig.yaml"
    
  # ============ IN MULTI-CLUSTER MODE ============
  curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash -s -- \
    --controller-kubeconfig $CONTROLLER_KUBECONFIG \
    --worker-kubeconfig $WORKER_KUBECONFIG
  

Error: ❌ ERROR: Worker installation requires UI to be installed.

Solution:

• Single-Cluster Mode: Install UI first, or ensure it’s detected:

  # Verify UI is installed
  helm list -A | grep egs-ui
    
  # If installed, the installer should detect it automatically
  

• Multi-Cluster Mode: UI is installed with Controller in the controller cluster. Use --controller-kubeconfig to specify the controller cluster (UI uses the same kubeconfig as Controller).

---

🔄 Reinstallation / Updates

Complete Reinstall

# 1. Uninstall existing components (if egs-installer-config.yaml exists)
bash egs-uninstall.sh --input-yaml egs-installer-config.yaml

# 2. Run installer again
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash

Update Components

The Quick Installer automatically handles upgrades. Simply re-run it to upgrade existing components:

# Re-run the installer - it will automatically upgrade existing components
curl -fsSL https://repo.egs.avesha.io/install-egs.sh | bash

---

📚 Related Documentation

• 📋 EGS License Setup - How to obtain and configure your license
• 🛠️ Full Installation Guide - For multi-cluster and advanced setups
• 📊 Configuration Documentation - Detailed configuration options
• ✅ Preflight Check - Validate your environment before installation
• 🌐 EGS User Guide - Complete product documentation

---