Deploying Workloads

Quick Launch templates and custom Docker/YAML deployments — ports, domains, GPU selection, and persistent storage.

12 min read

dockerkubernetesyamldeploy

Deploying Workloads on FluxEdge

FluxEdge provides two ways to deploy workloads on rented machines: Quick Launch for pre-configured templates from the marketplace, and Custom Deployment for running your own Docker images with full Kubernetes YAML control. Any hardened Dockerized application can run on FluxEdge — from Jupyter notebooks to large language models.

Quick Launch

Quick Launch deploys pre-configured projects from the FluxEdge Marketplace with a few clicks. Available templates include popular AI/ML tools ready to use out of the box.

Template	Description	Use Case
Jupyter Notebook	Interactive Python environment with GPU support	Data science, experimentation, prototyping
Ollama	Local LLM inference server	Running open-source LLMs (Llama, Mistral, etc.)
Stable Diffusion	Image generation with GPU acceleration	AI art, image generation workflows
Grafana	Monitoring and visualization dashboard	Infrastructure monitoring, data visualization
NVIDIA NIM	Optimized inference microservices	Production AI inference at scale

To Quick Launch: navigate to your rented machine, click Add New Project, browse the marketplace, select a template, customize settings if needed via the builder interface, and deploy.

Custom Deployment

Custom deployments give you full control over the container configuration. You can use the Builder UI (user-friendly form) or the YAML editor (direct Kubernetes YAML) — both are synchronized in real-time.

Configuration Options

1
Project Name
A descriptive name for your deployment.
2
Docker Image
The container image in namespace:tag format. Supports Docker Hub, GitHub Container Registry (GHCR), Google Container Registry, Azure Container Registry, and private registries.
3
Ports
Map container ports to public ports. Select TCP or HTTP protocol per port.
4
Domains
Attach custom domains to your deployment or specific ports. Multiple domains can be comma-separated.
5
Environment Variables
KEY=VALUE pairs passed to the container. Use these for API keys, configuration, secrets.
6
Commands
Override the container entrypoint and arguments for custom startup behavior.
7
GPU Selection
Enable GPU access, set the GPU count, choose manufacturer (NVIDIA or AMD), and optionally specify a model.
8
Resources
Allocate CPU cores, RAM, and storage for the Kubernetes pod.

Example YAML deployment — Ollama with GPU

apiVersion: v1
kind: Pod
metadata:
  name: ollama-inference
spec:
  containers:
  - name: ollama
    image: ollama/ollama:latest
    ports:
    - containerPort: 11434
      protocol: TCP
    resources:
      limits:
        nvidia.com/gpu: 1
        cpu: "4"
        memory: "16Gi"
      requests:
        cpu: "2"
        memory: "8Gi"
    env:
    - name: OLLAMA_HOST
      value: "0.0.0.0"

Deploying a corrupted or invalid YAML will prevent the pod from starting, but the machine rental will continue running and billing. Always validate your YAML before deploying. If a pod fails, fix the configuration and redeploy — or stop the machine if you no longer need it.

Supported Docker Registries

•Docker Hub — public and private images (default registry)
•GitHub Container Registry (GHCR) — ghcr.io/namespace/image:tag
•Google Container Registry — gcr.io/project/image:tag
•Azure Container Registry — myregistry.azurecr.io/image:tag
•Private registries — any OCI-compatible registry with authentication

Local Storage (Persistent Volumes)

FluxEdge supports local persistent volumes for workloads that need data persistence during a rental session.

•A volume is created on the provider's disk and mounted to the container at a configurable path
•Data persists for the duration of the lease only — it does NOT persist across leases
•Each container gets its own unique volume (shared volumes between containers are not supported)
•Each container is limited to one persistent volume
•Use the storage type filter (e.g., NVMe) to select machines with faster drives
•Quick Launch templates have local storage enabled by default

Data stored in persistent volumes is lost when the lease ends or if the workload migrates to a different provider. Always back up important data to external storage (S3, cloud storage, etc.) before ending your rental.

Deploying Workloads