Western Digital's Red 2 - 6TB NAS drives apparently aren't good for NAS use?

[GDRRiley]

3 minutes ago, leadeater said:

I mean how many posts are we at and you're still missing the point, thus proves I cannot force people to learn.

You can always force people to learn. The usually methods are just frowned upon. 
 

I’d hope WD would have done more testing than it appears they did. I swear WD don’t become like Microsoft 

[leadeater]

39 minutes ago, GDRRiley said:

I’d hope WD would have done more testing than it appears they did. I swear WD don’t become like Microsoft 

The disks failing to rebuild in a Synology NAS which is the primary use case target for these disks is a big fail. From memory I think the disks that come with a Synolgy NAS if you don't go diskless are WD disks, or at least some of them are. Wonder how that situation is going right now, Synology putting pressure on WD would get things moving real fast.

 

Edit:

Currently none of the WD Red 6TB and below FAX disks are on their HCL.

[GDRRiley]

10 minutes ago, leadeater said:

The disks failing to rebuild in a Synology NAS which is the primary use case target for these disks is a big fail. From memory I think the disks that come with a Synolgy NAS if you don't go diskless are WD disks, or at least some of them are. Wonder how that situation is going right now, Synology putting pressure on WD would get things moving real fast.

 

Edit:

Currently none of the WD Red 6TB and below FAX disks are on their HCL.

who forgot to test? I feel like that is true too.

I assume it is going badly for WD, given they are based about 20 miles from me and are likely having most people working from home.

yeah pressure from big OEMs does that.

[TechyBen]

3 hours ago, leadeater said:

I never said it was an invalid request, I said the reasoning applied was invalid. You can ask things for the wrong reasoning, it doesn't mean I'm saying nobody should ask for that information or asking for it is wrong.

 

Like I said if you want to know sure, we all do.

 

It's also my right to correct them if they go about spreading misinformation, be wrong in your head, do it out loud and you'll be corrected. You should not fear or be annoyed at gaining better understanding, we have all been wrong about something at some point, many times, and it'll happen to us all again. What you do with information matters, you can decide to take it on-board or you can ignore it. I cannot force anyone to learn.

 

I mean how many posts are we at and you're still missing the point, thus proves I cannot force people to learn.

Where am I afraid of better understanding? Someone got an SMR drive, 2 of them failed building an array/raid. They replaced them with PMR drives (of the similar manufacture/design) and it built/did not fail. Thus they requested PMR drives for their arrays/raid/nas.

 

Yes somehow, that's wrong? I'm wrong for agreeing with them that it would have been easier if they'd been told SMR/PMR so they could match up their raids.

 

You know, like tires, bulbs, etc etc. "Materially identical" in marketing/engineering speak, 99% of the time is not the same.

[JSaville]

Here's a blog post from Western Digital addressing the issue. 

 

https://blog.westerndigital.com/wd-red-nas-drives/?amp

 

[Lady Fitzgerald]

1 hour ago, JSaville said:

Here's a blog post from Western Digital addressing the issue. 

 

https://blog.westerndigital.com/wd-red-nas-drives/?amp

 

I'm going to print that out, shred it, and spread it on my yard. It should make even my gravel yard grow green.

[mr moose]

1 hour ago, JSaville said:

Here's a blog post from Western Digital addressing the issue. 

 

https://blog.westerndigital.com/wd-red-nas-drives/?amp

 

 

So they basically have given us no new information.    It will be interesting to see an in depth breakdown of what the issue might be, how many people experienced it and what the cause was or most likely is.  Until then, be careful buying WD reds for raid situations.

 

 

[Vishera]

On 4/17/2020 at 6:57 AM, GDRRiley said:

I swear WD, Toshiba and Seagate this is some let’s get a nice class actions suit. WD for new reds, Toshiba on their p300 and Seagate on whatever desktop models have it. 
If this doesn’t count as false advertising and failure to discloses what the product is, time to rewrite those laws. 

The P300 is not for NAS use,and only few specific models of the P300 lineup are SMR,

The P300 lineup is for consumer desktops,I actually own a P300 3TB CMR drive.

So only Western Digital is in the wrong.

