Imported from jeremylongshore/tons-of-skills-marketplace (plugins/saas-packs/vastai-pack/skills/vastai-upgrade-migration/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill vastai-upgrade-migration. Copyright stays with the author (MIT).

Vast.ai Upgrade & Migration

Current State

!vastai --version 2>/dev/null || echo 'vastai CLI not installed' !pip show vastai 2>/dev/null | grep -E "^(Name|Version)" || echo 'N/A'

Overview

Upgrade the Vast.ai CLI and Python SDK, handle API changes, and migrate between GPU configurations. The CLI is distributed via PyPI as vastai and tracks the REST API at cloud.vast.ai/api/v0.

Prerequisites

• Current vastai CLI installed
• Active instances inventory documented
• Backup of any custom scripts using the API

Instructions

Step 1: Check Current Version and Upgrade

# Check installed version
vastai --version
pip show vastai | grep Version

# Upgrade to latest
pip install --upgrade vastai

# Verify upgrade
vastai --version
vastai show user  # Verify auth still works

Step 2: Detect Breaking Changes

# Compare CLI help output before and after upgrade
import subprocess

def get_cli_commands():
    result = subprocess.run(["vastai", "--help"], capture_output=True, text=True)
    commands = set()
    for line in result.stdout.split('\n'):
        stripped = line.strip()
        if stripped and not stripped.startswith('-') and not stripped.startswith('usage'):
            cmd = stripped.split()[0] if stripped.split() else ""
            if cmd.isalpha():
                commands.add(cmd)
    return commands

# Run before and after upgrade to detect removed commands

Step 3: API Version Migration

# The REST API is at v0 — if Vast.ai introduces v1, update base URL
OLD_BASE = "https://cloud.vast.ai/api/v0"
NEW_BASE = "https://console.vast.ai/api/v0"  # Alternative endpoint

# Test both endpoints
import requests
for base in [OLD_BASE, NEW_BASE]:
    try:
        resp = requests.get(f"{base}/users/current",
                           headers={"Authorization": f"Bearer {api_key}"})
        print(f"{base}: {resp.status_code}")
    except Exception as e:
        print(f"{base}: {e}")

Step 4: Docker Image Updates

# Update GPU workload images to latest CUDA
# Old: pytorch/pytorch:1.13-cuda11.7-runtime
# New: pytorch/pytorch:2.2.0-cuda12.1-cudnn8-runtime

# Test new image locally before deploying
docker pull pytorch/pytorch:2.2.0-cuda12.1-cudnn8-runtime
docker run --rm pytorch/pytorch:2.2.0-cuda12.1-cudnn8-runtime python -c "import torch; print(torch.__version__)"

# Verify CUDA compatibility with target GPU hosts
vastai search offers 'cuda_max_good>=12.1 num_gpus=1' --limit 5

Step 5: Post-Upgrade Verification

#!/bin/bash
set -euo pipefail
echo "Post-upgrade verification..."

vastai show user && echo "  Auth: OK"
vastai search offers 'num_gpus=1 rentable=true' --limit 1 --raw | python3 -c "import sys,json; offers=json.load(sys.stdin); print(f'  Search: OK ({len(offers)} offers)')"
vastai show instances && echo "  Instances: OK"

echo "Upgrade verified."

Output

• CLI upgraded to latest version
• Breaking changes identified
• API endpoint compatibility verified
• Docker images updated to latest CUDA
• Post-upgrade verification passed

Error Handling

Resources

• vastai PyPI
• vast-cli GitHub
• Vast.ai Documentation

Next Steps

For CI/CD integration, see vastai-ci-integration.

Examples

Safe upgrade: Pin the current version in requirements.txt, upgrade in a test environment, run the verification script, then update production.

CUDA migration: Move from CUDA 11.7 to 12.1 by updating Docker images and filtering offers with cuda_max_good>=12.1.