It has been a while since I posted an update about my Proxmox homelab.
Memory
I upgraded the “Logos” from 64 GB to 96 GB after finding a reasonably priced 2×32 GB memory kit. After installation, I ran extensive MemTest86 checks; everything passed without issues.
Unfortunately, memory prices are going through the roof at the moment. Otherwise, I would have pushed this system to 128 GB without hesitation.
Storage
Over the past couple of weeks, I picked up six Samsung PM863(a) 1.92 TB enterprise SSDs; well, five Samsung drives and one HPE-branded variant.
While test-fitting the drives in the NOX 010 case, I ran into an unexpected issue. The drives themselves fit just fine, but the SATA cables I had did not. There is very little vertical clearance between the stacked drive positions, and straight SATA connectors collide with the drive mounted above them. That, in turn, leaves no room because of the standard drive bays above.
Great..
The solution turned out to be simple: 90-degree angled SATA cables. After test-fitting those, everything snapped into place perfectly.
Drives
I thoroughly tested all six drives. Five out of six report a drive health of at least 96%, which is excellent for enterprise SSDs that have already seen a few years of use.
One of those five, however, reports two relocated sectors.
Two relocated sectors do not necessarily mean that a drive is failing. It can simply indicate that the controller decided a particular sector was taking longer than expected to write to and proactively remapped it.
Unfortunately, there is no reliable way to determine when those sectors were relocated. Did it happen early in the drive’s life, or near the end, just before it was taken out of service? In the former case, this is perfectly acceptable. In the latter, it could indicate that the drive is starting its journey to the recycle bin.
HPE
And then there is the PM863a drive rebranded as an HPE drive.
This one reports almost nothing via SMART. Thanks, HPE, for the wonderfully crippled SMART readout.
With a power-on count of just 33 hours, this drive is either practically unused, or its history has been scrubbed and the counters have been reset.
With no drive health indicator and no reliable total-bytes-written metric, how am I even supposed to assess this drive’s condition? Not just now, but also over time.
This drive exposes so little information that I have nothing meaningful to base health monitoring on, nor any way to detect when it might start its journey toward failure.
So I ordered two additional PM863a drives from eBay. Hopefully, they’ll arrive soon.
See you in another update.
Brain out!