Am 08.04.2024 um 23:08 schrieb Stefan Monnier: > David Christensen [2024-04-08 11:28:04] wrote: >> Why LVM? > > Personally, I've been using LVM everywhere I can (i.e. everywhere > except on my OpenWRT router, tho I've also used LVM there back when my > router had an HDD. I also use LVM on my 2GB USB rescue image). > > To me the question is rather the reverse: why not? > I basically see it as a more flexible form of partitioning.
As an LVM-newbie (never used it before, i am more familar with ZFS), i did already collect quite a bit of misconceptions of mine/design problems with lvm. Therefore i would rather renew the question: Why? Just one example: In order to be able to use thin snapshots on my root partition, i did every thing i could, to have it inside a thinpool... until i noticed some weird problems booting from it (attributed to grub), so i setup a /boot outside, but the problems stayed (due to lvm's limitations). I came to use it to gain some flexibility (although it is an experiment) and found myself setting up zfs for its data integrity + flexibility, just to have a quality backup of the lvm-volume(s) on a zfs pool. > > Even in the worst cases where I have a single LV volume, I appreciate > the fact that it forces me to name things, isolating me from issue > linked to predicting the name of the device and the issues that plague > UUIDs (the fact they're hard to remember, and that they're a bit too > magical/hidden for my taste, so they sometimes change when I don't want > them to and vice versa). Even GPT brings you the chance to name hings (like part_label), only it does not force you. But i have been using that for 10+ years as a routine. DdB
