I have a VM host with image files around 100Gb. It's on `RAID5`, so it has some protection from a hardware failure, but not from accidental `rm -rf /`.
I want to setup periodic snapshot creation on a remote machine, but I think that `rsnapshot` won't do, because even if a small portion of a VM image changes, the backup server would have to store the whole file for each snapshot.
If I use `rsnapshot` with conjunction with `LVM` snapshots, it will solve the above problem, but `rsync` will still have to scan the whole image file for checksum comparison, which is bad for server's performance.
I want to use something similar to
Write-intent_bitmap to accumulate dirty blocks and sync them when I feel like creating a snapshot. The solution may still involve LVM thin snaphots on a remote disk.
Is it possible with mdadm or do I need some other software?
I've been using `LVM` thin volumes on some servers for about a year. They're very handy, but they get fragmented and unless you have SSDs, they're a performance killer. This is why I want them on a dedicated backup server.