The SMR Red's need to be rebadged as WD Browns ;)

Until WD removes all SMR drives from their "RED NAS DRIVE" line, I will not take them seriously, nor purchase WD any longer.

Shit man, 15 times longer for a resilver is wildly, wildly problematic. Not acceptable. Excellent work STH

Ever since in 2015 I bought my first SMR Ext. drive from Seagate, I was aware of the hanging/jerking behavior. So I totally shocked by when WD had made their RED drives to be SMR. They should discontinue the lower-end model of RED drives instead of using SMR to make profits or just label them as BLUE drive whatsoever. Labeling a SMR drive as NAS drive is a criminal.

Only 5 minutes in and this sounds damning. WD can ignore random people on the Internet but not STH.

I consider the bait and switch a total scam. What shady tactics and full of corporate greed.

This really feels like an example of false advertising on WD's part. If you're going to segment your product stack and advertise particular drives for specific use cases then you should ensure the drives are suitable for them. I feel like WD should offer a refund or exchange program for anyone who unknowingly purchased an SMR drive that was advertised for use in a NAS or RAID setup.

Smells like class action in the making

A friend of mine just almost bought one of those 4TB SMR drives for his Synology 5 bay NAS. I have stopped him in time and made him order an IronWolf instead. So glad I did that. The rebuild of Synology's 16TB formatted mdraid was finished after about 12 hours (a night basically).

WD also has a 2.5" 2TB "Black" drive that uses SMR. Black is supposed to be the performance brand.

9 days to resilver! And that's ZFS, which only rebuilds the 60% of filled data! @@

WD are not idiots regarding HDD technology. They had to understand the issues customers would encounter with these drives and that the use of SMR wouldn't stay secret long. This along with their customer unfriendly blog statement which attempts to gaslight customers and shift the blame to them makes me believe that WD made the decision willing to accept whatever blow back occurred. They must really be seeing some manufacturing cost savings on these things. What I don't know if they considered is that the small NAS segment includes many who run a small NAS at home and enterprise NASs at work. And that those who do are also the ones most likely to cheap abreast of news such as this. This will negatively impact their enterprise business. They are already on my do not buy list.

I specifically ordered some EFRX and received EFAX anyway. Returned them and went with a different brand.

WD needs to recall the SMR red drives, or justifiably we need to advise as many customers and friends as possible to avoid their products.

Good to know. But the video could've been 8 minutes without all those repititions again and again and again... 😉

Really appreciate your data and coverage of this. I was recently buying some cold spare WD Red 3TB models right before a lot of videos about the issue popped up. I was completely unaware of the problem. Just by chance the 2 used drives and new drive I got off eBay happen to be the older CMR models matching the 9 I have in production. Being "surprised" by a multi-day rebuild when your array is already degraded doesn't sound fun. Eventually I'll phase these out, but I'll be looking at IronWolf if I'm not getting the Pro models.

Great video, thank you all for all the hard work.

holy crap, these drives would even be shitty for consumers. SMR is the QLC of hard drives

No more WD red. Go Ironwolf for now.

Even if you pay attention and buy the CMR version of a WD Red 2-6TB drive, in case of a warranty replacement you'll very likely get an SMR drive. So perhaps best to avoid the WD Red line altogether? Not too many alternatives though...

I only learned of the drawbacks of SMR drives a couple of months ago. I've been extremely lucky in the fact that every drive I've purchased over the years have been CMR drives - including a pair of 12TB WD120EMFZ drives I recently purchased. I run openmediavault for my NAS/file storage needs, I'm also old school in the sense that I only run RAID1 arrays on a P410 SAS controller but its never let me down thus far so I have no need to change.

Thanks for all the hard work. I actually have a few 4TB WD Reds on my desk right now. Thankfully they seem to be of the CMR variety.

27mins of great information, thanks STH

Thank you for this. Awesome content as always.

Excellent test and video, Thanks STH

Outstanding information in this video. Thank you very much for posting this.

Red drives screwed me and WD support could have cared less. Cost me a couple grand to see if Data could have beee restored at a Tier 4 facility. They couldn't. WD lost a long time customer who has spent $10's of Thousands of $ with them.

