btrfs timeshift questions on latest KDE neon

[jfabernathy]

So I did a reinstall on real hardware.

jim@nucboxg3:~$ df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 1.6G 2.1M 1.6G 1% /run
/dev/nvme0n1p2 477G 6.3G 469G 2% /
tmpfs 7.7G 0 7.7G 0% /dev/shm
tmpfs 5.0M 16K 5.0M 1% /run/lock
efivarfs 192K 99K 89K 53% /sys/firmware/efi/efivars
tmpfs 7.7G 4.0K 7.7G 1% /tmp
/dev/nvme0n1p2 477G 6.3G 469G 2% /home
/dev/nvme0n1p1 500M 7.8M 492M 2% /boot/efi
tmpfs 1.6G 68K 1.6G 1% /run/user/1000

jim@nucboxg3:~$ sudo btrfs fi df -h /
[sudo] password for jim:
Data, single: total=8.01GiB, used=5.81GiB
System, DUP: total=8.00MiB, used=16.00KiB
Metadata, DUP: total=1.00GiB, used=199.27MiB
GlobalReserve, single: total=14.95MiB, used=0.00B

Now I'm going to modify fstab to add noatime and compress=zstd to both @ and @home subvols.

after I rebooted:

jim@nucboxg3:~$ mount | grep nvme
/dev/nvme0n1p2 on / type btrfs (rw,noatime,compress=zstd:3,ssd,discard=async,spac e_cache=v2,subvolid=256,subvol=/@)
/dev/nvme0n1p2 on /home type btrfs (rw,noatime,compress=zstd:3,ssd,discard=async,spac e_cache=v2,subvolid=257,subvol=/@home)
/dev/nvme0n1p1 on /boot/efi type vfat (rw,relatime,fmask=0022,dmask=0022,codepage=437,io charset=iso8859-1,shortname=mixed,errors=remount-ro)

jim@nucboxg3:~$ df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 1.6G 2.1M 1.6G 1% /run
/dev/nvme0n1p2 477G 6.5G 469G 2% /
tmpfs 7.7G 0 7.7G 0% /dev/shm
tmpfs 5.0M 16K 5.0M 1% /run/lock
efivarfs 192K 99K 89K 53% /sys/firmware/efi/efivars
tmpfs 7.7G 4.0K 7.7G 1% /tmp
/dev/nvme0n1p2 477G 6.5G 469G 2% /home
/dev/nvme0n1p1 500M 7.8M 492M 2% /boot/efi
tmpfs 1.6G 68K 1.6G 1% /run/user/1000

jim@nucboxg3:~$ sudo btrfs fi df -h /
[sudo] password for jim:
Data, single: total=9.01GiB, used=6.08GiB
System, DUP: total=8.00MiB, used=16.00KiB
Metadata, DUP: total=1.00GiB, used=204.61MiB
GlobalReserve, single: total=15.73MiB, used=0.00B

jim@nucboxg3:~$ sudo btrfs fi defragment / -r -czstd

jim@nucboxg3:~$ df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 1.6G 2.1M 1.6G 1% /run
/dev/nvme0n1p2 477G 4.2G 471G 1% /
tmpfs 7.7G 0 7.7G 0% /dev/shm
tmpfs 5.0M 16K 5.0M 1% /run/lock
efivarfs 192K 99K 89K 53% /sys/firmware/efi/efivars
tmpfs 7.7G 4.0K 7.7G 1% /tmp
/dev/nvme0n1p2 477G 4.2G 471G 1% /home
/dev/nvme0n1p1 500M 7.8M 492M 2% /boot/efi
tmpfs 1.6G 68K 1.6G 1% /run/user/1000

jim@nucboxg3:~$ sudo btrfs fi df -h /
Data, single: total=9.00GiB, used=3.71GiB
System, DUP: total=8.00MiB, used=16.00KiB
Metadata, DUP: total=1.00GiB, used=218.52MiB
GlobalReserve, single: total=16.11MiB, used=0.00B

Not sure I understand the df -h and btrfs fi df -h / output, but it went down this time.