[leadeater]

23 hours ago, TechyBen said:

Where am I afraid of better understanding? Someone got an SMR drive, 2 of them failed building an array/raid. They replaced them with PMR drives (of the similar manufacture/design) and it built/did not fail. Thus they requested PMR drives for their arrays/raid/nas.

 

Yes somehow, that's wrong? I'm wrong for agreeing with them that it would have been easier if they'd been told SMR/PMR so they could match up their raids.

I never said that part was wrong, I said the part where you are attributing the failure to DM-SMR is and not from the firmware error reported by one of the people having the issue who looked in to why it was failing.

 

It's wrong to jump to conclusions based on incorrect assumptions. And no doubt you'll do it again, probably as a reply to this so this will be my last rely to you. You don't want to get it, it's not even a case of you not able to, you just don't want to listen to the point being made and you don't actually care. It's fine not to actually care why there is a fault, not everyone has to care.

 

Not wanting to find out why it is failing is not gaining a better understanding however. You could of taken on-board what was said and information that was provided, you don't seem to want to though.

 

FYI Seagate Archive v2 disk which is a DM-SMR disk rebuilds in a mismatched RAID array without failing, failing is unique to the WD Reds. The rebuild is much slower, it does not fail.

[RejZoR]

@leadeater

Even if things don't fail, people don't buy new HW to be slower. That's like saying "Yes, I want to buy latest SSD so my system will be slower than with last one!".

[leadeater]

21 minutes ago, RejZoR said:

@leadeater

Even if things don't fail, people don't buy new HW to be slower. That's like saying "Yes, I want to buy latest SSD so my system will be slower than with last one!".

Yes and as I explained and have shown there are aspects of these new DM-SMR disks that are faster than the old ones. DM-SMR is not half the speed of PMR, it's really not that. While the PMR zone is not full and due to the larger caches they are faster than comparable priced less cache PMR disks.

 

DM-SMR are 100 times slower when the PMR zone is full and will only service cleanup operations and not host I/O commands.

 

DM-SMR disks are both slower and faster, are not perfect and neither are PMR disks. There is a reason there are WD Red (the old revision) and WD Pro and WD Gold or WD Ultrastar or any number of more expensive higher end PMR only disks and you can put all these disks through an array of testing and see the difference.

 

Just saying DM-SMR drives are slower really is not accurate at all. If you were to buy one of these and copy many small files to it it'll be 10 times faster than any PMR disk that doesn't also come with 256MB cache or higher and there are zero of those in this price category.

 

 

[2FA]

On 4/17/2020 at 7:59 PM, StDragon said:

Holy shit! That explains soooo much with a 6TB WD Red Pro dropping out of RAID. Spent countless hours years ago on the phone with a Synology engineer. RMA's of equipment, finger pointing, etc. In the end all got replaced with Seagate IronWolf drives as each Red dropped out one by one over the span of a year. And the real kicker, diags confirmed that not a single Red drive had any errors. All perfectly healthy.

 

Wow, total closure to that saga. And yet, I'm extremely pissed off at WD now!!!

Red Pros aren't SMR at all, only regular Reds. Your issue is different.

[StDragon]

1 hour ago, 2FA said:

Red Pros aren't SMR at all, only regular Reds. Your issue is different.

I'm thinking you're right, that the Red Pros use PMR (CMR) instead of SMR from what little info I can obtain online. WD doesn't specify eitherway.

 

That I'll said, this is what WD says

 

Exclusive NASware 3.0 technology

"Our exclusive advanced firmware technology, NASware 3.0 enables seamless integration, robust data protection and optimal performance for NAS systems operating under heavy demand."

 

*cough* BULLSHIT! *cough*

[spartaman64]

i dont understand whats the point of this. if a consumer bought a SMR drive and its slow or doesnt work in their raid array the companies might think they are pulling a fast one on them but in reality the consumer is going to think that the company makes shit products and not buy from them

[Jonotallica]

I'm not an expert on this stuff by any means, have never run a RAID setup in my life.  But I found the news about this a few weeks ago to be very frustrating.

 

I bought my first WD Red drive around 6 months ago, haven't used it much yet but will be using it soon.  To have to dig around and find out about whether it's CMR or PMR, SMR etc.. very frustrating.  It would have been nice if this stuff was stated upfront.  Fortunately I got the 10TB version.