This is good info Pat. I’m approaching the point where I’ll need to update my ZFS storage from 2 to 4 TB drives and was in the dark about the affects of mixing Red models. Considering stores aren’t open to verify the model being purchased, I’m considering jumping straight to 8 TB models.

I don't see a use case for SMR outside of the enterprise market (where it can fill a similar roll as tape). It should most definitely not be marketed towards consumers but WD even have a model of 2.5" Black series using SMR. The whole hard drive industry should be ashamed for their false marketing and they should make introduce new product lines for SMR where it is obvious for any buyer that these drives have problems with modifying data. Consumers pay with performance instead of with cash when they buy SMR drives.

I have been so utterly underwhelmed by WD's decisions for the last dozen or so years. I have no trust in their products at all any more. They used to be the gold standard. What a shame.

That sort of resilver time alone makes it not a NAS drive 😬

Man, a few days ago I had to buy 3x 6 Tb Ironwolfs and I was seeing a lot of chatter about SMR. I didn't know much about it other than it was bad, so I just quickly verified the Seagate line didn't use that tech and made my purchase. I thought it was a minor thing but it looks like I dodged the bullet here. I've always avoided WD anyway but damn, I'm glad this information is getting out there.

Finally good test, with conditions that mimic real usage.

Thank you. And thanks to the Hard Drive Manufacturers that now clearly label their drives as SMR or CMR. Im fairly new to private Disk Arrays, building my own soon, and such reports help learn the important facts regarding this matter.

Crap, I so just did this. I had a WD Red NAS array and replaced one drive without realizing they were SMR vice CMR. I found this video a few days after purchase and installing. I am going to go and drop another 105~ dollars and buy a CMR drive. Great video!, sometimes you learn the hard way.

good video! Also, it might be a good idea for ZFS or such file systems to CHECK for SMR drives before RAIDing them and then warn the user. eg It can have a list of the models etc. The NAS software could also warn you based on the list. I also did not realize this until after I purchased some RED drives...

This is some great tech journalism kudos to Will and STH. I was looking at two 6tb red drives on Amazon and was wondering why they had a $40 difference. It was CMR. Red drives are very quiet I already own 2 3tb REDs CMR and they've been fine for years (AFTER rmaing a few in a short timespan in the beginning lol) but I do know they're quiet which I took for granted so I'm buying a few more now.

The worst part of this is that only the smaller sizes use SMR, you know, the types of drives that home users with small arrays would use (i.e. arrays with the lowest redundancy).

Sad to hear, especialy WD destroyed such a good branding with this shit. I myself used WD - Red and WD Blues but in my next NAS, there wont be a WD in it, out of principe

Thanks for doing this testing. So happy to have you guys as a resource as I navigate my first home server build. Thankfully I would have avoided this as I am planning to use 10tb Reds but now I am wary of WD and my be shelling out big money for some HGST drives (even though they are WD) Would you guys still recommend shucking some WD externals (as long as they are larger) or should I start looking for segate to go on sale?

Thank you for this video, as i read this from your website also, and this cleared a bit, heh.. and the thing with 2 SMR 'nas, drives inserted to raid.. i wonder if double the 9days o.O

This is pretty damning.

So unfortunately I do have SMR drives in my system already. I have been no longer buying these drives for the last several expansions as I upped my capacity as the price points changed. Say I have a SMR RAIDz2 vdev and have a drive die. If I replace the dead drive and the new drive is CMR drive will I see these long rebuild times? Basically slowly moving my current SMR drives to CMR as they die. I think this would be the case as SMR should be fine for read performance but I'd love if STH could test this for myself and others who are in this situation.

Thanks for shining more light at that issue. It should be spread as a educational piece for all consumers and why you shouldn't be one. Was adding one 4tb drive into my home nas few months back. It was already populated with Reds so another Red was a no-brainer. I've been involved with IT ever since I was 8. Microsoft, Nvidia, AMD, Verbatim, Sony and many others that I don't remember from the top of my head pulled that sheet and got caught red handed. I have enough experience within IT and freaking shopping in general, that change of one letter in the product designetion lit up red flags everywhere. WD marketing boss must still be feeling the 80s with that kind of bravado.

0 dislikes a day after upload! WD representative hasn't watch this video yet good work! spread the word!

