Conversation

Jarkko Sakkinen

WiP: Archest Linux (EXT4 + LUKS2 - LVM2): https://codeberg.org/jarkko/archest-linux/src/tag/0.1.0-rc1 Boots to login and only minor glitches still left to fixup before tagging 0.1 šŸ³

I like how unlayered this is, i.e. at most two subsystems layered and stack is at its heaviest a file system + LUKS2 volume (i.e. no one to many relationships). Less risk of busy file systems that cannot be unmounted at least :-)
1
0
1

@jarkko I don't see why one would avoid LVM when using EXT4. The original post discusses btrfs not needing the volume management, but LVM + LUKS + EXT4/XFS is a fantastically stable system (I know, I use it on dozens of servers every day).

1
0
0
@mikebabcock Yep, breakin' the law I guess šŸ˜… But you know this WFM me best...
1
0
0
@mikebabcock and had no other choice than arch as ext4 is not favoured with this oddball combination choice of modern and legacy features. I do e.g. use snapshots but for that I use incremental backups to my NAS rather than pile them all over the place šŸ¤·ā€ā™€ļøšŸ’£šŸ„²
1
0
0
@mikebabcock Iā€™m going to use lvm2 after all. There is a useful commands depending on it: e2scrub. So it is a constraint then.
1
0
1

@jarkko I primarily use LVM so I don't need to allocate all my disk space up front, and to give me access to LVM cache. I use xfs for my filesystems so I have xfsdump for consistent backups.

1
0
1
@mikebabcock I also realized that it makes sense to have swap as LVM2 volume because I have essentially two different swap configurations depending on use and purpose: 2GB (non-hibernate) and 60-80GB (hibenate). LVM2 will help in this case later on tune between these choices a bit...
1
0
0
@mikebabcock I'm doing this partly to research what would be optimal configuration to enable hibernate, and possibly try out encrypted hibernate patch set, which never landed to the mainline. So I'm doing both reconfiguration for my host system and designing something for test target VM's at the same time :-)

One observation is that probably it is better to encrypt per logical volume than encrypt the physical volume because then the partition is reachable for so called restore kernel. I've previously encrypted the physical volume and created logical volumes inside it.
1
1
1

@jarkko I don't like encrypting anything with too much well-known structure in case it becomes an attack vector, so I do the same; disk -> lvm -> luks -> xfs in my case. I tend to not encrypt root filesystems on servers for recovery reasons, and often use "real" partitions for /boot and swap, again to improve recovery options.

1
0
0
@mikebabcock i have zero plans to use hibernation in my hosts but i'm interested to test it as a kernel feature :-) and generally i want a common image that works for my host system and test targets (usually VM's ,sometimes NUC's).
0
0
1