The idea in four steps
- Check that the app works before you touch it. Never patch something that is already failing.
- Take a snapshot so there is a way back.
- Run the updates.
- Check again. If the app still answers, keep going. If it does not, roll back to the snapshot.
- Checkthe app works now
- Snapshot
pct snapshot - Patch
apt-get upgrade - Check againsame health check
- Works: keep the snapshotdelete it after a few daysBroken: roll back
pct rollback, then start
What you need
- A Debian or Ubuntu container. The commands below use ID 105.
- Storage that supports snapshots for that container (for example ZFS, LVM-thin or Ceph).
- One check that proves the app works, such as a health URL.
Step by step
1. Check it works now.
curl -fsS https://app.example.com/health
2. Take a snapshot and wait for it to finish:
pct snapshot 105 pre-patch --description "before apt upgrade"
3. Run the updates inside the container:
pct exec 105 -- bash -c 'apt-get update && DEBIAN_FRONTEND=noninteractive apt-get -y upgrade'
4. Check again, giving services a few seconds to restart:
sleep 30
curl -fsS https://app.example.com/health
5a. It works. Keep the snapshot for a few days as a rollback point, then delete it:
pct delsnapshot 105 pre-patch
5b. It does not. Stop the container, roll back, and start it again:
pct stop 105
pct rollback 105 pre-patch
pct start 105
Limits you should know about
- A rollback loses what was written since the snapshot. That includes database rows, uploaded files and logs. For databases, rely on their own backups, not on snapshots alone.
- Some updates need a restart. A library update can take effect only after the service or the container restarts, so a check straight after the upgrade may not show the problem. Reboot, if you allow it, and check a second time.
- Do not stack old snapshots. Delete only the ones you created, and only after the check passed.
- Virtual machines use the same idea with
qm snapshot,qm rollbackandqm delsnapshot, and you run the updates over SSH or the guest agent. See theqmmanual.
Making it a monthly habit
The manual version is easy to skip. Pontavia's Proxmox Safe Patch ($59) is an n8n workflow that does exactly this. It checks your app answers, takes a snapshot, runs apt upgrade inside the guest, then checks again. If the app still works, the snapshot stays as a rollback point and older ones are cleaned up. If it broke, the guest is rolled back, started again, and you get an alert saying what failed.
It starts in report-only (dry run) mode, never patches a guest that is already failing a check, and never installs without a finished snapshot. It only deletes snapshots it made itself, and reboots only if you allow it. It was tested on Debian 12 and Ubuntu 24.04 containers, including a reboot. Virtual machines use the same steps but are not tested yet.
Frequently asked questions
Can I take a snapshot of a running Proxmox container?
Yes, on storage that supports snapshots. pct snapshot works on a running container. If your storage does not support snapshots, take a vzdump backup before patching instead.
What if my storage does not support snapshots?
Then pct snapshot will fail. Take a vzdump backup before patching and restore it if the update breaks something. It is slower but gives you the same safety net.
Should I use apt-get upgrade or dist-upgrade?
For routine patching, apt-get upgrade is the conservative choice because it does not remove packages or add new ones. dist-upgrade can change the package set and is better kept for planned upgrades.
Is this how I should patch the Proxmox host itself?
No. Host updates and reboots are a separate procedure. See the guide on finding Proxmox nodes that need updates or a reboot.
Related guides
- Find Proxmox Nodes That Need Updates or a Reboot
- How to Test That Proxmox Backups Actually Restore
- Stop Forgotten Test VMs: Proxmox Expiry Tags
- All guides
Questions about your own setup? Send a message or book a free 15-minute call.