Hacker News Reader

Comment by pedrocr | Hacker News Reader

The Toshiba and Seagate NAS lines seem well organized and better alternatives these days. And I guess this is all the more reason to get one drive per manufacturer for the typical ~4 drives in a home NAS. Between Seagate/Toshiba/WD/HGST there should be enough variation to still protect against this stuff.

View on HN

Edit: Seagate is doing the exact same thing [0], although not confirmed for NAS drives... yet. HGST is owned by WD (although the 3.5" side apparently went to Toshiba). So that leaves Toshiba.

It sucks to have to qualify drives (again?) in small businesses, thought this was a solved problem. It's more offensive they're doing it with the smaller capacities, too.

(I should note SSDs don't stand up to manufacturer's claims in error scenarios, too. But that's fairly well known I think?)

[0] https://blocksandfiles.com/2020/04/15/seagate-2-4-and-8tb-ba...

Seagate drives are notoriously poor quality, in addition to well published, large sample and long running studies by Backblaze - I ran a number of decent sized storage clusters for many years and by far found Seagate drives the most unreliable (including Seagate a rebranded as HPE etc...)

- https://www.backblaze.com/blog/hard-drive-stats-q2-2019/

- https://blocksandfiles.com/2020/02/13/seagate-12tb-drive-fai...

Agreed. I built a synology with 5x Seagate and all failed within 18 months with light workload. I will never buy another Seagate. Replaced with mix of HGST and Toshiba which are working for several years with no bad sectors. Now will buy Toshiba only until they go down the low quality road.

Buying the same brand / model drives is not a wise idea, especially if all are bought at once. Any manufacturing defects will be amplified and could cause a RAID array failure due to losing multiple drives in proximity to each other.

You are not wrong and I knew better but I was build for first time and bought all at same time. Now it's a mix of HGST, WD, Toshiba. Some are 34000h, with no issues. The WD is slower when data scrubbing but I don't trust WD for many years and that one was because other vendors were out of that size. I don't really know who to trust in that space any more so it's Toshiba for me on any new replacements.

They have a higher average failure rate but even backblaze still uses them. If you're defending against defects in product lines you want the most diverse set of drives possible to make the failures uncorrelated. Having to use drives with higher failure rates doesn't invalidate that benefit.

Anyone operating at the scale of Backblaze should also already have the right infrastructure to deal with a drive, machine, rack, PSU, etc failure without causing service downtime. (And NB that part of that will including diversifying between drive brands to guard against eg a firmware issue which affects many drives of a certain brnad in a limited time window).

What matters in that context is something similar to $/GB x (% of drives which make it to end of useful life in their application context). Cheaper drives which are less reliable might well win on that score. They might also expect to retire drives rather sooner than an average NAS user as the power cost of running older, smaller drives becomes an issue.

In an average small NAS setup the cost of a failed drive is potentially much higher. You incur at a minimum a performance penalty while the array is degraded and a significant performance penalty while it's rebuilding and have lower guarantees against data loss during that period and you can't average that out over thousands of drives.

If you've got 1,000 drives, a 99.5% or 98.5% success rate doesn't matter so much - you're going to be replacing a drives anyways. If you have 2 or 4 drives, dividing the probability that you have to replace any drives at all by a third is a big deal.

I'd say completely the opposite. If you're operating at Backblaze scale doubling your operations workload can be expensive. If you're at home doubling or tripling your chance of having to do some maintenance for an hour a year is trivial compared to the benefit of significantly reducing the chance that your whole array will fail from a common fault of all drives.

Oh my god. This isn’t true. There was one type of seagate drives that had a reliability issue and backblaze happened to have a bunch of them.

Do you realize how many different models there are, customized for different workloads? Do you realize how many different sets of firmware there are?

It’s like saying Honda has shit cars because one model year had a problem.

For a brand: P(problem | brand A) is the question. If Brand A has one product that's garbage, then selecting Brand A is now more risky.

I've been selling and doing DR on HDD for 20 years now, and frankly speaking, Seagate drives reliability isn't the best.

