Question 1

Documentation

Accepted Answer

Official Documentation

For Vast.ai official documentation, go to https://vast.ai/docs.

If you are just getting started, read our QuickStart Guide.

CLI documentation

Vast.ai has a command line interface written in Python. The Python script calls our API. For the latest documentation on the CLI, use the --help command.

To install the CLI, you will need to authorize your API key. Our documentation in the product uses pip. The Python script has it's own GitHub repo.

API Documentation

The complete API documentation is located below the CLI documentation on the CLI docs page. You can see the indivudal functions available on the left hand navigation.

Question 2

Instance is in scheduling status, but the same machine is available for renting

Accepted Answer

Observed Behavior:

The instance is in scheduling status
However, the machine is still shown as Available for rent on the portal (Machine ID: 37485).

Explanation:

When an instance is created, it is bound to a specific GPU ID. If the instance is stopped, it remains bound to the same GPU ID and waits for that GPU to become available again.
On machines with more than 1 GPU, other GPUs may still be free and available for allocation. As a result, the same machine can still appear as available on the portal, even though the original instance is stuck in Scheduling while waiting for its assigned GPU.

Suggestion:

We recommend using volume storage to persist data and enable access across multiple instances on the same host machine. If an instance gets stuck in Scheduling but other GPUs on the same host are available, you can create a new instance on that host and mount the same volume to continue working on it.

Question 3

Feasibility of Machine Hosting on WSL

Accepted Answer

Why You Can't Host on Vast.ai Using WSL: Technical Constraints and GPU Virtualization Limits

Q: Documentation

Official Documentation For Vast.ai official documentation, go to <a href="https://vast.ai/docs">https://vast.ai/docs.</a> If you are just getting started, read our <a href="https://vast.ai/docs/overview/quickstart">QuickStart Guide</a>. CLI documentation Vast.ai has a command line interface written in Python. The Python script calls our API. For the latest documentation on the CLI, use the --help command. To install the CLI, you will need to authorize your API key. Our <a href="https://cloud.vast.ai/cli/">documentation </a>in the product uses pip. The Python script has it's own <a href="https://github.com/vast-ai/vast-python">GitHub repo</a>. API Documentation The complete API documentation is located below the CLI documentation on the <a href="https://vast.ai/docs/cli/quickstart">CLI docs page</a>. You can see the indivudal functions available on the left hand navigation.

If you are looking to monetize your GPU hardware on Vast.ai, you might be tempted to use Windows Subsystem for Linux (WSL2) to avoid leaving the Windows ecosystem. However, according to the official Vast.ai hosting requirements, a native Ubuntu installation is mandatory. This isn't just a "recommendation"—it is a result of fundamental technical barriers within the WSL2 architecture, particularly regarding how it handles GPU resources and low-level system hooks.

${color}[#000000](1. The Virtualization Barrier: Para-virtualization vs. Native Access)

WSL2 operates on a highly optimized version of the Hyper-V utility VM. While it provides a Linux kernel, it uses GPU Para-virtualization (GPU-PV). In a native Ubuntu setup, the Nvidia driver communicates directly with the hardware. In WSL2, the "driver" inside Linux is actually a projection of the Windows host driver (dxgkrnl).

The Problem: Vast.ai’s hosting daemon requires NVML (Nvidia Management Library) to have direct, bare-metal access to the GPU to monitor temperatures, power limits, and clock speeds.
The Limitation: Under WSL, many NVML functions are restricted or report inaccurate data because the hardware is being managed by the Windows Windows Display Driver Model (WDDM), not the Linux native kernel.

${color}[#000000](2. Docker-in-Docker and Cgroup Constraints)

Vast.ai functions by running customer w

Question 4

Expanding /var/lib/docker Storage on a Vast.ai Host

Accepted Answer

For: Vast.ai host operators who need more space for /var/lib/docker. There are two different approaches, depending on your current setup. Pick the one that matches your situation: | Situation | Approach | Involves copying data? | Risk | | /var/lib/docker is a loop-mounted image file (e.g. /docker-xfs.img) sitting on a partition that already has enough free space | Option A: Grow the image file in place | No | Lower | | /var/lib/docker is its own separate disk/partition and you're moving it to a brand-new physical disk | Option B: Migrate to a new disk | Yes (full rsync copy) | Higher |

Both options share the same first three moves: confirm the machine is idle, stop/mask the same four services, and back up /etc/fstab. Those shared steps are written out once below, then each option continues on its own path.

Step 0: Confirm the machine is idle

"Idle" means no active rental, not just a container that happens to be stopped. A stopped container can still belong to an active rental, and the renter can start it again at any moment, including mid-migration. Check this on the machine's card in the Vast.ai console: look at the Occ, #Running, and #Stored fields. All of them need to read 0 (e.g. Occ: x, #Running: D: 0, I: 0, R: 0, #Stored: D: 0, I: 0, R: 0) before you can be sure there's truly no instance on the machine.

If you must proceed while a rental is still technically active (even if stopped), notify the renter of the maintena

Question 5

Alternative method to back up data using Rclone

Accepted Answer

What is Rclone?

Rclone is a command-line program to manage files on cloud storage. It is a feature-rich alternative to cloud vendors' web storage interfaces. Over 70 cloud storage products support rclone, including S3 object stores, business & consumer file storage services, and standard transfer protocols. Official website: https://rclone.org/

Why Rclone?

The following links will take you to Vast.ai, which already offers official documents for copying from and to multiple cloud storages: https://docs.vast.ai/documentation/instances/storage/cloud-sync https://docs.vast.ai/documentation/instances/storage/cloud-backups

This article discusses the benefits of using Rclone as an alternate data backup method.

How can we utilize Rclone in Vast.ai instances?

The majority of Vast.ai's templates already have rclones installed. Open Jupyter Terminal and enter "rclone version" to confirm the version.

(main) root@C.xxxxxxx:/workspace$ rclone version
rclone v1.60.1-DEV
- os/version: ubuntu 24.04 (64 bit)
- os/kernel: 6.8.0-59-generic (x86_64)
- os/type: linux
- os/arch: amd64
- go/version: go1.22.2
- go/linking: dynamic
- go/tags: none

For some cloud storages, the old version of Rclone may not work, so it is suggested to upgrade rclone to the latest version using this command: sudo -v ; curl https://rclone.org/install.sh | sudo bash

``` (main) root@C.xxxxxxxx:/workspace$ sudo -v ; curl https://rclone.org/

Question 6

Hosts cannot search for their machines

Accepted Answer

Issue

When a host searches for their own machine, the machine doesn't appear in the list (both CLI and portal)

However, when searching, other users can see the machine.

Root cause

The machine has an active instance of the host account. To get the machine to show up in the search list, that instance needs to be destroyed.

Featured & Frequently Read Articles

Browse All Categories