Skip to content

VM troubleshooting

Find your VM in Virtual Machine / Instances. Open its details to check the status and any error message, then find the matching problem below.

A VM does not start

Read the error in the instance details. Check that your wallet has enough credits for the plan you selected. For a GPU VM, enough GPUs must also be available for that plan. Adding credits does not reserve GPUs.

If the VM is still starting or stopping, wait for that operation to finish before trying again. Check its final status and use Start or Stop if needed. Avoid creating another VM just to retry the same request. If the VM still cannot start, contact support with its instance ID and error message.

The VM's disk can remain in your account after a failed start, and storage charges continue while you keep it. See Data and charges before deciding whether to keep or delete the VM.

SSH fails

Check that you are using the VM's public IPv4 address, the root username, and the password or SSH key you set for that VM. See Connect to a VM for the connection commands.

Use the SSH error message to choose what to check:

SSH errorWhat it meansWhat to check
Connection timed outSSH did not receive a responseCheck the VM's status and address. Open Console to check its network and firewall.
Connection refusedThe connection to the SSH port was rejectedOpen Console and check whether SSH is running and which port it uses.
Permission deniedSSH could not log you inCheck the username, password or selected key, and whether the VM allows that login method.
Host key changedThe server's identity differs from beforeCheck whether you rebuilt the VM or its address changed. Verify the VM's identity before accepting the new key.

If you can log in through Console, run these commands inside the VM to check Ubuntu's SSH service and network settings. They do not change any settings.

bash
systemctl is-active ssh.service
ip -brief address
ip route
  • The first command prints active when the SSH service is running. If it prints inactive or failed, the service is not running.
  • ip -brief address lists the network interfaces and their IP addresses. Check that the VM's network interface has an address.
  • ip route lists the network routes. A route beginning with default tells the VM where to send traffic outside its local network.

If you use a custom SSH port, include it with ssh -p PORT. Replace PORT with the port you configured. Keep Console open while checking or changing network, firewall, or SSH settings so you still have a way to access the VM. If you need help with the results, send support the command output and SSH error.

The console does not open

Check that the VM is powered on and allow popups for console.remotegpu.ai. Then select Console again to open a new session. Start a stopped VM before opening its console.

If the console opens but there is no login prompt, look for startup errors on the screen and send them to support. Rebuilding replaces the VM's root disk and deletes its data; use it only if you intend to start with a fresh OS.

VM stays on Powered On

Your VM is turned on, but it may still be completing its initial setup. Wait a few minutes, then try logging in through SSH or Console.

If you can log in but the status stays on Powered On, check the guest agent. This is a background service inside your VM that reports its status to RemoteGPU. Run these commands inside the VM to check whether the agent is running and Ubuntu has finished setting up the VM:

bash
systemctl is-active qemu-guest-agent.service
cloud-init status
  • The first command prints active when the guest agent is running. inactive or failed means it is not running.
  • The second command checks cloud-init, which configures the VM during startup. running means setup is still in progress; wait and check again. done means it has finished. Include any reported errors when contacting support.

If the guest agent is not running, or the status stays on Powered On after setup finishes, contact support with the output of both commands.

The console shows Running once the guest agent connects. This status does not check whether SSH or GPU drivers work; use the SSH checks above or the GPU checks below if you still have trouble with them.

Password reset stays pending

If your password reset is still pending, the new password may not work yet. Open Edit and check the status under Reset root password in Danger Zone. Start the VM if it is stopped, and follow any restart instructions shown there.

RemoteGPU uses the guest agent to change the password inside your VM. If that service is missing or not running, the reset cannot finish. Use the guest agent checks above if you can still log in.

Try logging in with the new password to confirm the change. If the reset remains pending or reports an error, contact support with its status and error message.

Storage is smaller than expected

Run lsblk inside the VM to see the disk and partition sizes, then run df -h / to check the filesystem that holds your OS and files. If the disk shows the new size but the partition or filesystem is still smaller, it may need to be expanded too. Follow Increase storage.

GPU is unavailable after a plan change

Changing from a CPU plan to a GPU plan keeps your existing OS and does not install a GPU driver. If the plan change finished but your application cannot use the GPU, follow GPU drivers to check for a driver and install one if needed.

If the VM is still waiting for available GPUs, it must start before you can check the driver inside it. Installing a driver will not resolve a lack of available GPUs.

Contact support

Select Contact Support in the RemoteGPU console. Include:

  • The VM name, selected plan, and instance ID. You can find the ID in the URL when viewing the VM's details.
  • What you tried to do and when, including your timezone.
  • The status and exact error message you see.
  • For a software problem inside the VM, the OS and kernel versions and the output of the relevant checks above.
  • Any plan change, OS upgrade, or driver installation made before the problem started.

Remove passwords, private keys, tokens, and application secrets from anything you share. Send the relevant error messages instead of the VM's entire log.

RemoteGPU customer documentation