 

In my use case, it might not make much difference.. but when spending that much money.. why risk it?  I now have a second 10TB, and hope to get a third in the near future, but will be avoiding the other models like the plague.  Not because they might fail or create errors (they probably wouldn't.. my drives get plenty of idle time).. but still.. I hate the idea of taking the risk.  Or to change something because to me if it ain't broke, don't fix it.  I was lucky.. but it was frustrating because if I chose the 6TB model for example.. I'd be really cheesed off.  WD's marketing and PR have only further annoyed me.

 

My simplified logic.. all Pro and Gold models use it.  My current (many years old) Toshibas use it.  I've never had a problem or lost any data from a HDD before.  I value my data and want it to stay that way.  But mainly.. I like to be aware of the choices I am making in advance.  There's literally no way of knowing that when you are buying a 6TB Red, that you aren't getting the same thing as the 10TB.  Until now..

 

I stuck with the 10TB because I really like it's power usage and noise characteristics.  I hope to have many years of good experience with them.  And stacks of storage space.  I haven't tested it yet, but I'm also hoping for a nice reduction in idle wattage, or to the very least.. less power used per TB of space in my system.  But yeah..

 

It's not just WD either.  Others are doing it too.  If one compares the options.. sees a reduced price and says "I'll go with the SMR".. that's different.  I bought my first 10TB Red around six months ago with no idea that I was making this choice.. and it's only luck that helped me.  I would have just got the Red Pro's had I of known.  If there was warning in advance ie.. honest marketing and specifications.

 

Apologies if this makes the thread cluttered, it's ok for the mods to edit it.. but I'll post this table here.  I recently saved it to a text file earlier today.

 

D10EFRX     Red     1TB     CMR     3.5inch
WD20EFRX -old     Red     2TB     CMR     3.5inch
WD20EFAX     Red     2TB     SMR     3.5inch
WD30EFRX -old     Red     3TB     CMR     3.5inch
WD30EFAX     Red     3TB     SMR     3.5inch
WD40EFRX -old     Red     4TB     CMR     3.5inch
WD40EFAX     Red     4TB     SMR     3.5inch
WD60EFRX -old     Red     6TB     CMR     3.5inch
WD60EFAX     Red     6TB     SMR     3.5inch
WD80EFAX     Red     8TB     CMR     3.5inch
WD100EFAX     Red     10TB     CMR     3.5inch
WD101EFAX     Red     10TB     CMR     3.5inch
WD120EFAX     Red     12TB     CMR     3.5inch
                
WD20EZAZ     Blue     2TB     SMR     3.5inch
WD60EZAZ     Blue     6TB     SMR     3.5inch
                
WD10SPZX     Blue     1TB     SMR     2.5inch
WD20SPZX     Blue     2TB     SMR     2.5inch
                
WD10SPSX     Black     1TB     SMR     2.5inch
                
WD2002FFSX     Red Pro     2TB     CMR     3.5inch
WD4003FFBX     Red Pro     4TB     CMR     3.5inch
WD6003FFBX     Red Pro     6TB     CMR     3.5inch
WD8003FFBX     Red Pro     8TB     CMR     3.5inch
WD101KFBX     Red Pro     10TB     CMR     3.5inch
WD102KFBX     Red Pro     10TB     CMR     3.5inch
WD121KFBX     Red Pro     12TB     CMR     3.5inch
WD141KFGX     Red Pro     14TB     CMR     3.5inch
                
WD1005FBYZ     Gold     1TB     CMR     3.5inch
WD2005FBYZ     Gold     2TB     CMR     3.5inch
WD4002FYYZ     Gold     4TB     CMR     3.5inch
WD4003FRYZ     Gold     4TB     CMR     3.5inch
WD6002FRYZ     Gold     6TB     CMR     3.5inch
WD6003FRYZ     Gold     6TB     CMR     3.5inch
WD8003FRYZ     Gold     8TB     CMR     3.5inch
WD8004FRYZ     Gold     8TB     CMR     3.5inch
WD101KRYZ     Gold     10TB     CMR     3.5inch
WD102KRYZ     Gold     10TB     CMR     3.5inch
WD121KRYZ     Gold     12TB     CMR     3.5inch
WD121KRYZ     Gold     12TB     CMR     3.5inch
WD141KRYZ     Gold     14TB     CMR     3.5inch

 

TLDR version.. all use CMR except for the Blues, Black and the Red 1TB-6TB.  There's older versions of those (that are still floating around some places) that use CMR.

