Joblet is a secure Linux job execution platform that simplifies running isolated processes with comprehensive resource management. It handles process isolation, resource limits, GPU acceleration, and monitoring automatically, letting you focus on your applications rather than infrastructure complexity.
Complete Documentation: /docs/README.md
Provides sandboxed code execution with GPU support, resource controls, and complete process isolation for AI workloads.
- 🔒 Isolation: Run in secure sandboxes, separate namespaces, preventing interference between workloads
- ⚡ GPU Acceleration: Native NVIDIA GPU support with automatic CUDA env. setup for AI/ML workloads
- 🎛️ Resource Management: CPU, memory, I/O, and GPU resource limits with fine-grained control
- 🌐 Network Isolation: Networking options from full isolation to custom networks for secure communication
- 💾 Storage Solutions: Persistent volumes for data retention and temporary storage with automatic cleanup
- ⏰ Job Scheduling: Execute jobs immediately, schedule for specific times, or set delayed execution
- 🛡️ Enterprise Security: mTLS encryption with certificate-based authentication and role-based access control
- 🐍 Runtime Environments: Build Python ML, Java, and custom runtime environments from declarative YAML specifications
☁️ Cloud Deployment: Joblet can be deployed directly on AWS EC2 with automated installation DynamoDB and CloudWatch integration. See the AWS Deployment Guide for one-click deployment with user data scripts.
# Download and install the latest release
wget $(curl -s https://api.github.com/repos/ehsaniara/joblet/releases/latest | grep "browser_download_url.*_amd64\.deb" | cut -d '"' -f 4)
sudo dpkg -i joblet_*_amd64.deb
# Start the service
sudo systemctl start joblet
sudo systemctl enable joblet# Download and extract the release
curl -L -o rnx-linux-amd64.tar.gz $(curl -s https://api.github.com/repos/ehsaniara/joblet/releases/latest | grep "browser_download_url.*linux-amd64.tar.gz" | cut -d '"' -f 4)
tar -xzf rnx-linux-amd64.tar.gz
cd rnx-linux-amd64
# Run the installation script
sudo ./install.sh
sudo systemctl start joblet# Add the Joblet tap
brew tap ehsaniara/joblet https://github.com/ehsaniara/joblet
brew install rnx # Installs RNX CLI
# Copy configuration from server
scp user@joblet-server:/opt/joblet/config/rnx-config.yml ~/.rnx/# Download the Windows binary
Invoke-WebRequest -Uri "https://github.com/ehsaniara/joblet/releases/latest/download/rnx-windows-amd64.exe" -OutFile "rnx.exe"
# Add to PATH
New-Item -ItemType Directory -Force -Path "$env:USERPROFILE\bin"
Move-Item rnx.exe "$env:USERPROFILE\bin\"
[Environment]::SetEnvironmentVariable("Path", $env:Path + ";$env:USERPROFILE\bin", [EnvironmentVariableTarget]::User)
# Copy configuration from Joblet server to %USERPROFILE%\.rnx\rnx-config.yml# Test the connection and run your first job
rnx job list # List current jobs
rnx job run echo "Hello from Joblet!" # Execute a simple test job
rnx monitor status # Check server health
# Install: pip install joblet-sdk
from joblet import JobletClient
with JobletClient(host="your-joblet-server") as client:
# Run a job programmatically
job = client.jobs.run_job(
command="echo",
args=["Hello from Python!"],
name="python-job"
)
print(f"Job started: {job['job_uuid']}")# Execute pre-configured workflows
rnx workflow run jobs.yaml:ml-analysis # Python ML analysis
rnx workflow run jobs.yaml:hello-joblet # Java application
rnx workflow run ml-pipeline.yaml # ML pipeline with GPU support
# Monitor execution status
rnx job status <job-uuid>
rnx workflow status <workflow-uuid># Basic job execution with resource isolation
rnx job run echo "Hello World"
rnx job run --max-cpu=50 --max-memory=512 python3 script.py
# Build and use runtime environments
rnx runtime build ./examples/python-3.11-ml/runtime.yaml # Build Python ML runtime
rnx runtime build ./examples/java-21/runtime.yaml # Build Java 21 runtime
rnx job run --runtime=python-3.11-ml python analysis.py # Use Python ML runtime
rnx job run --runtime=openjdk-21 java App.java # Use Java runtime
# GPU acceleration for compute-intensive workloads
rnx job run --gpu=1 --gpu-memory=8GB python train_model.py
rnx job run --gpu=2 --runtime=python-3.11-ml python distributed_training.py
# File uploads and environment configuration
rnx job run --upload=data.csv --upload=script.py python3 script.py
# Environment variables (visible and secure)
rnx job run --env=NODE_ENV=production node app.js
rnx job run --secret-env=API_KEY=secret python app.py # Hidden from logs
# Network isolation modes
rnx job run --network=none secure_task.sh # No network access
rnx job run --network=isolated wget https://api.com # External access only
rnx job run --network=bridge api_server.py # Inter-job communication
# Persistent storage management
rnx volume create mydata --size=1GB
rnx job run --volume=mydata python3 process_data.py
# Job scheduling options
rnx job run --schedule="1hour" backup.sh
rnx job run --schedule="2025-12-25T00:00:00" maintenance.pyThe Joblet Admin UI is available as a standalone package at joblet-admin repository:
- System Monitoring: Real-time CPU, memory, disk, and network metrics with visual dashboards
- Job Management: Comprehensive job listing, filtering, control, and log streaming capabilities
- Workflow Visualization: Interactive dependency graphs and execution timelines
- System Administration: Intuitive volume, network, and runtime environment management
- Direct gRPC: Connects directly to Joblet server via gRPC
To use the admin interface:
# Clone the joblet-admin repository
git clone https://github.com/ehsaniara/joblet-admin
cd joblet-admin
# Install and run
npm install
npm run dev
# Access at http://localhost:3000Learn more: See the Admin UI Documentation
- Operating System: Linux kernel 4.6+ (Ubuntu 18.04+, CentOS 8+, or equivalent distributions)
- Hardware: Multi-core CPU, 1GB+ RAM, 5GB+ disk space (scales with workload)
- Network: Port 50051 accessible for gRPC communication
- Privileges: Root access required for namespace and cgroup management
- Dependencies: cgroups v2, iptables, iproute2, bridge-utils (standard on most distributions)
- GPU Support: NVIDIA drivers (optional, required only for GPU workloads)
- Operating Systems: Linux, macOS 10.15+, Windows 10+
- Resources: Minimal - 50MB+ RAM, network connectivity to server port 50051
- Distribution: Single binary executable with no additional dependencies
- Quick Start Guide - Get up and running quickly with step-by-step instructions
- Installation Guide - Comprehensive installation procedures for all platforms
- AWS Deployment Guide - Deploy Joblet on AWS EC2 with CloudWatch integration
- GPU Support Guide - Complete guide to NVIDIA GPU acceleration and CUDA integration
- Runtime System Guide - Build and manage runtime environments from YAML specifications
- Job Execution Guide - Detailed job management and execution patterns
- RNX CLI Reference - Complete command reference with examples
- Security Guide - Security configuration and best practices
- Working Examples - Production-ready workflow templates and runtime configurations
Developing a custom client? Start with the protocol definitions to generate language-specific bindings.
MIT License - see LICENSE file for details.