A cheatsheet for immutable fedora variants, usually about ostree and rpm-ostree

Finding all damaged objects. --delete flag will delete corrupted files and mark commits with damaged files as "partial", which then allow to redownload it.

ostree fsck -a

Current deployment status

rpm-ostree status

1. Boot into the last booted deployment before the currently booted deployment, and make it your default deployment

rpm-ostree rollback

2. Make a specific deployment your default deployment, 0 is the deployment index

ostree admin set-default 0

1. Download last 5 commits (parameter --commit-metadata-only is optional but makes things faster)

ostree pull --commit-metadata-only --depth 5 fedora fedora/38/x86_64/kinoite

2. (Optionally) pin current deployment if there is chance of need to return to it. 0 is index of current deployment. to list deploymends run rpm-ostree status or ostree admin status

ostree admin pin 0

3. List downloaded commits to choose the best one, basing on the timestamp or anything elese

ostree log fedora:fedora/38/x86_64/kinoite

4. Deploy selected commit using its hash (or version, like 38.20231008.0)

rpm-ostree deploy 520e744c643b85fd14817a3eb948f200e7dec902cad4157e411dfeda2c6d7aab

5. Reboot into the new deoloyment (to see it before reboot, run rpm-ostree status)

6. (Optionally) unpin previously pinned deployment (id will be 1 if it is listed as the one before currently booted one)

ostree admin pin -u 1

1. List trees of the ostree repository (fedora example)

ostree remote summary fedora

2. Rebase the system into the selected tree

rpm-ostree rebase fedora:fedora/39/x86_64/kinoite

3. reboot into the new deployment

Main tool for cleanup is rpm-ostree cleanup command. To see all options run:

rpm-ostree cleanup --help

Full cleanup (including all deployments other than currently deployed)

rpm-ostree cleanup -p -b -r -m

ostree also have cleaning tool

ostree prune

Adding package fzf to the (new) deployment

rpm-ostree install fzf

Removing packgage that was additionally installed

rpm-ostree remove fzf

Removing package that is in the system by default

rpm-ostree override remove firefox

Restore default package that was removed with override

rpm-ostree override reset firefox

Temporarily

rpm-ostree usroverlay --transient

Make the changes to /usr persist reboots

rpm-ostree usroverlay --hotfix

It is very important for performance to have trimming working. To test it run the below to trim all mounted filesystems (can take long time, like over 10,20 minutes) (if root disk is not trimmed, it have discard not enabled at some level):

fstrim -av

If trimming is not working, check the below:

• /etc/crypttab if it have discard set for root
• discard=async, discard=sync, or discard flags set in /etc/fstab
• lsblk -o +DISC-GRAN,DISC-MAX - the additional columns should have non-zero values
• cryptsetup luksDump /dev/nvme0n1p3 - if flag allow-discards is set (optional but best way to make cryptsetup allow discards by default)

Setting performance improving flags on luks volume:

cryptsetup refresh \
    --allow-discards \
    --perf-no_read_workqueue \
    --perf-no_write_workqueue \
        --persistent \
            /dev/mapper/<disk>

Arch wiki about workqueues: https://wiki.archlinux.org/title/Dm-crypt/Specialties#Disable_workqueue_for_increased_solid_state_drive_(SSD)_performance

Recommended flags for btrfs volume mounting (compression flag optional, I guess):

noatime,space_cache=v2,discard=async,compress-force=zstd:1

If for some reason space_cache is not set properly on volume mount, do this , (fs need to be unmounted):

• clear any v1 and v2 space_cache
• mount with space_cache=v2 flag
• check if flag is set correctly (maybe also create some file there?)

btrfs check --clear-space-cache v1 /dev/mapper/<device>
btrfs check --clear-space-cache v2 /dev/mapper/<device>

mount -o space_cache=v2 /dev/mapper/<device> /mnt
mount | grep space_cache
echo "hello" > /mnt/some/path/file.txt
umount /mnt