Michael Rohleder <mike@rohleder.de> writes:

> Hey Simon!
>
> Simon Josefsson <simon@josefsson.org> writes:
>> I'm thinking that having the root file system on LVM may be unsupported,
>> so that manual could say that, but I'm not even sure that is true.  I
>> haven't tried it though.  Perhaps 'lvm2' should be pre-installed on the
>> default installation image to help move LVM support along.
>>
>> Another thought may be that constructing LVM partitions through the
>> mapped-device mechanism is not supported:
>>
>> http://guix.gnu.org/manual/en/html_node/Mapped-Devices.html
>>
>> however as far as I can tell, LVM is working just as well as LUKS/RAID
>> here: LVM just doesn't need any command to start the devices.  The
>> mapped-device approach appear to require out-of-band device creation for
>> LUKS and RAID, just like is required for LVM.  A no-op
>> `lvm-device-mapping' could be added for completeness, and it might run
>> 'pvscan --active ay' or something like that if we really wanted to but
>> I'm not sure when that would be useful.
>
> I use a root fs on lvm.
> What helped me a lot was this posting, where I stole most things:
>
> https://lists.gnu.org/archive/html/bug-guix/2020-02/msg00285.html

Thanks for pointer!  Maybe some examples like this could be added to the
manual.  I'll experiment a bit too and see if I can propose something...

> What (currrently) isnt working is the lvm device-mapper and (for me more
> important) lvmcache...

What do you expect from a lvm device-mapper?  When I thought about it,
it is mostly a no-op since the kernel handles this automatically, or?

Working lvmcache does not seem required to claim support for LVM to me.
Why doesn't lvmcache work?  Also sometimes the performance improvements
doesn't seem that significant:
http://strugglers.net/~andy/blog/2017/07/19/bcache-and-lvmcache/

/Simon