Hacker News Reader

Comment by numpad0 | Hacker News Reader

(with a tinfoil hat on) I'm convinced that Backblaze is intentionally withholding and ambiguating data to prevent producing too-easily understood visualization that Seagate is consistently the worst of the last 3 remaining drive manufacturers.

Their online notoriety only started after a flooding in Thailand that contaminated all manufacturing clean room for spindle motors in existence, causing bunch of post-flood ST3000DM001 to fail quickly, which probably incentivized enough people for the Backblaze stat tracking to gain recognition and to continue to this date.

But even if one puts aside such models affected by the same problem, Seagate drives always exhibited shorter real world MTBF. Since it's not in interest of Backblaze or anyone to smear their brand, they must be tweaking data processing to leave out some of those obvious figures.

View on HN

I don't think so, their posts still have all the details and the Seagates stick out like a very sore thumb in their tables:

https://backblazeprod.wpenginepowered.com/wp-content/uploads...

and graphs:

https://backblazeprod.wpenginepowered.com/wp-content/uploads...

Since it's not in interest of Backblaze or anyone to smear their brand

It is if they want to negotiate pricing; and even in the past, Seagates were usually priced lower than HGST or WD drives. To me, it looks like they just aren't as consistent, as they have some very low failure rate models but also some very high ones; and naturally everyone will be concerned about the latter.

To Seagate's credit though, their warranty service is excellent. I've had the occasional exos drive die (in very large zfs raids) and they do just ship you one overnight if you email an unhappy smart over. Also their nerd tooling, seachest, is freely downloadable and mostly open source. That's worth quite a lot to me...

(And if anyone is curious about their tools – https://github.com/Seagate/openSeaChest is the link. Lots of low level interesting toys!)

OTOH, Seagate never sold customers SMR drives mislabeled for NAS use.

Ugh, source on that? In the market for a new NAS/Homeserver soonish (realized my drives are almost at 10 years of power on time) and would like to have spinning rust behind ssd for larger storage.

It was a whole thing a while back. This was maybe the original article, but once it landed this was the headline of all tech news for a couple of days. https://blocksandfiles.com/2020/04/14/wd-red-nas-drives-shin...

SMR drives aren’t inherently bad, but you must not use them in a NAS. The may work well, up until they don’t, and then they really don’t. WD snuck these into their Red line, the one marketed at NAS users. The end result after a huge reputational hit was to promise to keep the Red Pro line on HMR, but the plain Red line is still a coin flip, AFAIK.

I will not use WD drives in a NAS. It’s all about trust, and they violated it to as astonishing degree.

Ugh, that's sad. The 10 year old drives for me are WD:RED's so can't complain about the old ones.

Looking at the Pro/Plus PDF's it seems that they do specify CMR officially so sneaking in SMR disks sounds like lawsuit material.

I'll probably go for them again, just be vary with research!

Not 100% sure about SMR situation, but granted, Seagate was never not technological front runner nor untrustworthy nor unfaithful company; their 5k4 drives were always more cost effective than anybody and they're the first to ship HAMR drives right now as well. It's __JUST__ that the MTBF was always statistically shorter.

Which is a significant “just”, to be sure! But in my experience, if an Iron Wolf survives a RAID rebuild, it’s probably going to work for many more years. I’ve had 3 WD Reds claim to keep working, and still pass their SMART short and long tests, but tank in performance. I’d see the RAID usage graphs and all drives would be at like 5% IO utilization while the Red was pegged at 100% 24/7. The whole volume would be slow as it waited for the dying-but-lying Red to commit its writes.

In each case, I yanked the Red and saw volume wait times drop back down to the baseline, then swapped in an Iron Wolf. Fool me thrice, shame on all of us. I won’t be fooled a 4th time.

I’m not a Seagate fanboy. There’s an HGST drive in my home NAS that’s been rocking along for several years. There are a number of brands I’d use before settling for WD again. However, I’d make sure WD hadn’t bought them out first.

Didn't the ST3000DM001 fail because of a design flaw on the read head landing ramp?

According to Wikipedia: https://en.wikipedia.org/wiki/ST3000DM001

Somewhat tangent: Imagine my dismay after googling why two of my drives in my NAS failed within a couple of days of one another, and I came across a Wikipedia page dedicated to the drive's notoriety. I think this is one of the few drives that were so bad that it had it's own dedicated Wikipedia page.

Same thing happened to me with this drive: Lost the data on a RAID because two drives failed at the same time.

I did not even have a raid, I was too cheap to allocate drives to parity. So I had two of these and a couple of other drives in a JBOD array fused together into a logical filesystem using something I can't remember now.

Maybe my approach was better in hindsight: I only lost the data on those two drives and everything else was left untouched. I guess if you know the drives are dodgy, put them in a JBOD array as it will limit the blast radius in case of failure.

Reply on news.ycombinator.com