I started using LVM to take advantage of snapshots and want to make sure I have my setup proper and what can I improve? It's working so far. I use *.raw format since easier to backup, restore and move around. Looked online and tons of good help but no straight tutorial on using *.raw files and LVM with snapshots so hoping this should work well.

KVM has worked very well for me with raw files on just ext4 partitions without LVM for few years. Want some live backups going forward on the weekdays and offline still on the weekends.


1) 3 servers (*.raw file) running from 3 nvme drives (1 drive per raw for I/O performance).
2) Created 1 GPT partition per nvme drive per Redhat recommendations and made sure aligned.
3) Ran pvcreate /dev/sdX 3 times on each drive.
4) Created 1 VG and LV per nvme drive using 75% of the drive partition (rest for snapshot)
4a) I made the LV 10GB larger than the *.raw file. (this my main question if okay or not? Don't know if there's some kind of overhead.)
5) Formatted each LV ext4 with defaults
6) Created 3 directories in /mnt for each raw file an modified fstab with UUID of each LV.
7) Copied working server1.raw file to each drive (1 per drive).
8) Modified (swappiness = 10) since have 64GB ram (that was fun to figure out).
9) I wrote a small script to take a snapshot every night with 20G snapshot (in case Windows Updates, already screwed me during copy using 1G).
10) I remove the snapshot after the *.raw file is copied to external drive since I know there is a performance hit per snapshot.

Does this look okay to where I shouldn't have any weird issues in the future?

Other Questions:

A) Is there a certain sector/cluster size I should use since only 1 raw file per LV for performance?
B) Does running pvcreate 3 times keep them separate or adds them to a pool if I did pvcreate /dev/sdX /dev/sdY /dev/sdZ? So I know in the future.
C) Even though I set swappiness to 10 in sysctl and rebooted it still uses swap file even though I have 16GB of ram free with all VMs booted and linux system. I'm assuming this is normal, but used to not seeing swap file touched. Verified it's still at 10.

Thanks!

A day and a half, a couple of hundred views - may everyone else had the same reaction as me; hmmmm ...

It looks like, from that description, you are mounting all the guests in the host and doing nightly LVM snaps of the guests. From the host.
Is that correct ?.

Yes this is correct. Want to make sure this is a good process or is there something I should improve? Also the questions about cluster sizes in the LV since only storing a single RAW file. I understand basics of clusters, but never messed with them before so don't know what I can set to try to increase performance. Or does that bring in it's own level of complexities?

I cannot imagine any situation where this will provide valid (current) backups.
ext4 is not cluster aware, nor is LVM without the CLVM extension.

Without having tested this, I would expect the cache for the lv to be instantly out of date on the host, and there is no way for it to know of updates to the filesystem. Even if you remounted the filesystem on the host immediately before snap it would always be missing the data in the guest cache and (ext4) journal.
I repeat, I can never see this as being a valid backup procedure. Do your snaps and subsequent backup from the guest(s).

Note the use of "verified"in my sigline.