Hard drive makes noise formatted to EXT4 - but not BTRFS

Ask Fedora
1

Hello all

This is a very odd question.

I’m moving my WinSLOP 10 Jellyfin server to Fedora 44 Workstation and everything went almost well - yes of course there’s an Nvidia GPU :squinting_face_with_tongue: - and then i started moving files around and getting the hard drives formatted to EXT4.

And then suddenly after mocking around with some bad words Nvidia drivers, because i had some smaller graphical glitches and some freezing, so i had to do something - then one of my drives started sounding like it had been dropped and makes that weird noise of running back and forth the same track.

But every test i have run says that there’s no errors - AND it also worked fine under Win10.

I have been in the PC technical bush for around 40 years, so i know my way around the hardware, systems and troubleshooting. I have also from time to time run Linux.

But i have never experienced this odd behaviour.

So i started doing all the test i could find and do, but nothing showed anything wrong. I then tried to format it again, but that didn’t do anything either. Then i tried formatting it to BTRFS and suddenly the weird running back and forth in the same track went away :thinking:

Before i switched to BTRFS, i also did some test on partition sizes, which gave some very odd results. The smaller the drive, the faster the noise stopped by itself.

To my knowledge, Linux doesn’t have the same indexing thing Win10 runs all the time, so the drive will spin up again and again even not in use.

So what’s going on here?

I will just after posting this be swapping the SATA port to check that as well, as the other two drives does not share this odd behaviour running EXT4.

I doubt that will make a difference, but who know with those ghosts in the machine :wink:

All the best
Carsten

2

have you checked smartctl output, esp related to reallocated and current_pending sectors ?

I’d suspect that the hdd may have some area that coudl have issues with growing defects,. ext4 has a different layout and now you’ŕe hitting that area more often than with ntfs before.. Or that area was 100% used by a file, and now it’s available…

Maybe perform a 100% surface check with writes, to be sure.
Was anything logged in journal ?

3

Maybe your disk is failing?

What is the output of sudo smartctl -x /dev/YOUR-DISK?

Also worth looking at the output of sudo dmesg when you hear the disk noise. Do you see errors from the disk being reported?

4

Thank you Mark and Barry :slight_smile:

Everything looks just fine and dandy, the only blood that flows in the logs, are from Nvidia and that nouveau crap. Nothing in regards to the drive.

I have already sent an RMA, that sound gives me nightmares. Haven’t heard it for years, and only know it to be a bad omen.

Looking through the log, when creating the test drives in EXT4, there’s a lot of read around in regards to the nouveau driver - i don’t suppose that can have an effect?

Is the surface write the hardest test i can throw at it?

I just want to throw a nuke at it, it needs to stand the test or break.

Where i bought the disk, they have more in storage. Toshiba themselves, says doing it through them could take a while because of the state of affairs. And also, i’m in Denmark, we have a 2 year warranty out of the box, and it’s not even a year old. Bought it october 2. last year.

5

Have set it to do a: badblocks -wsv /dev/sdd and going to bed, then i’ll see what has happened in the morning :wink:

6

Hmm, how long should a badblocks -w take?

The first 3% didn’t even take 9min and now 9 hours later almost, it has only done 55% - the math says 5 hours, but reality says 18 :thinking:

And there’s still the original question; What’s the difference between how EXT4 and BTRFS handles a drive - why does one make it noisy and the other not?

I know you, Mark, wrote, that maybe EXT4 uses more of the disk than NTFS - and yes, i removed the small gaps Windows leaves on a drive, but again BRTFS doesn’t make gaps either and should therefor be as EXT4 in that way…

7

https://www.linux.org/threads/info-if-you-have-large-hard-disks-and-they-keep-grinding-at-every-reboot.40906/

8

until it’s done :slight_smile:

a smartctl long test would have been my preferred test.
journalctl -c <dev> shows the capabilities and rough estimates
Try a short test first followed by the long test

see SMART tests with smartctl - Thomas-Krenn-Wiki-en

It’s a jellyfin server? You might be better off using XFS anyway.

badblocks, did you adjust the block size (-b) default is 1024 , you might be writing each block 4 times on a hdd with 4k physical sector size.

Please open a new topic about nvidia drivers. there are one or two pitfalls.

9

The health of the drive will be reported in the S.M.A.R.T counters from sudo smartctl -x /dev/sdd output.

badblocks may find something. Modern drive remap bad sectors to spare good sectors, so this scan may not yield insteresting data.
Any remapping is in the smartctl output.

10

I remember reading about badblocks doing 4 kinds of writes, and i can see that it has reset the percentage, and can also see that it says pattern 0x55 now.

And nothing so far.

And my point in regards to Nvidia has nothing to do with the drivers, i will sort that out.

But it has to do with WHETHER they can influence the functioning of a hard drive?