[mr moose]

4 hours ago, Jonotallica said:

snip

There are performance testing videos in the other thread,  the easiest way to tell the reds apart is that the SMR models are all AX and the CMR are all RX. 

 

For the most part if you are using it as a single drive you likely won't notice any difference at all unless you are copying an extraordinary amount of data in one go. (which is why they don't list it int he specs)

[TechyBen]

This part is important in the latest video:

Despite certain claim here, no one has ever said SMR is unusable as a HDD. But that certain NAS, and in this cas ZFS, tank performance to the requirement users need to know if the drive they are buying is SMR or not (drive managed or OS managed).

 

I'll still dive out of this "discussion", but I do have to note this information, should anyone else look into this thread for info, and think SMR is fantastic, or think it's rubbish, and not just think sensibly and check their use case before deciding.

[StDragon]

2 hours ago, TechyBen said:

This part is important in the latest video:

Despite certain claim here, no one has ever said SMR is unusable as a HDD. But that certain NAS, and in this cas ZFS, tank performance to the requirement users need to know if the drive they are buying is SMR or not (drive managed or OS managed).

 

I'll still dive out of this "discussion", but I do have to note this information, should anyone else look into this thread for info, and think SMR is fantastic, or think it's rubbish, and not just think sensibly and check their use case before deciding.

If the drive is rated for NAS, then that is the use case. Hardware or software RAID doesn't matter.

 

Western Digital has over 61,000 employee on the payroll. There's a whole lot of fail there.

 

IMHO, there's a bell curve of sorts. There's an optimum number of employees and once you pass the number, it goes down-hill from there and will even work against you with internal fiefdoms forming and competing against each-other.

[leadeater]

23 minutes ago, StDragon said:

If the drive is rated for NAS, then that is the use case. Hardware or software RAID doesn't matter.

Not all software RAID is the same or act the same, neither do hardware RAIDs for that matter either. A disk can still be designed for NAS usage and have low performance and may even have compatibility issues with certain products on the market. Compatibility issues can be fixed and even if they cannot that's where product advisories come in, that still doesn't make something not NAS rated because it doesn't work optimally in some cases.

 

If WD want to sell a bunch of SMR disks with NAS firmware and vibration compensation in them who are you or anyone else to say they are not allowed to do that? As long as you know what you are getting what is the problem with that? If I want a giant archive NAS where the disks are mostly always spun down with heads parked is SMR not ideal for that? Archive disks sold on the market to date are not designed for RAID and don't have the firmware optimizations for that and some are HM-SMR which means they would outright not work at all on a hardware RAID controller and if you want to use software RAID it has to issue all the SMR commands which none do currently, this is why DM-SMR exists and to a lesser extent HA-SMR.

 

If WD want to make a product line of disks nobody wants to buy let them it's their company and their decision to make. As long as buyers know what they are getting there isn't an issue, it may even turn out that's what a lot of people actually want.

 

Edit: What makes Linus's video so much better than all the other media coverage of this is he actually took the time to look at what is going on and explain it, where it is applicable to and how it relates to the market. He didn't just jump on in pitchfork raised declaring it's all bad burn it to the ground.

 

On the HGST ZFS conference there is a little bit that needs addressing though, as far as I was aware the comments about needing significant development time to support SMR was for HM-SMR and HA-SMR (implementing the SMR commands so effectively HM-SMR). HGST has always held the position that the proper way to utilize SMR is to do it fully managed but the problem being faced is extremely few are implementing SMR awareness and management so you get HA-SMR without issuing SMR commands which is the same thing as DM-SMR.

[StDragon]

1 hour ago, leadeater said:

Not all software RAID is the same or act the same, neither do hardware RAIDs for that matter either. A disk can still be designed for NAS usage and have low performance and may even have compatibility issues with certain products on the market. Compatibility issues can be fixed and even if they cannot that's where product advisories come in, that still doesn't make something not NAS rated because it doesn't work optimally in some cases.

