Linux – Redundancy: LVM+mdadm or LVM internal RAID

linuxlvmmdadmsoftware-raid

The task is to create LVM storage for virtual machine images. So LVM will be used in any case.

There are two ways LVM providing redundancy:

block mirroring

lvcreate -m 1 –mirrorlog mirrored -n logical_volume -L 50G vg0
internal RAID mechanism

lvcreate –type=raid1 -n logical_volume -L 50G vg0

Second way should provide reading performance boost.

As alternative we can create raid1 with mdadm and use Linux Soft Raid device as physical volume for LVM.

Which way is better and why?

Best Answer

mdadm raid is better for raid1 and on top of that you can put lvm for flexibility. Lvm has problems when a disk goes bad, mdadm is the tried and proved solution for raid1.

Related Solutions

Lvm – Optimizing ext2 filesystem for use on LVM + RAID device? Stride, stripe-width, LVM IO size considerations

I could suggest checking this answer for ideas. I provided some feedback there regarding aligning xfs and someone commented about ext3. It may give you some clues.

Also, before aligning file system make sure your volume is aligned to stripe size as well. It can be done by properly sizing metadata and optimizing an extent size. Here is an example:

pvcreate -M2 --metadatasize 2048K --metadatacopies 2 <raw_device>
vgcreate --physicalextentsize 256M --autobackup y vg10 <pv_device>

Lvm – XFS on LVM or directly on MD array

I think the performance impact is really minimal for LVM, except if you use snapshots which can have a big impact - best if snapshots are used during backups only. This answer agrees.

For a detailed discussion of issues with LVM, with some mention of XFS, Advanced Format, RAID, etc, see LVM dangers and caveats

Best Answer

Related Solutions

Lvm – Optimizing ext2 filesystem for use on LVM + RAID device? Stride, stripe-width, LVM IO size considerations

Lvm – XFS on LVM or directly on MD array

Related Topic