Hi,

I've bee a long term user of unraid and never had any real problems, I have two arrays one for long term storage which is powered on periodically and I sync from a smaller ssd based one. The larger array has several spinning disks and an older CPU so consumes a huge amount of power. Its working fine, however reasonably recently I've been having problems with the smaller array. It has 6 ssds in it and has performed flawlessly for several years, however recently a disk was dropped by the array, which I suspect was a controller issue as opposed to a disk issue as the disk passed its smart tests without issue, however I replaced it anyway and have been using the disk in an external enclosure to move data around and its fine.

When I lost the disk I noticed that things were not right, performance would drop through the floor on writes (and reads), I noticed some errors in the syslog that looked like they might indicate contention on the controller as the motherboard only had 4 slots so I used a cheapo ASM116x based controller for some of the ssds. Anyway I moved things around tried different controllers and managed to clear that error but performance was still shocking, plus the wait for i/o seems high, can be as high as 25%.

I did some googling and reading through the forums on here and most advice seemed to be to move to a LSI controller so I picked one up and installed it, moved all the disc to it but it was no better, I even disabled the onboard sata ports incase that was part of the problem still no joy and one of the suggestions in here was to do with pci lanes and limited etc. I also read that shfs can be a problem so have created some disk shares and that has made no difference. Even transferring from disk to disk either manually on the command line or using unbalanced takes an age. With transfers averaging around 1.5 Mb/s

When I replaced the controller I did a parity check and obviously its read only but that completed at 465 Mb/s according to the stats, what is curious is that cpu usage as a whole is maybe only 25% however I suspect thats because its just waiting for i/o. I use a cache drive and even transferring from that to the main shares can take an age.

What is curious though is that the performance is "lumpy", sometimes a file copy will complete at 80+ Mb/s, other times it crawls and I've seen disk reads and writes as low as 300 Kb/s (sustained).

Initially I did wonder if one of the SSDs is aging but they all pass their smart tests (if I am reading them correctly) and there doesn't seem to be any pattern, ie its only when one particular disk that is involved but that doesn't appear to be the case. The only common disk would be the parity disk but again watching the array this morning I could see at one point while the mover was running it was doing over 50MB/s

I've started to move things around in the array to see if that helps, but thats made no difference. The problem started probably about a month ago and showed up as stalled video playback while streaming even small (<500MB) files from the array or the whole GUI (and docker containers) becoming unresponsive for multiple minutes with the CPU pinned at 100%, that issue seems to have gone away but I'm left with this weirdness. Even when the array is showing as practically idle streaming will stall (there is no trans-coding etc its just a straight read into local osmc) and the cpu on the array is showing as less than 25%

I'll attach the diagnostics dump as soon as the forum software will allow me.

box-diagnostics-20250716-1118.zip

Edited July 16, 2025Jul 16 by Jollyp
Added diags

I would suspect an SSD problem, most likely with those no brand Chinese ones, though also note that SSDs on the array cannot be trimmed, so write performance degradation can occur.