FYI: Recently bought a WD MyCloud EX2 Ultra and it shipped with two WD Red SMR drives, made mid 2022... so obviously they still think SMR drives are just fine for NAS usage, and/or they're attempting to get rid of their stock of SMR Red drives by putting them into their own NAS models and hoping that customers won't notice or care.

I was telling my friends this as soon as it came out. We all run FreeNAS. I needed to replace a drive and if the news didn't break when it did, I would have had one of those SMR WD Reds. If I experienced such long rebuild times I would have ended up RMAing it with cross shipping, and there would be a lot of issues. In the end, I ended up buying a WD Red 8TB, since eventually, I want a 6x6TB RAIDZ2 vdev in my pool, but seeing at how 6TB is SMR now, 6x8TB is my only option. I didn't want to spend so much money on these drives, I can't afford it, so it means it's going to take longer for me to finish this upgrade, but unfortunately for me everything I have is WD, and I don't really trust Seagate (the numbers, in my opinion, haven't added up). I'm glad the tech community raised the alarm as soon as it happened to I knew to look into this.

At least Newegg is now specifically stating which models of the WD Red series are SMR and which are CMR. Props to them

Interesting that these results differ so much from the ones obtained by NASCompares. In his results the SMR drives took ~2x longer when full, and were the same when empty. Does ZFS do something different to regular RAID when rebuilding?

so how do the SMR drives go with NON ZFS raid arrays???? to me that is an important question... Reason I ask is I am currently using SAS drives on my server, and SAS drives in my storage enclosure but later on (when the sas drives eventually fail) I will be looking for replacing the sas units with sata units.... My current setup does not use the ZFS file system.... Is this only an issue for ZFS?

I'm running a supermicro dual Xeon e5 4u 24bay centos 7 rig with an lsi megaraid card with 22 14tb ironwolfs raid 60 3 strips of 7 drives each for the nas storage with 1 hot spare. Then 2x 1tb raid 1 for 2 vm's and the os runs off a 128gb sata SSD. My initial build of the array took 3.5 days. A few months back I had an ironwolf drive that was getting a little loud and making more vibration than I liked. When finally got the guts to swap it out it took 31 hours to rebuild over the weekend with minimal usage except a scheduled off-site incremental backup. I couldn't imagine the stress of that pushing past the 2 day mark.

I know this is an old video but would using a WD Red SMR drive be ok to back up a NAS while rebuilding the array with the older CMR drives?

Did you factor in the smr-group-cylinder-cluster size?

What if the ZFS array is built with SMR drives initially, and then we use CMR drives to replace failed ones? What would be the rebuild time then? (Say, a 3+1 configuration with 4 TB drives). That way we could save money buying the initial drives upfront, and pay more for CMR replacements later as the drives start to fail.

So, that means I had a good "nose" a year ago buying WD REDs for my professionnal back up boxes Is there a good source to know what drives are SMR ? thinking about Seagate here. I know ex Archive drives are SMR, but what about the rest of the EXO family ?

is SMR okay for older games, gigabit Ethernet, music, photos, and things like that? i want to get a 6tb SMR hdd, but I'd rather get a 4tb CMR one if the performance degradation is noticeable

WD shows this on their site now. Apparently there are SMR and CMR models for the regular red non-pro.

Hi I have a WD red 4GB. Just a single drive in a 2-drive system. Sometimes i notice that the NAS is often accessing the drive hard, minutes at a time while i am not using the NAS (reading/writing to NAS). In my earlier system with a WD red 2 GB i never had this behavior. Has this also to do with this SMR/CMR issue? And since when is WD shipping these SMR drives? Hope you can answer this.

How about on a basic 2 drive nas raid 1 setup who already this past december 2019 bought these 4TB smr reds for a new nas? I am thinking it is ok to be using them on a new nas, but when replacing a drive , replace with something else. Like many, I also fell into the trap of relying on past history.

So glad I switched all my new buys to Seagate. Wd has always been a little shady but this is insane.

I used to run WD red 4x6TB in my ZFS raid. Every scrub, ZFS would repair between 64 and 256 KB, always in different places. Scary. I've had no more issues after switching to Seagate IronWolf Pro drives.

Wow, seeing this now as I'm orientating in getting a new NAS with WD RED's in it and looking at the 4TB 256mb version of the disk. The 64mb version is CMR. Would you recommend these over the SMR or skip WD in general? I've have bad experiences with Seagate in the past and rather stick to WD instead. What about Toshiba's N300 disk? Thanks.

