- #Linux
- #Fedora
- #Arch Linux
- #Checklist
When I moved my dev box from Arch to Fedora, I spent the first few days on the stuff you can see: the desktop, keybinds, the browser, the editor. All of that came over fine.
The stuff that broke was quiet. Nothing errored, nothing popped up, and everything looked fine until I went looking.
Here's what bit me and how to check for it before it bites you.
1. Your timers didn't come with you
On Arch I had a set of systemd user timers doing the boring, important work, like backups and health checks.
I moved my home folder and my dotfiles. I did not move the units in ~/.config/systemd/user/ and the scripts they called. Nothing failed, because nothing was scheduled to fail. It was days before I noticed my backups had stopped.
Check it now, on the old box:
systemctl --user list-timers --all
systemctl list-timers --all
crontab -lWrite that list down before you wipe anything. Then on the new box, run the same commands and compare. A missing backup timer will never send you a warning.
2. The PATH your timers see isn't your PATH
Once the timers were back, some of them failed anyway. A user unit doesn't get your shell's PATH. On Fedora, systemd's user PATH is:
~/.local/bin:/usr/local/bin:/usr/bin:/binMy tools lived in ~/bin, which works fine in a terminal and is invisible to a timer. I symlinked each tool a timer calls into ~/.local/bin. You could also set Environment=PATH=... in the unit.
Test it like the timer will run it: systemctl --user start thing.service, then journalctl --user -u thing. Running the script by hand in your shell tells you nothing.
3. The default file limit will quietly shrink your work
The default soft limit for open files is 1024. A scan job of mine hit that limit, skipped hundreds of files, and then exited like it had scanned everything. A green check that only covered part of the job.
# ~/.config/systemd/user/dep-scan.service.d/limits.conf
[Service]
LimitNOFILE=65536If a job touches a lot of files, set the limit on purpose and check the count of what it actually processed, not just the exit code.
4. Postgres doesn't like a glibc downgrade
A rolling release had put a newer glibc on Arch than Fedora ships, and Postgres text sorting comes from glibc's collation rules. Copying the raw data directory across a collation change is a good way to end up with indexes that quietly return the wrong rows.
The safe move is a logical dump and restore:
pg_dump -Fc mydb > mydb.dump # on the old system
pg_restore -d mydb mydb.dump # on the new oneThen verify it. I compared exact row counts table by table between old and new. A restore that "worked" and a restore that has every row are two different things.
5. Your notification tool is gone
My alert scripts called dunstify. Fedora didn't have dunst, so the alerts went nowhere. A two-line shim that calls notify-send fixed it. Grep your scripts for the notification tool they call, and make sure it exists on the new box.
6. Symlinks still point at the old disk
A folder can look like it moved when it's really a symlink back to the old disk. Everything keeps working, so nothing tells you.
find ~ -maxdepth 2 -type l -printf '%p -> %l\n'Look at where every top-level link points. Anything pointing at the old disk still lives there, and it goes with the disk when you wipe it.
7. Inventory the old disk before you wipe it
Before I touched the Arch disk, I listed everything on it that wasn't already on the new system. That came to about 64 GB I would have lost. Most of it is the stuff that's easy to forget: hidden folders in your home directory, app and browser profiles, game saves, repos with uncommitted work, and local databases.
Do that inventory before you decide what dies. Copy what matters, verify each copy, and only then wipe.
The short version
Your desktop will tell you when it's broken. Background jobs won't. Moving to a new distro means moving the scheduled work along with the files, so check these before you call the move done:
- Timers and cron listed on the old box, all present on the new one
- Each timer run once through systemd, logs read
- PATH and file limits set in the units that need them
- Databases moved by dump and restore, row counts compared
- Alert and notification commands exist
- No symlinks pointing at the old disk
- Old disk inventoried and verified before the wipe
Why I made the move in the first place is in why my dev box runs Fedora now.
- Linux
- Fedora
- Arch Linux
- Checklist
