Zpool on sparse files


New Member
Nov 2, 2020
Hi everyone!

I have an NFS share, where I'm planning to store my backups. I created a raidz2 zpool on 7 sparse files on that NFS share and transferred snapshots using syncoid to that zpool.
It seems to me that it's a working solution, but I haven't done that before and wanted to hear your opinion on this topic.
I'm now aware that failing NFS server can make the server I connected it to unresponsive, so, to minimize the chance of data corruption in that zpool, I'm planning to mount and import zpool before the backup process started and export+umount when it's finished.
Other than that, are there any disadvantages or pitfalls in having a zpool on sparse files?

I'm using ZoL 0.7.5 on ubuntu 18.04

Thank you for replies in advance :)


Active Member
Sep 13, 2019
For data you care about, don't use *files* as zraid members.

I have used files as zraid members as part of a migration plan at home. (while still retaining enough "real" drives in each vdev to maintain file integrity).

This is a "running with scissors" thing. Unless you really REALLY know ZFS (and the given implementation) AND the host file system where your "drives as files" will live AND you can absolutely positively guarantee that any two files that act as drives for the same pool won't themselves end up on the same physical disk, you really, really, really shouldn't do this.

This is 100% riskier than running ZFS without ECC, and will almost certainly lead to weird/annoying performance problems.

I also suspect you have a "hammer to get ice off of a windshield" situation here ;) (As in, you may need to backtrack some steps and re-examine your problem and how to solve it.)