ZFS NFS server with ZFS and backup strategy

S

spectrum48

Hello,

I'm going to replace my main workstation with:
  • a headless system exporting disks with NFS;
  • two less powerful workstations, in two different rooms, mounting my home directory from the NFS server.

I'd like to:
  • use ZFS on the NFS server, with one SSD for the OS (FreeBSD 11.0) and two 6TB disks in mirror mode;
  • backup my data with zfs send/receive on two different external USB3 HDDs, both using ZFS, alternating them day by day.

My doubts are:
  1. Is it possible to scrub a 6TB disk connected with USB3 or will it take ages? If it's not practical, do you think a single ZFS disk without any scrubbing is better than using rsync to copy my ZFS system on a standard UFS backup disk?
  2. I've read that using a single disk with ZFS is worse than using standard filesystems because without redundancy errors can be detected but not fixed and corrupted metadata lead to losing the whole ZFS pool. (Source: one of the moderators on https://forums.freenas.org/index.php?threads/single-drive-zfs.35515/). Other sources say that a single ZFS disk is always better than other filesystems since loosing the whole pool for metadata corruption is very unlikely. However, if a single USB3 ZFS disk has small advantages over an UFS one (due, for example, to the impossibility of scrubbing) I'd prefer not taking this risk. Did some of you weighted pros and cons of single ZFS disk external backups?
  3. I read that applications like NFS requires synchronous writes and can benefit from a SLOG (Source: https://www.ixsystems.com/blog/o-slog-not-slog-best-configure-zfs-intent-log/). Do you think this is true even when there is a single NFS client connected or would it be overkill?
Of course, if someone thinks my proposed setup is fundamentally wrong critics and suggestions are welcomed.

Thank you to everybody who will help to clear one or more of these doubts!
 
SirDice said:
You can have redundancy with a single drive if you set copies to 2 or higher.
Click to expand...

Yes, at the beginning I was thinking about 2 x 4TB mirrored disks on the server and an 8 TB external disk with copies=2 for backup. But I would really benefit from 2 X 6TB disks on the server which makes me think redundancy on the backup disk is impossible.

SirDice said:
https://wiki.freebsd.org/ZFSTuningGuide#NFS_tuning
Click to expand...

Thank you, I didn't found this guide before. But I don't understand this passage:

The best solution is to put the ZIL on a fast SSD (or a pair of SSDs in a mirror, for added redundancy). You can now enable/disable ZIL on a per-dataset basis.
Click to expand...

What's the point of putting the ZIL on a fast SSD if I'm going to disable it? Wouldn't it be simpler to simply disable ZIL? Or am I missing something obvious?

The only application requiring data consistency is a database running on the server but it will use the SSD with the operating system, not the exported ZFS disks. So I'm only concerned about avoiding system corruption.
 
You must log in or register to reply here.
Back
Top