Numerous new technologies led to severe issues like : 5400.6/7200.4 platter dust issues killing heads the more they heat, the infamous LBA 0 fw issue on desktop drives, motor bearing seizure on 7200.10/12, lots of bad batchs on 7200.14 ST DM001/002/003, more than WD or even Hitachi. On SAS drives, the Cheetah 15K.x are not very reliable (replaced LOTS of ST3600057SS), but the Fujitsu and Hitachi are not very either.

Toshiba drives are not great either. The 2.5" MK series had an important return rate, more than the competitors. The MQ01ABD is better, but faces platters demagnetization that leads to bad sectors. The 3.5" ACA drives are a Hitachi prod line that was "given" to Toshiba when WD acquired HGST. They are quite basically Hitachi drives with a Toshiba firmware.

Edit : and now DR on Seagate drives like the Rosewood family (like the ST1000LM035) is the absolute worst nighmare for any data recovery guy. Ultra brittle head stack, easy platter damage when shocked, self encrypted firmware and SMR. If those disks makes some motor noise or head clicking, we don't even bother to open them anymore.

What would you buy for your own NAS? :-)

A mixed set of PMR drives ready for 24/24 usage. And f*cking reliable, frequent backups.

Customers here often mistake redondancy for backups. Had to save a 8 years-work worth excel file on a failed Kingston USB key today.

Okay, thanks.

Thanks for lots of valuable information, finally someone who has some data not just a story. Could you please share more? Any source on Toshiba ACA being made using HGST tech? What do you think about reliability of helium drives? On paper they should fare much better than air drives.

Of the last 4 Seagate drives I purchased[1] 2 failed in less than 1 year - one failing after a week(!) of use. This is pure anecdote, but I don't have a lot of confidence in Seagate's quality.

[1] These were bought from 3 different retailers at slightly different times, so it's not a "bad batch" either.

Thing about Seagate drives though is that they're usually less expensive than the competition. From my perspective, they do not have a high enough failure rate to justify spending extra on a brand like HGST. If you're doing things right, you have both redundancy and backups for your critical stuff, and in my experience Seagate is good about replacing failed drives within the warranty period.

Anecdotally, the only Seagate disks that have caused me significant headache are the Archive line that definitely uses SMR and is marketed appropriately because of it.

I've been using Seagate's Exos enterprise line for video processing and am very impressed with the performance. I get sustained 230MB/s write over terabytes, and while it's supposed to be slightly louder, I can't hear any difference with the case closed.

Compared to their high end consumer IronWolf Pro/NAS series they have better tech, no RAID drive limit, higher rated workload, and longer estimated MTBF. Looks even better now with the 14TB Ironwolf at $460 and the 14TB Exos at $340 on Amazon.

HGST = WD. And with seagate being kinda shitty, there seems to be only Toshiba left.

They're the same company but not (yet?) the same drives.

Do we really know? I see no reason to trust them

In the past year I've had these failures: 2x WD RED 4TB (EFRX), 2x WD RED 6TB (EFRX), 1x Seagate 4TB. I have a fairly even mix that includes HGST (and a few HGST He). No HGST failures yet. I won't be buying WD RED's for a while, if ever.

To counter that, I've been running 11x 4TB WD Red (WD40EFRX) since 2013 and zero failures, with 10 drives in RAIDZ2 and 1 warm spare.

I have them in a 4U chassis in the attic, with 5 case fans blowing through; two on the rear and three across all the bays. It's a 24 bay box, but I've not yet had pressing need to fill the rest of the bays.

Lucky. Maybe I had a bad batch :(

Disclaimer: I worked for HGST about 3 years ago (but in the Enterprise storage (flash)/Object store fields, not consumer/prosumer HDDs).

At the time, WD was assimilating HGST's drive tech where there were cost/performance/reliability benefits to the point where quite a few 'WD' products were just relabeled HGST ones, possibly with minor firmware/logic board tweaks. I seem to recall that at the time some WD Black, and maybe Gold, models at least were rebadged HGST units. The aim at the time was to fade out the HGST brand so that everything was WD and I pretty much think that has since been done.

I have a few Seagate Ironwolf 8TBs and I think (hope?) they are not SMR.

Reply on news.ycombinator.com