So to answer my last question. I don't have to worry about my currently running drives to fail but I need to make sure my replacement drive is a CMR drive?

Some time ago I got my hands on an Adaptec 72405 for almost nothing. So I said „let‘s RAID5 what‘s around“, which were some Seagate Archive 8TBs and some new Toshiba N300 8TBs, eight drives in total. It worked so far, RAID build times were like 12 hours or so. I didn‘t think about the implications of mixing CMR and SMR at that time, but seeing your video now makes me think I should swap the Seagate SMRs for more N300s or similar. OTOH every swap will necessitate a rebuild... Oh boy oO

I'm also wondering about the SAS drives market, I personally either go with shitty Seagates 5TB from a backup plus portable or full blown SAS HGST or EXOS from Seagate, haven't used WD much as of lately myself, but i'm curious to know what else is also SMR. I really don't like the sata drives for anything serious, always preferred sas myself.

Can someone test how SMR drives fair with sequential resilvering? I would like to know if it makes a difference. Thank you

I had to replace the 2 2tb hard drives on my home-made nas using ubuntu and mdadm. I had so many trouble with the newer smr drives that I thrown them and bought HP enterprise-grade hdds. mdadm will always end up marking as bad smr drives, either Seagate or WD.

I don't really follow the high enterprise market for hdd but I am interested in the purple line since I install on a regular basis surveillance systems. Is the purple line affected? It would be counter productive since these drives do continuous writes . Thanks.

I did not know about this. I was curious why rebuilds had been taking a long time though. I logged into my Synology device and went to Storage Manager and lo-and-behold, I have EFAX 6TB drives in every bay..

I have 4 3TB WD RED hard drives 5 months ago i got 4 more 3TB WD Reds from 2019... i put them in and they all faild doing a raid expanshuon on raid 6... Lucky i had a back up i built the RAID and then spent 8hr coppying over all the data. After that the raid comtroler was showing drive errors and warning.. I had to return all 4 drives. All my older 3TB drives still work fine....

Thank you for your work

Thank you again! It does seem I need to find a new vendor when my Red's kick the bucket. A real shame as they have been reliable in ZFS up until this point...

Is there a website that has a table of all the drives?

I just bought two Seagate Barracuda 2tb. Freshly formatted started Crystal Disk Mark and gotten 220mb/s read 200mb/s write. I just put some games of my steam library on it ~160gb and did the test again only to find out it's now at 10mb/s read and ~100mb/s write. And that's all the time no matter if I let it sit for hours (Taskmanager performance 0%). How can I use TRIM on Windows 10?

While I don't think a Z2 is going to be very common for a 4 bay NAS, I would expect it to be MUCH more common for an 8 bay NAS. I'd be interested in knowing just how bad a double failure would be in a fully populated 8 bay (4tb ea) array. It'll take a long time to test probably, but still would be interesting to know. THANK YOU so much for looking into this and getting concrete results into our hands!

Problem is not NAS, problem is RAID. These drives rely on knowing which areas are "empty" (and can be overwritten) and which areas are "occupied" (and must be relocated). When the RAID controller initializes the array, it writes all over the drives, so they appear as "full"; then, any subsequent write operation implies relocating some data. The resilver procedure did not take 15 times longer just because the drives are 15 times slower, it tool 15 times longer because all that data relocation required 15 times more work. Not only these drives are much slower in a RAID array, they are going to die pretty quickly.

Could you guys make a video explaining ZFS

Just about to click "order" on a WD EFAX when I came across this video. *whew*. So, I see EFRX (CMR) are still available. This is what I currently have. I wonder if I should get one of those, or get a different brand CMR drive... Hmmm...

What about WD MyPassport wdbs4b0020bbk-wesn? I can't find witch technology is used..

When I shared this on Facebook it says SMR vs SMR and not CMR.

I saw this video like 5 times in one year. I can't stop myself... I will ask it.. I have 10TB music archive. I am sharing my files with Soulseek. My computer is 24/7 open and not always but most of time it does read due to uploads. And sometimes i download new music files. Now... Due to disk plates will be more on higher capacities and i want less head, surface fails, i will buy 4TB drives. In addition i am trying to buy CMR/PMR drives. WD40PURZ/WD40PURX (I don't know which one is PMR/CMR), WD40EZRZ and WD40EFRX are my options. So which one do you recommend me in these? Thank you.

