Attaching an external disk shelf

[Flupp]

I'm using a TrueNAS Core system with the following specs as our main production NAS:

- EPYC 7713P 64c/128t
- 256GB ECC RAM
- 2x25Gbe in LAGG
- 24 2.5" SAS SSDs + 2 NVMEs for OS

Having set this up in 4 vdevs of raidz2, we are slowly running out of space and one of our issues is how long its currently taking us to move data off production for archival. To solve this, I want to get a new TrueNAS system to be used as kind of an on-site archive (which will also archive to S3 Deep Glacier).

Instead of getting a separate server, I was thinking of getting an external disk shelf and attaching that directly to this server using an HBA with external connections. Probably something with 24 3.5" bays and setting it up in 2 vdevs of RAIDZ2. I could then SSH into the server and rsync the data from the production pool to the archival pool. Basically trying to avoid saturating the servers NIC as I don't want it to effect production while we are archiving.

So I have 2 questions in my head about this:
- If I were to do this, would those external disks be visible to the TrueNAS as individual disks? I assume so as it's effectively a JBOD situation?
- Is there anything that screams out BAD about my plan?

 

[Ericloewe]

- If I were to do this, would those external disks be visible to the TrueNAS as individual disks? I assume so as it's effectively a JBOD situation?

Yes, of course, unless you do something silly like install a RAID controller and attach the disk shelf to it.

- Is there anything that screams out BAD about my plan?

No, but why rsync and burn CPU time? Just replicate it all.

 

[Flupp]

Oh yeah, in my head ZFS replication was for server to server. But can you replicate a specific folder or does it have to be the entire dataset?

Let's say we got a dataset Main and it has a bunch of folders at it's root, one for each project. I want to be able to archive specific projects, not the entire set. In the GUI i can see that I can do the set but can't seem to be able to drill down from that.

 

[Ericloewe]

But can you replicate a specific folder or does it have to be the entire dataset?

No, it's per dataset, that's one of the reasons "more datasets" is a god rule of thumb. For instance, I organize stuff at work first by category, then by year - that way I can backup smaller chunks at a time, if the need arises.

 

[Flupp]

Rsync it is then I suppose, or maybe Rclone?