Google Cloud VM Manager vs pyinfra

Google Cloud VM Manager manages operating systems across Windows and Linux VM fleets running on Compute Engine. Its tools cover patch management, OS inventory and OS policies. Patch jobs apply updates on demand or on a schedule, report compliance, and can use selected approvals and pre- or post-patching scripts. OS inventory collects hostnames, OS and kernel versions, installed packages, available updates, Windows applications and OS vulnerabilities. OS policies centralize software configuration deployment, maintenance and reporting. Teams can use the Google Cloud console, Google Cloud CLI or OS Config API; Terraform can extend scope across projects. The service supports projects inside VPC Service Controls perimeters, but patch jobs operate on VMs in one project at a time. A free tier covers 100 VMs per Cloud Billing account per month, and limited-functionality mode is available at no cost for unlimited VMs. After the free tier, the listed charge is $0.003 per hour per VM with an active OS Config agent.

pyinfra is a free, Python-native tool for automating infrastructure across servers. It connects over SSH to run commands concurrently and idempotently, using Python deploys and inventories. Operations can describe a desired host state or issue specific commands, and `--dry` previews per-host changes before applying them. The project says it works from one host to 10,000, with parallel execution and real-time output. Target machines need a shell and SSH, but no agent, daemon, state file, or control plane. Documented connectors include SSH, Docker, Terraform, and the local machine; operations cover package managers, files, services, databases, security, and containers. It runs wherever Python runs, with Mac, Linux, and Windows listed as control systems. Python 3.10 or later is required; pyinfra 3.5.1 and later supports Python 3.10 through 3.13. The software is MIT licensed. The installation guidance recommends keeping pyinfra and its dependencies current and installing as a regular user rather than as root or with sudo. Help is directed to Matrix and Stack Overflow, while bug reports go to GitHub issues.