Hello,

I use my unraid server as a backup for large files, however since a few days the transfer speeds are non existent. Usually, depening on the machine I'm moving the files over are betwwen 100 Mb/s to 130 Mb/s which is the expected speed usually. However yesterday I've started a new large transfer, for a few minutes it reached between 80 - 100 Mb/s however running it longer, it wen't down to a max of 5 Mb/s. What I've also observer, however I hadn't paid any attention to it before. Is that only the parity drive and one disk receives writes, all other drives only reads, no writes on them.

I've attached the diagnostics ZIPs as I don't know what the problem could be.
Hope someone can point me in the right direction.

unraid-diagnostics-20250705-1237.zip

7 minutes ago, exxis said:

. Is that only the parity drive and one disk receives writes, all other drives only reads, no writes on them.

That is normal for standard write mode, try turbo write, if still slow, transfer to the pool to see if it's any different.

2 minutes ago, JorgeB said:

That is normal for standard write mode, try turbo write, if still slow, transfer to the pool to see if it's any different.

if you referr to "reconstruct write" then that's always active, as I've wrote, before the transfer speeds where normal now they are non existent. Also started a parity check right now (without anything transfering), it would take a few hundret days to complete it as the speed here is around 800 Kb/s, before a parity check would perform with around 218 MB/s

4 minutes ago, exxis said:

Also started a parity check right now (without anything transfering), it would take a few hundret days to complete it as the speed here is around 800 Kb/s, before a parity check would perform with around 218 MB/s

Repost diags after that's been running for a while.

5 minutes ago, exxis said:

if you referr to "reconstruct write" then that's always active

I'd try disabling that and trying to read/write something from each individual disk to see if one is slow. Reconstruct writes (and parity check) are constrained to the speed of the slowest drive, if one is failing and having issues reading its data that can slow everything down in those cases.

57 minutes ago, Kilrah said:

Repost diags after that's been running for a while.

I'd try disabling that and trying to read/write something from each individual disk to see if one is slow. Reconstruct writes (and parity check) are constrained to the speed of the slowest drive, if one is failing and having issues reading its data that can slow everything down in those cases.

Here would be a new diags after running parity check for some time.

How would I check each disk, with krusader and copy files between them or would there be a tool/app in unraid to check each disks speed?

Also what was really weird, is that, yesterday after a reboot, which took a really long time, the machine renamed itself, usually I just have it called unraid, but after the reboot it renamed itself back to Tower, could that be related or would that be a different issue.

unraid-diagnostics-20250705-1322.zip

So performed a disk speed test, the slower ones are all WD Red, however even at the slowest rate which would be 80 MB/s it's still faster then the transfer of the maximum of 5 MB/s I'm currently getting, don't know if it's still an issue, however never had these slow transfer speeds before.
However what looks at least weird to me, from the second screenshot that If I transfer data to unraid, it writes to the slowest drive only disk 9 (sdc) but still slower as the max benmarked speed.

Edited July 5, 20251 yr by exxis
added screenshot

Parity check was going on during the last diags, once it finishes, transfer a large file from a pool/LAN to the array and post new ones during the transfer.

7 hours ago, JorgeB said:

Parity check was going on during the last diags, once it finishes, transfer a large file from a pool/LAN to the array and post new ones during the transfer.

On 7/6/2025 at 10:03 AM, JorgeB said:

Parity check was going on during the last diags, once it finishes, transfer a large file from a pool/LAN to the array and post new ones during the transfer.

on the last one yes, as Killrah requested that I post the diags while the parity check is running.
Will post new transfer speeds, I've replaced the slowest drive sdc, as something must be up with that drive if the average speed is that low during the disk speed test.

On 7/6/2025 at 10:03 AM, JorgeB said:

Parity check was going on during the last diags, once it finishes, transfer a large file from a pool/LAN to the array and post new ones during the transfer.

on the last one yes, as Killrah requested that I post the diags while the parity check is running.
Will post new transfer speeds, I've replaced the slowest drive sdc, as something must be up with that drive if the average speed is that low during the disk speed test.

WDC WD50NPZZ-00A is a 5TB 2.5" 5400RPM SMR drive with only 8MB of cache.

It is just a very, very slow disk under extended loads and when closer to full it has to perform SMR writes, which can slow the the KB/s range.

Go through each disk and check if they are SMR, they should be your targets for future upgrades.

14 hours ago, Faceman said:

WDC WD50NPZZ-00A is a 5TB 2.5" 5400RPM SMR drive with only 8MB of cache.

It is just a very, very slow disk under extended loads and when closer to full it has to perform SMR writes, which can slow the the KB/s range.

Go through each disk and check if they are SMR, they should be your targets for future upgrades.

Yep that's the one I've already replaced because it was in general the slowest drive, currently rebuilding after that I do another speed test.

So in general as unraid when data is copied to the array picks one disk, since all the old drive I still use were already full it only copied to the new drives I added, so the array in general was already slow but because it copied to the new freshly added drive I never noticed it?

Also drives I will replace next are the WD60EDAZ and the WD40EZRX as they are also really old ones.

Edited July 8, 2025Jul 8 by exxis

I believe the WD60EDAZ is also SMR, so it would have been a bottleneck to write to in some cases too, nowhere near as slow as the 5TB you removed but still measurably slower than the higher end drives you have in other slots.

The next slowest are also obvious on the disk speed chart you posted, so target those as the next upgrades/removals.

so replaced three slower drives, now transferring files to the array. Still really slow, don't know what the issue could be. New diags during file transer.

unraid-diagnostics-20250714-1230.zip

Diags show writes at around 80MB/s, which is reasonable, but there could be a controller bottleneck, for example. Did you run the diskspeed controller tests as well?