Network Storage Considerations (2023) - All-in with Synology

Background

The 6-disk TrueNas has been causing me a plethora of issues. I’ve gone through 5+ HDD RMAs through Western Digital (all without incident) just this year alone.

• The HDD models varied from WD Red Pro, to Gold, to Ultrastar models. No obvious pattern across the disks, themselves.
• I’ve also replaced the motherboard on this server earlier this year, because the previous one had a one bad SATA port (of the 6 total on the board).
• I’ve re-cabled and played musical chairs with the disks themselves, within the server. In an effort to isolate a bad cable/SATA port/anything.

No obvious patterns arose, with the above.

Looking through the FreeNAS archival forum, it looked like those using Node 304 cases were no longer using them (based on rudimentary investigation via signatures). Reading into that finding more deeply than I should, I decided to call it quits on this storage server. There simply wasn’t enough time to continue dealing with the debugging of this mystery.

Enter Synology.

I already transitioned my larger (power-hungry) dual-Xeon TrueNAS server to the ds1821+ a few months ago. Given the zero issues that arose from that, I decided to go in on another. Completing this second migration would bring me to a post-TrueNAS world. And while my 6-disk TrueNAS didn’t need it, I went with another ds1821+ and popped in a spare Intel X540-T2 for the 10Gbps link with the M1 Air.

Desired State

We did some formal renaming along the way (aka finally given hostnames), given how Synology limits what can be exposed via SMB shares.

• 6-disk is now tank
• 8-disk is now dozer

The connection map is found below:

While this is technically wired up, the restores are slowly running–limited by the 1G link on dozer.

Minor Details

I ran into system crashes when trying to do restores on the M1. I didn’t dig too deeply into it and just trusted the SMB stack on Windows and was reminded that Arq restoration has a license-free Restore-only mode that I could use. So I did the majority of my restorations via my Windows 10 desktop. I still did a second pass in macOS as a precaution, before calling it “Ready to Use”.

Also, there was one particular backup plan that I didn’t run very often (3 records this year). Looks like all backup records had at least one corrupt file in them, so there were ~10 missing files total. Not good. Not terrible, given that this was a backup plan that wasn’t critical. And it’s good to know that Arq can recognize basic issues like this.

What’s Next

If I find myself running more restores, I’d be very tempted to get another 10 Gbps NIC for dozer so that future dozer<>tank operations would run faster. At that point, I may also pair this with an upgraded switch with more 10G ports.

Otherwise, I would expect that storage considerations are done for the foreseeable future. Until an all-flash (all-NVMe, ideally) 12-bay chassis comes along to run TrueNAS again.