**** Looks like it's actually causing corruption the other way, when copied/read from NAS into another device (even when streamed from NAS)*************

Have been pulling my hair out for a week now so I figured it was time to ask for help.

Running Unraid Unleashed 7.2.4

Specs: i5 4690k, z97 G55 Motherboard, 1080ti, 4 x 4GB 1866MHz RAM

Storage: 1Tb Cache NVME SSD, 4 x 24TB Exos HDDs

HBA: 9207-8i

NIC: Intel X540

Everything connected locally via 10GBe (Switch etc.. also 10GB)

Essentially the issue is as stated in the title. I initially noticed some quality degradation when playing files via Jellyfin or from the NAS with VLC and thought it was related to jellyfin or network issues. Eventually realized the hashes didn't match between files so something was going wrong in the file transfer. I am using Windows file transfer, robocopy and teracopy for each of the following tests (have done every combination so many times)

Here's what i've done so far:

windows 11 Desktop -> Unraid NAS Cache -> array = MD5 doesn't match

disable cache for array and then transfer = Md5 Mismatch

Windows machine to another drive on the same desktop = MD5 Match

Windows Desktop -> Share on Cache = Mismatch

At this point I reran memtest for 28 hours , no issues. I also took out the NIC and changed the ethernet cable and Switched to another port on the switch (just in case). All of the above results were the same.

Next i tried transfers within unraid.

Array -> cache and cache to array = Mismatch

Cache ->array = mismatch

then i used chatgpt to help diagnose the issue and it made me create some 1gb files in ram and hash those a few times to make sure they matched. They did

Then made it create a .bin of that on the array and get the hash of that to see if it still matched, it did.

When i compare hashes generated (of the same file) within unraid to the hash that teracopy or powershell creates they are different (same as the one in windows explorer - i know it's some utility i added before, but i don't remember what it is). This is confusing. Additionally when i generate MD5 within unraid of a target file it matches perfectly to the same file that I move within unraid (as long as i check the hash with the unraid terminal).

But this is where it gets weirder. If i delete that file and copy from the windows machine to the NAS again. The hash is once again different as to the desktop (and different to the last copy where it was also different) but when i run the hash within unraid terminal, it actually matches the previous runs of the same file before deletion. Yet the file is definitely corrupt and the quality loss is clear.

Point being i have no idea what is going on and desperately need help.

Thank you

Edit: here's something interesting. On one test file i transferred from PC to NAS the hash matches but only if i compare the locally derived hash to the locally derived hash (windows side md5 for the windows copy and terminal MD5 for the unraid copy) If I compare them with a utility like teracopy that's supposed to compare them automatically, it doesn't match. And if i use the windows utility to generate MD5 for both of them it doesn't match. I thought the tool might be broken, but unlikely that powershell, file explorer and tertacopy are all faulty. does calculating md5 for a network attached file not work?

EDIT 2: Seems like the real issue is in reading files from the NAS. However it's done results in corruption:

• taking a hash of the file from windows reads as a hash mismatch because of some error during the read.

• transferring the file from unraid to windows result in hash mismatch.

• It also looks like jellyfin reading the files and streaming them results in some kind of corruption

Edited March 20Mar 20 by Thepn

Attach Diagnostics ZIP to your NEXT post in this thread.

diagnostics-20260319-1242.zip

Assuming both PCs have gigabit NICs as well, do a peer-to-peer connection using them and retest; that would rule out the 10GbE LAN completely.

14 minutes ago, JorgeB said:

Assuming both PCs have gigabit NICs as well, do a peer-to-peer connection using them and retest; that would rule out the 10GbE LAN completely.

I Am not quite sure how to do that. Would the nas still show up as it does right now? or is it a whole new thing to setup

Use an IP from a different subnet to set them up, then connect via that IP, that will always use those NICs.

3 minutes ago, JorgeB said:

Use an IP from a different subnet to set them up, then connect via that IP, that will always use those NICs.

I apologize for my noobness here but I don't know what that means or how to do that

Same as this, just in this case you would use the gigabit NICs in peer-to-peer:

5 hours ago, JorgeB said:

Assuming both PCs have gigabit NICs as well, do a peer-to-peer connection using them and retest; that would rule out the 10GbE LAN completely.

Same behaviour as before

6 hours ago, Thepn said:

Next i tried transfers within unraid.

Array -> cache and cache to array = Mismatch

So, this not network relate. I have same issue in a new build before, after troubleshooting few days, problem gone suddenly and no more again ( I have RMA the mobo later ).

6 hours ago, Thepn said:

does calculating md5 for a network attached file not work?

If you hash remote file through network got mismatch, then it should be other issue.