I Just boubas 2x4tb from them.. should i be worried ?

*Newbie question* Would SMR have the same problems in a ZFS Mirror rebuilds(resync)??

I wonder how SMR affects the platters in the long haul once platter degradation becomes a factor especially for laptops where capacities have been stagnant for over half a decade. I can live with sata 1 speeds but I deeply loathe data loss and performance degradation as the drives age.

This video is long, yes. But exact description of the test procedure is crucial. However, you repeat yourself a lot.

Is there a compiled list of these SMR drives so that I can avoid them?

Do We know, if it would have affected the rebuild time for the worse, had all drives been SMR, not just the one that was replaced?

I don't understand, if you have a drive failed shouldn't you backup your data first before rebuilding?

I had this when it first came to light. I had to do the research. I accidentally bought a 6tb SMR and it was hell. Took an age to rebuild, then performance on my FreeNAS when this resilvered was awful. I'm moving to Ironwolf ST6000VN001 when I need to replace a drive.

Hmm, these results seem to differ from other similar rebuild tests I've seen on YT. In my own tests it seems that having separate SSD CACHE and LOG devices helps alot and is recommened always when using ZFS. And rebuilds on normal RAID1/5/6 setups work just fine on SMR drives.

resilvering should be more performant if the raid layout fits the shingle sizes, so they are not rewritten?

What about the other way around? Nas made of SMR with a new dmr?

I bought one of these Just a week before the story broke. (4Tb) not so happy. Thanks for testing.

love the subject and like the video but everytime I watch this channel I fall asleep :)

1. Why does re-silvering cause so many randomish writes? I thought it would be technically possible to rebuild the drive in a linear fashion. 2. Is rebuilding mirrors also slowed down?

Has anyone done any testing with the 10TB white label WD drives? SMR or CMR?

Recently swapped a bunch of 3TB drives (mix of WD Red and Seagate NAS) for 6TB Reds Speed went from "scan: resilvered 2.55T in 1 days 03:44:02 with 0 errors" to "scan: resilvered 2.56T in 1 days 06:11:03 with 0 errors". Were all WD60EFAX-68SHWN0. 11 drives replaced fine. Didn't consider a bit over a day as too bad. 12th drive failed with write errors though. Ran badblocks, smart test, WD Diag, all good, so tried again, few errors and then got stupid slow, so detached. Tried again with a replacement drive, this time a WD60EFAX-68JH4N0. Once again, drive all good with badblocks, etc. was going fine at a decent speed till faulted with write errors again. Thinking that the 12th WD60EFAX-68SHWN0 was just faulty and the WD60EFAX-68JH4N0 is technically fine, just with the SMR being so slow, writes timed out causing ZFS to label it faulty? Believe all the WD60EFAX-68SHWN0 should be SMR though? Definitely got slower the more of them that were in the array, but not 15+ times slower. In my case, was replacing active drives though, not faulted/removed ones. So not sure if the not so bad speed was that or not all EFAX drives as bad as each other? Might be worth testing replacing active drives as a comparison? Interestingly, same AsRock Rack board, although PCI-E HBA, replacement drives using MB SATA though before being swapped.

I just won't be specifying WD Reds in any future builds. Even if I pick the right model number, can we trust someone in a warehouse to pick the correct product from the shelf? The other use case is that someone is buying a new NAS, and then they will migrate their data from their existing smaller capacity NAS, or they want to organise their existing data from desktop storage and external USB devices into a central location. Most businesses don't start with a NAS from day one. It is something they do when they reach a certain size and their data starts to get unmanageable. So it would seem to me that any NAS drive is going to have a period of high-volume workload when it first comes out of the box. Data migration is something you need to be able to start running on Friday evening and come back to on Monday morning to do the final steps before starting work for the week.

What about 3tb? I bought 7 wd reds 3tb and ran in freenas the last few years. Lost 1 after a year and just recently found out the 3 more are bad in my server. Shut server down. Tested and they have bad sectors. Why? I hardly used the server, the drives just sat there idling for a couple years.

I found this in testing over a year ago. Been thoroughly discussed on the FreeNAS forum. You should check it out.