 

If WD want to sell a bunch of SMR disks with NAS firmware and vibration compensation in them who are you or anyone else to say they are not allowed to do that? As long as you know what you are getting what is the problem with that?

My point is that WD advertised a series of drives as NAS suitable. In reality, they should have created NAS-DDF and NAS-ZFS sub-models 

 

I guess that's why Synology limits the make/model of drives they recommend; because they clearly have to vet the usage case scenario with a QA/QC process that essentially calls BS to any marketing label slapped on a spindle. Obviously, no?

[leadeater]

42 minutes ago, StDragon said:

In reality, they should have created NAS-DDF and NAS-ZFS sub-models

DDF?

 

You don't need sub models at all, just a product line. You don't need to make it confusing like that, just don't use SMR in Red, job done. WD Orange, WD Brown, WD Pink, we have plenty of colours and no reason to not use them.

 

42 minutes ago, StDragon said:

I guess that's why Synology limits the make/model of drives they recommend; because they clearly have to vet the usage case scenario with a QA/QC process that essentially calls BS to any marketing label slapped on a spindle. Obviously, no?

Synology had them listed as supported before, they only recently got removed. That's the problem with how QVL's work, new iterations of existing products don't generally go through actual testing nor need to as the same disk with the same firmware built the same way just with more platters or denser platters don't generally need to be tested and just get added to the QVL list, because it's the same but bigger. DM-SMR is not the same but bigger.

 

Synology implements mdadm RAID with BTRFS filesystem on top of that which is actually a rather odd/uncommon thing to do as BTRFS itself can do RAID but here it is just being used as a filesystem. Like ZFS BTRFS is a Copy-On Write (actually Redirect-On Write both of them) filesystem which if I had to guess plays a part in compatibility issues with DM-SMR. Redirect-On Write always writes data to a new area of the disk, even when modifying existing data, which is about the worst possible thing you can do with DM-SMR looking at how the firmware of these disks work.

 

If you are constantly writing data to new areas of the disk you're not going to be utilizing any of the existing data in the PMR cache zone of the disk for write operations so you've turned what was a read/write cache in to a semi write-back cache and writing more data in to it than otherwise, you're also requiring more background cleanup of the SMR zone as well.

 

I think ZFS is hit harder than Synology/BTRFS because of the data chunk size ZFS uses which is quite small, I think it was Arstech? that did a test using traditional RAID but using a much smaller stripe size and also got way worse rebuild times. COW/ROW filesystems and small stripe sizes look to be not ideal at all.

 

Most other NAS's use mdadm and ext3/ext4 or something else not COW/ROW and RAID stripe sizes used are 256KB or larger so avoiding that double up pain of ZFS.

[mr moose]

If a company makes a product and it works in 95% of the cases it's advertised but for some reason it doesn't work in 5% and as a result WD give those customers a refund/exchange for a drive that does work.  how is that false advertising?

 

 

[StDragon]

1 hour ago, leadeater said:

DDF?

Common RAID Disk Data Format ; it's a common striping format that can be implemented in both software (fake RAID) and hardware based HBAs.

 

Striping size can be changed, but it's really a sliding scale that can sacrifice IOPS for throughput and vice versa (as I understand it). Optimal is somewhere in the middle and is usually defaulted based on drive size and count within a given RAID type. I would imagine that striping size would have a huge impact with SMR based drives.

[leadeater]

13 minutes ago, StDragon said:

Optimal is somewhere in the middle and is usually defaulted based on drive size and count within a given RAID type

You changed it based on workload pattern and I/O type, but that's kinda old school now as SSD has wiped all this away basically. Database server arrays used 64KB stripe size as that is the size of the I/O operations that is used so each operation would nicely fit on a single disk therefore multiple operations span nicely across the array. General purpose file servers used much larger stripe sizes to reduce load on the RAID controller (usually parity RAID) and to decrease fragmentation and I/O padding, basically everyone defaulted to either 256KB or 1MB stripe size. You don't actually change the stripe size based on the number of disks or the size of them, it's all about the I/O operation size you are doing.

 

ZFS from memory does 128KB operations, double check that though as I'm not 100% on that but it's not very big anyway.

[leadeater]

20 minutes ago, Escanor said:

I wish software raid supported trim under windows.

It does, use Storage Spaces.