[jfabernathy]

I found something over on the Fedora forum. where a person had the same 1 minute delay issue.

This happens probably because of the notification that Timeshift tries to show (but it doesn’t work), and it stalls the process for a bit.

Notice when you take a snapshot, a notify-send process is run as root:

notify-send -t 10000 -u low -i gtk-dialog-info TimeShift BTRFS Snapshot saved successfully (0s) -h string:desktop-entry:timeshift-gtk

This stalls the timeshift process until the notify-send process ends.

What is puzzling to me is when I test this on Linux Mint 22, which is based on Ubuntu 24.04, timeshift has no delay. Of course the Mint folks now own timeshift along with Cinnamon

[jfabernathy]

jim@nucboxg3:~$ sudo btrfs fi df /
[sudo] password for jim:
Data, single: total=20.01GiB, used=17.14GiB
System, DUP: total=8.00MiB, used=16.00KiB
Metadata, DUP: total=1.00GiB, used=708.86MiB
GlobalReserve, single: total=49.41MiB, used=0.00B
jim@nucboxg3:~$ df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 1.6G 2.1M 1.6G 1% /run
/dev/nvme0n1p2 460G 19G 441G 5% /
tmpfs 7.7G 0 7.7G 0% /dev/shm
tmpfs 5.0M 16K 5.0M 1% /run/lock
efivarfs 192K 99K 89K 53% /sys/firmware/efi/efivars
tmpfs 7.7G 3.6M 7.7G 1% /tmp
/dev/nvme0n1p2 460G 19G 441G 5% /home
/dev/nvme0n1p1 300M 7.8M 292M 3% /boot/efi
tmpfs 1.6G 92K 1.6G 1% /run/user/1000

Next time I do an install, I use both commands before I compress.

[oshunluvr]

You should use: sudo btrfs fi df /

[jfabernathy]

I miss typed.

df -h

is what I used

[oshunluvr]

My first suspicion is using "du" instead of "sudo btrfs fi du -s <subvolume>".

Did the whole file system increase in size or just the subvolume?

Try the btrfs du on all your subvolumes and have a look at where the space is used. Maybe the forced compression broke reflinks??

cd into your subvolume and/or snapshot folders and run:

sudo btrfs fi du -s *

[jfabernathy]

Thanks, if I do manual snapshots on this system it's instantaneous. it seems to be some delay in timeshift on *buntu.

The size I compare with du -h.

This is a new install so I'm assuming scub/balance is not an issue at this point.

[oshunluvr]

Not seeing this here. Snapshots take nearly zero time to make or delete. Sounds like a timeshift problem to me. Not sure why fstab options would cause a timeshift delay. I've never used timeshift.

My / mount options show as: rw,noatime,compress=lzo,ssd,discard=async,space_ca che=v2,autodefrag,subvolid=901,subvol=/@KDEneon

I have no guess about the extra space usage because you didn't say how you measured it. If you used sudo btrfs fi du -s <subvolume> to check space used then you're likely correct. Otherwise, probably not.

Also the type of compression you used could make a difference also. Some files are not compressible and actually can get larger when compression is forced. Although I doubt you'd eat up 6G of space that way.

Have you run scrub and/or balance lately?

[jfabernathy]

I was looking to test Plasma 6.2.3 on a *buntu system without snap. It seems that Neon has snap but no packages are installed. I did a BTRFS install so I have @ and @home subvolumes. I installed a few flatpak packages I like and also installed timeshift with apt.

What I noticed was taking any snapshot in timeshift take 1 minute. Deleting a snapshot takes 1 minute. I don't know what is causing this since on Debian 12 and Archlinux timeshift snapshots create almost instantaneously when using btrfs.

I tried setting up the fstab for compress and noatime, and then compressing "/". That made no change in timeshift timing.

However, it increased the amount of disk space used by / from 13G to 19G. On Endeavour OS (Archlinux) it goes down as you'd expect.

Anyone has an answer for this?