Skip to content

Atomic (transactional) update for openSUSE systems with read-write root filesystems

License

Notifications You must be signed in to change notification settings

pavinjosdev/atomic-update

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 

Repository files navigation

atomic-update ⚛️

Never let another update break your read-write openSUSE system!

Synopsis

atomic-update is a simple single file program with just one external dependency that can be used to perform atomic (transactional) updates of openSUSE systems with read-write root filesystems such as Tumbleweed and Slowroll. It can also be used on Leap, even though Leap-specific update commands are yet to be added.

atomic-update uses btrfs subvolumes and snapper snapshots to safely perform updates to a new root filesystem snapshot while minimizing any side-effects to your currently running system/snapshot.

How it works

  • On performing an update or running a command using atomic-update, a new root filesystem snapshot is created
  • The new snapshot is used to boot an ephemeral container to see which services are in a failed state, for later comparison
  • All changes are made against this new snapshot and not to the currently running system's snapshot
  • The snapshot is booted again in an ephemeral container to see if the changes broke any new services
  • If the changes are successful, the new snapshot is set as the default snapshot. The changes can be either applied live or the system rebooted into the new default snapshot
  • If the changes are unsuccessful, the new snapshot is discarded

Performing updates like this have a number of benefits:

  • Prevent a broken system due to system crash, power loss, or other unforeseen events during an update
  • Prevent updates from impacting the currently running system, causing the currently running programs, including but not limited to the desktop environment to crash
  • Prevent non-interactive updates from breaking system due to conflicts/errors causing zypper to abort (the default action on conflicts/errors) in the middle of an update
  • Prevent updates from causing an inconsistent system state due to failing scripts during an otherwise successful update
  • Avoid having to reboot into read-only grub snapshots to perform rollback

Downsides:

  • Updates must be either applied live or the system rebooted shortly thereafter to avoid losing changes made to the old root filesystem.

Acknowledgements

atomic-update is heavily inspired by the excellent transactional-update package for read-only root filesystems. All the credit goes to them 🤗

Even though transactional-update works on read-write systems as of version 4.6.0, it's not officially supported and the lead developer has stated support may be removed in the future if there are conflicts with read-only filesystem features.

Installation

  1. Install external dependency for booting snapshots in an ephemeral container to check for issues. systemd-nspawn is part of systemd and very small 👼
sudo zypper install systemd-container
  1. Install atomic-update, just a single python script you can read through in a few minutes 📜
curl -s https://raw.githubusercontent.com/pavinjosdev/atomic-update/main/atomic-update | sudo tee /usr/bin/atomic-update > /dev/null
sudo chmod 755 /usr/bin/atomic-update

Usage

Type in atomic-update --help for usage help.

Usage: atomic-update [options] command

atomic-update provides safer transactional operations
for openSUSE systems with read-write root filesystems.

Commands:
  dup                 - Perform distribution upgrade
  run <cmd>           - Run a command in a new snapshot
  rollback [number]   - Set the current or given snapshot as default snapshot

Options:
  --reboot              - Reboot after update
  --apply               - Switch into default snapshot without reboot
  --shell               - Open shell in new snapshot before exiting
  --continue [number]   - Use latest or given snapshot as base
  --no-verify           - Skip verification of snapshot
  --interactive         - Run dup in interactive mode
  --debug               - Enable debug output
  --help                - Print this help and exit
  --version             - Print version number and exit

Video examples

  1. Perform distribution upgrade on Tumbleweed/Slowroll and apply it live without rebooting

atomic-dup

  1. Test what happens when an update breaks

break-snapshot


Examples

  1. Perform distribution upgrade on Tumbleweed and reboot
sudo atomic-update --reboot dup
  1. Perform distribution upgrade on Tumbleweed and apply it live without rebooting
sudo atomic-update --apply dup
  1. Update packages on Leap and reboot
sudo atomic-update --reboot run zypper update
  1. Drop into a bash shell in a new snapshot
sudo atomic-update --shell run true
  1. Run bash script in a new snapshot and drop into a bash shell in the same snapshot afterward
sudo atomic-update --shell run bash -c 'date | awk "{print \$1}" && whoami'
  1. Troubleshoot a failing dup by running dup interactively and dropping into a bash shell afterward
sudo atomic-update --shell --interactive dup
  1. Continue making updates to the previous snapshot in a new snapshot
sudo atomic-update --shell --continue run true

Without --continue option, atomic-update would always base the new snapshot from the currently booted snapshot. Use this option to not lose changes made to a previous snapshot. Option --apply implies continue.

  1. Rollback to currently booted snapshot
sudo atomic-update rollback

Uninstallation

  1. Remove atomic-update
sudo rm /usr/bin/atomic-update
  1. Optionally, uninstall systemd-nspawn
sudo zypper remove systemd-container

Troubleshooting

Specify the --debug option for troubleshooting. atomic-update is intended to catch SIGINT (Ctrl+C) and properly cleanup. If for some reason it does not cleanup such as when receiving SIGTERM or SIGKILL, current or future operations should not be affected. atomic-update keeps its working directory in /tmp/atomic-update_*, so a reboot would always cleanup.

Known issues

  • When switching to a new snapshot without reboot using the --apply option, future updates to the bootloader (prior to a reboot) such as running update-bootloader script must be performed from a new atomic snapshot.