Why Auto Updates Kill GPU Host Uptime (and How to Disable Them)

Ubuntu ships with unattended upgrades turned on. For a laptop, that is a sensible default. For a GPU host with renters on board, it is a time bomb: a background update can replace your NVIDIA driver, restart Docker, or reboot the machine while a customer is mid-job. Here is why hosts turn it off, and how to do it cleanly.

What unattended upgrades actually do

The unattended-upgrades package checks for security updates daily and installs them without asking. Two side effects matter for a rental machine:

Service restarts. Package hooks can restart Docker, containerd, or the display stack. Every running rental on the machine dies with them.
Kernel and driver drift. A new kernel installed overnight will not match your NVIDIA driver modules until the next reboot. Your GPUs can vanish from nvidia-smi, and the marketplace marks the machine as broken.

Renters rate reliability with their wallets. A machine that drops jobs earns a worse reliability score, which pushes it down the search results and costs you future rentals. One surprise update can undo weeks of good uptime.

Turn it off

Three steps, about two minutes. You need sudo access on the host.

1Stop and disable the timers that schedule the daily runs:

sudo systemctl stop unattended-upgrades
sudo systemctl disable unattended-upgrades
sudo systemctl mask apt-daily.timer apt-daily-upgrade.timer

2
Tell apt not to run automatic updates at all:
/etc/apt/apt.conf.d/20auto-upgrades
```
APT::Periodic::Update-Package-Lists "0";
APT::Periodic::Unattended-Upgrade "0";
```
3
Verify nothing is scheduled anymore:
```
systemctl list-timers | grep apt
```
No output means you are done.

You still need updates, on your schedule

Turning off automatic updates does not mean never updating. It means you choose the moment. Patch when the machine has no active rentals, and reboot immediately after a kernel or driver update so problems surface while you are watching.

For the full walkthrough with explanations of each file, see the disable unattended upgrades guide in our open source Linux tools collection.

Patch on purpose instead

A simple monthly routine replaces the automation you just removed:

Pick a low-demand window. Check your rental history for the quietest weekday hour.
Pause new rentals before you start, so a renter cannot grab the machine halfway through your maintenance.
Run sudo apt update && sudo apt upgrade, reboot, and confirm nvidia-smi sees every card before you go back on the market.

The uptime mindset

Every practice in this guide serves one number: rented hours. Renters cannot buy time on a machine that is rebooting, and the marketplace remembers instability. The hosts who earn most treat their rigs like production servers. Updates are scheduled, changes are tested, and nothing happens on the machine that the host did not decide.

The same logic applies to pricing. A machine with perfect uptime still earns nothing while it sits idle at the wrong price. Once your host is stable, read our guide on why utilization beats listing price to put those rented hours to work.