TR Forums

Some really weird stuff happened to my WD external drives: I downloaded a 80 GB file with qbittorrent directly on a 3 TB external WD drive. When I wanted to run a surface scan on that drive (with HDTune pro), the read speed on that segment was really slow 20-35 MB/s, only to jump back to 100 MB/s when that portion ended. I tested other WD drives and the same thing happened. But the actual transfer speed remained unaffected: meaning, while I was scanning that portion with max 35 MB/s, when I deleted the file and I tried to copy some other files in that location, the files were transferred at the normal 80-100 MB/s.
The scan speed returned to normal after I overwrote the section where I transferred the torrent data (I used HDSentinel to reinitialize those sectors), but can someone try to guess (and tell me) what the hell is going on?
I can understand a cable going bad, or the controller going bad (in which case the speed would be bad over the whole drive), or a drive developing bad sectors... but what the hell was that???

That's pretty weird. The only semi-plausible theory I can think of is, if the torrent client is (as I assume) using multiple download streams from multiple sources, and writing everything to disk as it comes in, the file may not be getting laid out sequentially on the disk. So when you go to read it back, there's some extra seeking involved, which impacts throughput.

just brew it! wrote:

That's pretty weird. The only semi-plausible theory I can think of is, if the torrent client is (as I assume) using multiple download streams from multiple sources, and writing everything to disk as it comes in, the file may not be getting laid out sequentially on the disk. So when you go to read it back, there's some extra seeking involved, which impacts throughput.

Yes, that is the phenomenon of fragmentation, but it does not apply in this case, because a surface scan with tools like HDTune pro or HD Sentinel does not actually try to read the files per se, but reads the blocks in sequential order to see if they are readable.

Are these new-ish drives? If so they may be drive-managed SMR, in which case all bets are off. Some SMR implementations (and IIRC WD's is one of them...) can do all sorts of weird sector/zone remapping behind your back, so even sequential LBAs may not necessarily be sequential on the physical media. IOW, you can potentially have fragmentation at the drive level (invisible to the OS other than as a performance hit), depending on how the data was written, and how the drive's firmware deals with SMR GC operations.

If you know the full model number of the bare drive inside the external enclosure we may be able to confirm/refute whether it is SMR.

just brew it! wrote:

Are these new-ish drives?

Yes.

If you know the full model number of the bare drive inside the external enclosure we may be able to confirm/refute whether it is SMR.

WDC WD30NMZW-11GX6S1

Some SMR implementations (and IIRC WD's is one of them...) can do all sorts of weird sector/zone remapping

Slowly but surely WD are making their way on my **** list...

I'm afraid it looks like that HDD SKU is infected with shingles.

Krogoth wrote:

I'm afraid it looks like that HDD SKU is infected with shingles.

It's SMR?

Noldor wrote:

Krogoth wrote:

I'm afraid it looks like that HDD SKU is infected with shingles.

It's SMR?

Looks like it.

Unfortunately Seagate isn't a solution either, since they also have a number of SMR drives in their lineup. And manufacturers have been known to switch the model of bare drive used in a given model of external drive, so buying the same model of external as one that was non-SMR in the past is no guarantee either.

The only way to be sure these days is to do some research, buy a bare drive that you KNOW is non-SMR, and put it in a third-party enclosure.

Known to be infected (AVOID): Most (if not all) of Seagate Barracuda Compute (non-Pro) line; select models of WD Blue and Red (seems to be mostly the ones with 256MB cache)

Known clean (at least for now): Seagate Ironwolf; WD Red Pro; WD Purple

The days of being able to just buy any random model of HDD, and expect it to perform, well, like a normal HDD, are gone. Probably for good, unless someone manages to challenge this sort of behavior in court and forces the manufacturers to note in their marketing literature which drives are SMR. Most of the non-enterprise ones don't even mention it in the technical data sheet, which is unforgivable IMO.

At least be thankful you're not trying to use it in a RAID array. Running a workload that does a lot of random writes to a RAID array containing drive-managed SMR drives can actually cause the entire array to degrade and fail due to I/O timeouts on the SMR drives.

just brew it! wrote:

Noldor wrote:

Krogoth wrote:

I'm afraid it looks like that HDD SKU is infected with shingles.