11

why and how?

badblocks sector size,
badblocks won’t start on a 4k native drive, but it won’t complain with drives that still emulate 512 sector size like that one here

Sector Sizes: 512 bytes logical, 4096 bytes physical

update: This may be even worse if you performing this test on a SMR hdd.

12

It like that you had a bad luck and you have SMR HDD. What it means, for example: google “SMR disk compare CMR”.

13

I hadn’t come around to checking what the difference was between the drives physically yet. I noticed looking through the logs, that the newer drive had a different part number, HDWQ140 and the two older ones, HDWG440. All are N300 drives.

If i remember correctly, wasn’t WD that got caught selling SMR and not telling buyers that they were, a couple of years ago.

As far as Toshiba goes, they swear that they’re all CMR drives, and the benchmarks all show the same results on all drives. NAS Hard Drive for Home Office & Small Business | Toshiba N300

So if i may, i would like to reset the conversation back to my original question;

What is the difference between EXT4 and BRTFS?

My point is, that i might not have noticed when i build my streamer, whether the older drives did the same thing.

I did some more testing creating different sizes of partitions. I have an odd idea that the noise weird read/head movement pattern of the drive will eventually go away - as it has done with all the smaller test partitions i have made.

I also formatted in most of the other formats possible in GParted.

Do EXT4 do some checking or indexing that all the other file systems don’t?

14

If the specification document don’t mention CMR than it’s most likely SMR these days.

ext4 / btrfs totally different layout and strategy. There are enough info available on the web.

Hot take, I’ve stopped using ext4 for anything besides /boot and very small devices. I would consider xfs or btrfs over ext4 any time.

gparted , why bother with partitions? Consider using lvm, much more flexible, you can extend logical volumes any time if you keep extends in reserve. But it all depends on what your plan is with those three drives.

15

See, that’s why i supplied a link :wink:

And now an image: https://i.postimg.cc/FFfzGmBf/Screenshot-From-2026-08-09-14-50-14.png

At the moment they’re just single drives, not in any form of RAID, the will be at some point, when hard drives come down in price again :slight_smile:

From what i read, Jellyfin doesn’t like BTRFS that much, hwne i made my streamer, but maybe it has changed, and now with the new RC ver. 12, it might be even better.

And yes, there’s a lot of information, but when searching for hours, almost an entirre day and not finding something, then it’s time to ask someone :wink:

16

UPDATE: As predicted, now a couple hours later the hard drive stopping running back and forth in the same track - OR, it did not have a moment - it’s doing something, but what is it :thinking:

It does do something. I have now tried unmounting it and remounting it. Nothing happened. Then i tried rebooting the machine. Nothing happened.

When remounted, i can hear it reads something, and then just silence.

So whatever happens when a drive is formatted to EXT4, it does something afterwards when mounted.

And this might not be hear’able on most drives. These N300 are known to be more noisy than most drives.

17

That’s an interesting difference between ext4 and Btrfs. Since the drive passes the usual health checks, I’d compare what is accessing the disk with tools like iotop or iostat. If the noise only happens with ext4, checking background filesystem activity and mount options may help identify what is keeping the drive busy.

18

Ahh, ok - i’ll have to reformat the drive again then and start iotop and iostat up before mounting the drive again :slight_smile:

19

I don’t click on everyone’s links :laughing:

I’ve never used or setup a Jellyfin server, so I can’t really comment on that, but they seem to mention ZFS only in their documentation.

regarding ext - btrfs etc. I had written some lines but decided to not post that. But you can get a good idea by reading these links.

Regarding the seeks you coudl hear, It’s not clear what you were doing at that time, still copying continuosly to the drive either large files or many files?
The ext4 journal is located at a fixed position on disk as a circular buffer. and needs constant updating. With default settings only metadata are written to the jourmal, this includes changes to inodes, extended inodes etc.
WIth a large filesystem this could be the seeks you were hearing, while with a smaller filesystem size the area for the heads to move is way smaller and less auduble?

But IMHO, you shouldn’t let ext4 anywhere near a multi TB drive, especially if it’s a filesystem for mainly large video files. EXT filesystem checks are no fun at all and you will run into them at some point (automatic checks) in time regardless of an active journal or not.

My advice would be XFS.

20

Just to take the file system first, i have looked more into BTRFS and i like the benefits of it, especially in regards of a Jellyfin media server.

Yes, ZFS they’re not that happy about in regards to something with the database and datasets - whatever it does to that.

So, back to my drive.

I thought it was clear, that it was an empty just formatted drive; It is - not a single thing, besides the lost&found folder.

So if you care to look at this link, it’s a 1.21min recording of what happens in iotop, when mounting a newly formatted EXT4 drive: Dropbox

I’m not sure what it’s all about, and wondering why it says something with BTRFS - and there’s a clear indication of the disk in writing mode, and when that stops, the noise stops. The smaller the drive, the faster it stops.

It says btrfs-endio-write - and you can actually hear the noise/writing on the recording.

This partition is only 16GB - as written earlier, a 4TB drive takes a couple of hours :grimacing: