PCD Ops

An operations dashboard for Platform9 Private Cloud Director (PCD). Provides day-to-day operational visibility and automated actions across your PCD environment.

Features

Feature	Description
Resource Reclamation	Identify idle VMs, unused floating IPs, orphaned volumes, and stale snapshots
Capacity Planning	Cluster-level vCPU / RAM / storage headroom with what-if analysis
Inventory	Searchable, filterable view of all resources across projects
Snapshot Management	List, age, and delete snapshots with bulk operations
VM Right-Sizing	AI-assisted recommendations based on actual utilization from Prometheus
Anomaly & Drift Detection	Alert on metric anomalies and configuration drift
Log Analysis	AI-assisted log triage and pattern detection
Scheduled Jobs	Background data collection with job history and status

---

Deployment

PCD Ops runs as a dedicated appliance VM in your PCD environment — no Kubernetes, no containers.

VM requirements: 2 vCPU, 4 GB RAM, 20 GB disk. Needs outbound internet access for self-updates.

Ports: Only port 80 (HTTP) needs to be open. The appliance does not run sshd — management is via the VM console (VNC) or the web UI.

---

Method 1 — Prebuilt image (recommended)

No build tools required. Download the image, upload it to PCD, and launch a VM.

1. Download

Download the latest pcd-ops-alpine.qcow2 from Releases.

2. Upload to PCD

pcdctl image create --insecure \
  --container-format bare \
  --disk-format qcow2 \
  --property os_type=Linux \
  --public \
  --file pcd-ops-alpine.qcow2 \
  pcd-ops

3. Launch the VM

In the PCD dashboard, create a new instance from the pcd-ops image:

• Flavor: 2 vCPU / 4 GB RAM or larger
• Network: your management network
• Security group: allow inbound port 80

Or use pcdctl server create if you prefer the CLI.

---

Method 2 — Build from source

For contributors or custom builds. Requires Packer and QEMU/KVM on the build host.

Prerequisites

# Install Packer
# https://developer.hashicorp.com/packer/install

# Install QEMU (Ubuntu/Debian)
apt install qemu-system-x86 qemu-utils

1. Clone and build

git clone https://github.com/jscott-pf9/pcd-ops.git
cd pcd-ops

./deploy/build-image.sh
# Output: deploy/packer/alpine/output/pcd-ops-alpine.qcow2
# Build time: ~5 minutes with KVM acceleration

2. Upload to PCD

pcdctl image create --insecure \
  --container-format bare \
  --disk-format qcow2 \
  --property os_type=Linux \
  --public \
  --file deploy/packer/alpine/output/pcd-ops-alpine.qcow2 \
  pcd-ops

3. Deploy with Terraform

cd deploy/terraform
cp terraform.tfvars.example terraform.tfvars
# Edit terraform.tfvars — set auth_url, credentials, network_name, flavor_name
./deploy.sh -auto-approve

The VM IP is printed on completion: app_url = "http://192.168.1.50"

---

Initial Setup

After the VM is running, open its IP in a browser:

1. The app opens to Settings → System Status
2. Go to Settings → PCD & Metrics — enter your PCD and Grafana credentials, click Save
3. Back on System Status, click Test Connections to verify
4. Click Collect Now to run the first data collection (~1–2 minutes)
5. Navigate to Inventory — data is now populated

Grafana / Prometheus: In PCD, Prometheus (VictoriaMetrics) is only reachable via the Grafana datasource proxy:

• GRAFANA_URL → https://your-pcd-host/grafana/
• GRAFANA_TOKEN → a Grafana service account token (glsa_...)
• PROMETHEUS_URL → https://your-pcd-host/grafana/api/datasources/proxy/1

---

VM Console

Open the VM console (VNC) in the PCD dashboard for appliance management — no SSH needed. The console shows service status, the web UI URL, and a menu:

• Main: Restart app · Reboot · Shutdown · Advanced options
• Advanced: Force update · View logs · Network info · Emergency shell

---

Self-Update

From the web UI: Settings → Software → Check for Updates → Update & Restart

From the VM console: A (Advanced) → U (Force update)

Both pull the latest code from GitHub, rebuild the frontend, and restart the service (~2 minutes).

---

Local Development

Prerequisites

• Python 3.10+, Node.js 18+
• Access to a PCD environment + Grafana

Backend

cd backend
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"

cp ../.env.example ../.env
# Edit .env with your PCD + Grafana credentials

uvicorn app.main:app --reload
# API at http://localhost:8000

Frontend

cd frontend
npm install
npm run dev
# UI at http://localhost:5173 (proxies /api/* to :8000)

Tests & Lint

# Backend
cd backend && pytest
ruff check . && ruff format .

# Frontend
cd frontend && npm test && npm run lint

---

Stack

• Frontend — React 18 + TypeScript, Vite, TanStack Query, Recharts, React Flow
• Backend — Python 3.10+, FastAPI, openstacksdk, prometheus-api-client
• AI — Pluggable provider: Claude (Anthropic) or Ollama
• Data Sources — PCD APIs + Prometheus/VictoriaMetrics via Grafana proxy

Project Structure

pcd-ops/
├── backend/app/
│   ├── api/          # One router per feature domain
│   ├── services/     # PCD, Prometheus, AI providers
│   ├── agent/        # Background job runner
│   └── main.py
├── frontend/src/
│   ├── api/          # Fetch wrappers per domain
│   ├── pages/        # One page per feature domain
│   └── components/   # Shared UI
└── deploy/
    ├── packer/alpine/ # Alpine VM image build
    ├── terraform/     # VM deployment (used with Method 2)
    ├── nginx/
    ├── openrc/
    └── scripts/       # update.sh, appliance-console.sh