[alias_neo]

> Wasn't Western Digital caught selling NAS drives with some shittier technology than they were advertising

WD switched their (larger; over 4TB iirc) WD Red drives from CMR to SMR without changing the model number at all, this is why I switched to buying Seagate.

The SMR problem became particularly apparent when NAS users (like myself) switched out a failed drive with one of the same model in their ZFS pools, and resilvering would fail.

Interestingly, the Seagate drives I switched them with just a couple of years ago have now started failing (one of them at least) so that didn't work out.

Does anyone know if it's possible to check this runtime data on the drives? According to the article it's not in the SMART data which has been reset in the case of the drives they're talking about.

I'm thinking of switching my NAS to solid-state as I've never had an SSD fail yet I'm replacing disks in my RAID1 ever couple of years on a home NAS that sees fairly light load other than some VMs and Kubernetes clusters writing logs etc since I'm not actively using much of it for 90% of the time it's on.

[marxisttemp]

> I'm thinking of switching my NAS to solid-state

My NAS is mostly read-only and I’m very keen to do this. My understanding is that since SSDs are still readable when they fail, you don’t need the same degree of RAID parity to avoid data loss.

[alias_neo]

My NAS isn't really a NAS, it's a fairly hefty server with a Ryzen 9 on an ASRock Rack board in a CS381 case with 8 hot-swap bays occupied with SAS and SATA HDDs as well as 2 nvme drives in the board itself. It's a server doing server things and one of those things happens to be roleplaying as a NAS.

It runs a bunch of VMs with their primary drives on the HDDs but most of the VMs are idle most of the time unless I'm working on something, like deploying to the Kubernetes clusters so it's just Ubuntu, Debian, whatever else I'm running doing it's thing with logs etc.

Some of the drives are used for storage in the traditional NAS sense so they get hit for backups nightly.

The HDDs keep failing on a pretty regular cycle every few years so switching them to SATA SSDs might be an idea. I only used the "server" part of it once a month or so when I want to test some deployment etc, the only reason it's on 24/7 is because of its NAS usage.

I use mostly Z1 mirrors at the moment, 4 pairs of drives, as for SSDs, I'm yet to have a single one fail in any machine I own or have owned since SSDs were first a think. I tend to stick to good-brand drives such as Samsung EVO Pros or if I'm trying to save a bunch of cash and the data isn't I buy Crucial.

[ciupicri]

How good is the CS381 case? Do you have any problems cooling the hard disks?

[alias_neo]

I'm very happy with it overall; the only potential issue I have with the case is the short clearance between the CPU socket and the HDD cage above, means you need a very low profile CPU cooler; I use a Noctua cooler, don't recall the name, but it's a very flat one, with a 10mm depth fan on top, all in the cooler is no more than about 40mm deep.

I've never had an issue cooling the HDDs, I swapped out the fans it came with for the back of the case for some Noctua Redux PWM fans (I think those rear ones are 120mm) and I let the motherboard control them, the HDD backplanes also have a connector for a fan for each side of 4 bays. Not sure if they're always on, never used them, but looking at my drive temps, the HGST 10kRPM helium drives average 41C, and Seagate IronWolfs averaging 34C.

I also have an HBA, Quad-Intel GbE NIC and a Tesla P4 in there all putting out a tonne of heat. I've down-specced the CPU to 64W for longevity and power savings, but I've not seen any temperatures I'd be concerned about yet, other than the passively cooled P4 which I printed a bracket for and strapped a high-static-pressure fan to push air through it and out the back, but that's GPUs for you.

EDIT: For the record, I bought it in 2020, and it has been running 24/7 since, with regular HDD swaps, and the Tesla was added just recently.

[dml2135]

I think its actually 8TB and under that WD switched to SMR. All their drives over 8TB are CMR, last I heard.

[alias_neo]

You're probably right, I forget which way around it was, but if I was to take an educated guess, I'd say that SMR on the lower capacity models lets them cut costs because the added density of SMR lets them use fewer platters?

[dml2135]

I'm not sure why SMR would save costs on lower capacity drives but not higher capacity drives. It may be a technical reason, but I'd also find a corporate-strategic reason to be plausible -- i.e., they were like, any consumers that give a shit about something as technical as SMR vs CMR are probably the ones buying our high-capacity drives.