It's SMR?

Looks like it.

Unfortunately Seagate isn't a solution either, since they also have a number of SMR drives in their lineup. And manufacturers have been known to switch the model of bare drive used in a given model of external drive, so buying the same model of external as one that was non-SMR in the past is no guarantee either.

The only way to be sure these days is to do some research, buy a bare drive that you KNOW is non-SMR, and put it in a third-party enclosure.

Known to be infected (AVOID): Most (if not all) of Seagate Barracuda Compute (non-Pro) line; select models of WD Blue and Red (seems to be mostly the ones with 256MB cache)

Known clean (at least for now): Seagate Ironwolf; WD Red Pro; WD Purple

Until now Seagate did not give me any trouble, though.

Noldor wrote:

Until now Seagate did not give me any trouble, though.

I would assume you have not encountered any of the SMR Barracuda Compute drives then. They are horrible. I got burned quite badly buying a pile of them a couple of summers ago to upgrade the capacity of my home server. Had to replace them all. I now use those drives solely for external backups, and just put up with the crappy performance since I don't really care if the backup (and verify... I verify my backups) takes all night (or even longer).

just brew it! wrote:

Noldor wrote:

Until now Seagate did not give me any trouble, though.

I would assume you have not encountered any of the SMR Barracuda Compute drives then. They are horrible. I got burned quite badly buying a pile of them a couple of summers ago to upgrade the capacity of my home server. Had to replace them all. I now use those drives solely for external backups, and just put up with the crappy performance since I don't really care if the backup (and verify... I verify my backups) takes all night (or even longer).

What does this poor performance consist of? Slow writes, slow reads or something else?

As for WD, to be fair, this is more of a annoyance than a problem because the transfer speed is not affected, but, on the other hand, if one day I want to do a surface scan to check for bad sectors, I don't want it to take the whole day or more.

Noldor wrote:

just brew it! wrote:

Noldor wrote:

Until now Seagate did not give me any trouble, though.

I would assume you have not encountered any of the SMR Barracuda Compute drives then. They are horrible. I got burned quite badly buying a pile of them a couple of summers ago to upgrade the capacity of my home server. Had to replace them all. I now use those drives solely for external backups, and just put up with the crappy performance since I don't really care if the backup (and verify... I verify my backups) takes all night (or even longer).

What does this poor performance consist of? Slow writes, slow reads or something else?

Painfully slow writes at times, and need to make sure the "noatime" option (this is Linux with ext4 file system) is set when the volume gets mounted or things get even worse.

Edit: I see some read slowdowns similar to yours as well. There's a backup verifying as I type this, and it is getting ~20 MB/sec, with utilization of the drive hovering just below 100% (according to the Linux iostat command).

just brew it! wrote:

unless someone manages to challenge this sort of behavior in court and forces the manufacturers to note in their marketing literature which drives are SMR

Wasn't that roughly what happened a month ago? (With some yet to join the fray.)

meerkt wrote:

just brew it! wrote:

unless someone manages to challenge this sort of behavior in court and forces the manufacturers to note in their marketing literature which drives are SMR

Wasn't that roughly what happened a month ago? (With some yet to join the fray.)

AFAIK they haven't updated their marketing materials yet. Burying it in a corporate blog post doesn't help the average Joe much.

Edit: I stand corrected. The product page for WD Red does indeed now indicate which ones are SMR!

Apparently Toshiba also 'fessed up at the end of April:

https://blocksandfiles.com/2020/04/29/t ... -smr-list/

Fortunately, I just bought the N300 for a new system build. It's on the clean list.

just brew it! wrote:

Noldor wrote:

Krogoth wrote:

I'm afraid it looks like that HDD SKU is infected with shingles.

It's SMR?

Looks like it.

Unfortunately Seagate isn't a solution either, since they also have a number of SMR drives in their lineup. And manufacturers have been known to switch the model of bare drive used in a given model of external drive, so buying the same model of external as one that was non-SMR in the past is no guarantee either.

