[lfs-support] LFS-7.0 with LVM

Qrux qrux.qed at gmail.com
Sun Jan 29 18:30:09 PST 2012


On Jan 29, 2012, at 5:41 PM, Baho Utot wrote:

> On Sunday 29 January 2012 08:08:58 pm Bruce Dubbs wrote:
>> Baho Utot wrote:
>>> For me it is ever try to manage 16 regular partitions?
>> 
>> How about two regular partitions: / and /boot, and lvm for everything else.
>> 
>> And yes, I do manage 16 regular partitions:
>> ...
>> So I can experiment with standard jfs, xfs, and reiser filesystems and
>> boot from them without an initramfs.
>> 
>>   -- Bruce
> 
> What I do is to create a new lvm partition for the system under test. . .
> ...
> LVM also allows snapshots of your running system.

If you're not testing kernel operations that handle bare-metal (e.g., physical hardware drivers), you can do FS-related testing (or anything higher-level than the FS) even more easily (from a disk-space perspective) with VMs.  If you use Xen with disk images, for example, after you're done with a VM, you clean up like this:

	rm -f vm1.img

Clone an OS like this:

	cp vm1.img vm2.img

Snapshot like this:

	cp vm1.img vm1-now.img

No resizing partitions or filesystems.  Even easier than LVM.  ;)  The only case I see for needing to boot into LVM is for managing partitions over 2TB, though it seems like parted could handle that pretty easily (I can't verify; I'm a 32-bit fdisk fan).  I'm just a user, but I tend to agree that for LFS, purity (and simplicity) is good.  Not even sure that md would be a good inclusion for LFS.

Maybe consider a "Booting- and Hardware-Related Options" section for BLFS?  It could include md, lvm, interactions with hardware RAID setups, (e.g., filesystem options like stride), other hardware options (like SSD trim and 'discard' in ext4), multiboot kernels (like the Xen hypervisor), needs of other virtualization technologies, etc.  Or, maybe LVM could fit into the "Virtualization" section we talked about for BLFS...

	Q




More information about the lfs-support mailing list