A quicker way to make test ( no need copy whole file and hash ), is use rsync in therminal, copy file, it will recopy again if problem happen.

rsync /aaa/bbb/sourcefile /ccc/ddd/destfile --progress

Edited March 19Mar 19 by Vr2Io

1 hour ago, Vr2Io said:

So, this not network relate. I have same issue in a new build before, after troubleshooting few days, problem gone suddenly and no more again ( I have RMA the mobo later ).

If you hash remote file through network got mismatch, then it should be other issue.

A quicker way to make test ( no need copy whole file and hash ), is use rsync in therminal, copy file, it will recopy again if problem happen.

rsync /aaa/bbb/sourcefile /ccc/ddd/destfile --progress

I'm trying to use rsync but i keep getting ssh_dispatch_run_fatal: Connection to 192.168.1.200 port 22: message authentication code incorrect

25 minutes ago, Thepn said:

I'm trying to use rsync but i keep getting ssh_dispatch_run_fatal: Connection to 192.168.1.200 port 22: message authentication code incorrect

Use build-in terminal also fine.

5 minutes ago, Vr2Io said:

Use build-in terminal also fine.

how?

Also some updates. It seems the issue exclusively occurs when trying to read from the NAS. If i copy files to the nas the hash matches (as long as i check the hash from within unraid - as if i don't i am reading the file from the nas and the issue occurs)

22 minutes ago, Vr2Io said:

Use build-in terminal also fine.

2 minutes ago, Thepn said:

how?

Depends on the type of theme, you will found a icon in top right coner or most left sdie

For example

5 minutes ago, Vr2Io said:

Depends on the type of theme, you will found a icon in top right coner or most left sdie

For example

yes but what should i do in the terminal?

12 minutes ago, Thepn said:

Also some updates. It seems the issue exclusively occurs when trying to read from the NAS. If i copy files to the nas the hash matches (as long as i check the hash from within unraid - as if i don't i am reading the file from the nas and the issue occurs)

I can't remember the behavior of copy file corrupt in that time, but it seems not always, it look like as below. So you need systematic to perform test

• hash remote file through network no problem

• copy file from harddisk to hardisk ( same controller ) no problem

• copy cross between media ( i.e. nvme <-> harddisk ) have problem, even with other nvme

3 minutes ago, Thepn said:

yes but what should i do in the terminal?

I means this is a quick way to make different test.

Edited March 20Mar 20 by Vr2Io

6 minutes ago, Vr2Io said:

I can't remember the behavior of copy file corrupt in that time, but it seems not always, it look like as below. So you need systematic to perform test

• hash remote file through network no problem

• copy file from harddisk to hardisk ( same controller ) no problem

• copy cross between media ( i.e. nvme <-> harddisk ) have problem, even with other nvme

I'm sorry you lost me. I don't understand what you want me to do

3 minutes ago, Thepn said:

I'm sorry you lost me. I don't understand what you want me to do

Could you use rsync to copy file from array to cache and found rsync indicate re-copy again ? Then use rsync to copy file hardisk-to-hardisk, and check any different.

Array -> cache and cache to array = Mismatch

Edited March 20Mar 20 by Vr2Io

41 minutes ago, Vr2Io said:

Could you use rsync to copy file from array to cache and found rsync indicate re-copy again ? Then use rsync to copy file hardisk-to-hardisk, and check any different.

Array -> cache and cache to array = Mismatch

Ok there's no issues when coppying internally on the nas as of now. the reason i thought so before was because i used the windows md5 utility to check hashes. But when i check internally on unraid they're fine. However when i trey to transfer from unraid to windows that's when they get corrupted and hashes break

OK, then this not same issue as my case.

If you have another PC ( Windows ), then I would suggest you verify does file transfer still corrupt.

Edited March 20Mar 20 by Vr2Io

2 minutes ago, Vr2Io said:

OK, then this not same issue as my case.

If you have another PC ( Windows ), then I would suggest you verify does file transfer still corrupt.

Yes i checked on another windows laptop and when i transfer from nas to that one the file still corrupts

This doesn't make much sense, if the files don't corrupt locally, if the LAN is not the problem, and they still get corrupt being read from 2 different PCs, there must still be something in common causing the issue.

10 hours ago, JorgeB said:

This doesn't make much sense, if the files don't corrupt locally, if the LAN is not the problem, and they still get corrupt being read from 2 different PCs, there must still be something in common causing the issue.

yes it's quite confusing. Chat gpt is saying something in the motherboard on the NAS maybe or the operating system itself.

Unsure if you read the updates but it looks like the issue exclusively exists when copying from the NAS to another system. Locally it's fine, writing to the nas is fine, but reading from it is where the corruption exists (except when copying to a flash drive, so only when reading from the nas on something network related). I have tested multiple different network setups (NICs, cables, ports) and those all result in the same errors.

Rsync also fails, recognizing the corruption

You need to replace some parts or try a different PC as the server, for example, until you find the issue.