The only way to be sure these days is to do some research, buy a bare drive that you KNOW is non-SMR, and put it in a third-party enclosure.

Known to be infected (AVOID): Most (if not all) of Seagate Barracuda Compute (non-Pro) line; select models of WD Blue and Red (seems to be mostly the ones with 256MB cache)

Known clean (at least for now): Seagate Ironwolf; WD Red Pro; WD Purple

The days of being able to just buy any random model of HDD, and expect it to perform, well, like a normal HDD, are gone. Probably for good, unless someone manages to challenge this sort of behavior in court and forces the manufacturers to note in their marketing literature which drives are SMR. Most of the non-enterprise ones don't even mention it in the technical data sheet, which is unforgivable IMO.

At least be thankful you're not trying to use it in a RAID array. Running a workload that does a lot of random writes to a RAID array containing drive-managed SMR drives can actually cause the entire array to degrade and fail due to I/O timeouts on the SMR drives.

Seagate enterprise drives do not have this. Get a IronWolf - no SMR.

If I was in the market the IronWolf's are probably what I'd be looking at. Fortunately the 4x2TB WD Red's I got about two years ago are still doing well in the backup box.

DiMaestro wrote:

Seagate enterprise drives do not have this. Get a IronWolf - no SMR.

Not entirely true. Some of their enterprise drives are SMR, but they do (and AFAIK always have) mentioned it in the specs when this is the case. It was the consumer line where all the obfuscation occurred.

And yes, IronWolf should be safe.

just brew it! wrote:

DiMaestro wrote:

Seagate enterprise drives do not have this. Get a IronWolf - no SMR.

Not entirely true. Some of their enterprise drives are SMR, but they do (and AFAIK always have) mentioned it in the specs when this is the case. It was the consumer line where all the obfuscation occurred.

And yes, IronWolf should be safe.

You're correct. I should have specified Iron Wolf at first. Does WD have any drives that they say are SMR free?

DiMaestro wrote:

You're correct. I should have specified Iron Wolf at first. Does WD have any drives that they say are SMR free?

After all the brouhaha in the tech press, I believe they have updated their site to indicate which model numbers of internal drive are CMR and which are SMR. But I don't think this helps with USB externals (which this thread was originally about). Hence my recommendation to put an internal drive in a 3rd party enclosure. Unfortunately this costs more than just getting a USB external from the get-go.

just brew it! wrote:

DiMaestro wrote:

You're correct. I should have specified Iron Wolf at first. Does WD have any drives that they say are SMR free?

After all the brouhaha in the tech press, I believe they have updated their site to indicate which model numbers of internal drive are CMR and which are SMR. But I don't think this helps with USB externals (which this thread was originally about). Hence my recommendation to put an internal drive in a 3rd party enclosure. Unfortunately this costs more than just getting a USB external from the get-go.

Man I must be old - I can't imagine using a USB enclosure or dock for anything I consider important. Dockwise if it doesn't support Esata I won't even bother with it. USB 3 is ok - but I've seen it just fall over itself when the CPU is in high use.

Nothing but a real NAS/Storage machine for me.

DiMaestro wrote:

just brew it! wrote:

DiMaestro wrote:

You're correct. I should have specified Iron Wolf at first. Does WD have any drives that they say are SMR free?

After all the brouhaha in the tech press, I believe they have updated their site to indicate which model numbers of internal drive are CMR and which are SMR. But I don't think this helps with USB externals (which this thread was originally about). Hence my recommendation to put an internal drive in a 3rd party enclosure. Unfortunately this costs more than just getting a USB external from the get-go.

Man I must be old - I can't imagine using a USB enclosure or dock for anything I consider important. Dockwise if it doesn't support Esata I won't even bother with it. USB 3 is ok - but I've seen it just fall over itself when the CPU is in high use.

Nothing but a real NAS/Storage machine for me.

Lots of people shuck external USB drives and put the drive in a NAS.

Funny class action text (linked from here):

The hard drives are completely worthless for their intended purpose—and are in fact dangerous to customer data.