JorgeB

According to the diags there's no link on either 10GbE NIC: be2net 0000:01:00.0 eth0: Link is Down be2net 0000:01:00.1 eth2: Link is Down

That's from mcelog, not a problem for Unraid.

I don't have experience with Intel 10GbE, only Mellanox, but I seem to remember some issues with the out of tree driver with some releases, if you are on v6.10 try 6.9, or vice-versa, alternatively boot with for example an Ubuntu flash drive to confirm the NICs are linking.

Jun 8 18:58:56 Tower kernel: shfs[19677]: segfault at 14a9360000a0 ip 000014a94f319b49 sp 000014a94db3fbb0 error 4 in libc-2.33.so[14a94f2b1000+15e000] Shfs is crashing, every time this happens you'll lose the shares, open the syslog tail window and keep monitoring as you work, to see if you can find what's triggering the crash.

That means the cache floor was hit, that's the minimum free space set for the pool.

Please post the diagnostics.

Forgot to mention, for speed Unraid has pools, currently only btrfs is supported with raid0/1/10/5/6 profiles, soon ZFS will also be supported.

Seems about right to me. Also as already mentioned Unraid will always be slow for internal transfers, it's because the way parity works, you can find more info here, turbo write part is not relevant since it can't currently be used for internal transfers, which by itself is a shame since it would be faster.

Probably best to go with the actual disk then, unless you have a spare you could use, to check the actual disk contents stop the array, unassign disk4, start the array (leave disk4 unassigned), stop the array, then you can mount actual disk4 with the UD plugin to see if it mounts and contents look OK, note that the array can't be started or the disk won't mount without first changing its UUID.

Plugin might just not be reporting a correct/stable average, the actual speed is what really matters, at 100MB/s you can transfer around 300GB per hour, if you get anywhere close to that I would consider a great speed for an internal transfer.

No HDD can write at 500MB/s.

I would expect better performance with those, especially since it looks like you're transferring large files, though like mentioned it will never be close to disk speed, maybe 50/60MB/s at best, I believe the plugin uses rsync internally for the transfers, but never tested myself, you could try for example with midnight commander (mc on the console) to compare, but note that it would be much faster transferring to a pool or an unassigned device then transferring back to the array with turbo write enable, I always try to plan things to avoid internal transfers in the array.

Not looking great, but just start the array as is in normal mode and check the contents of the emulated disk4, look for a lost+found folder and see if there are many things there.

Unraid is always pretty slow when transferring data inside the array, because of how parity works, and currently it's not possible to use turbo write for internal transfers, though I would still expect it to be faster than that unless you're using SMR drives, what drive model is parity and the disk you're writing to?

Enable the syslog server and post that after a crash, hopefully some clues there.

It doesn't, this is also not a good sign: And the total uncorrected errors as well.

With vt-d disable it should be perfectly safe, and you can't enable vt-d with that CPU even if you want to, also note the is is a kernel bug, not an Unraid issue, same exact problem was reported for Ubuntu with ZFS for example.

There have been some reports of permission changes with docker and AD after updating, so you're probably on the right track.

Never format, that's not part of a rebuild. First run xfs_repair and depending on the result decide if you rebuild on top or not, i.e., if the emulated disk is fixable and contents look correct and there's nothing or not much in a lost+found folder.

Did you re-start the array in normal mode? If yes post new diags.

Depending on how simple or complicated your current install it could be mostly OK to update directly to latest, especially if you use Unraid mainly as a NAS, but some changes might require some config changes, like with VMs for example. I would also suggest reading the release notes from the major releases, v6.4.0, v6.5.0, v6.6.0, etc, those will have the most important things to consider.

Yes, and only if it asks for it use the -L option.

Are you sure it was ever formatted? Please post the diagnostics, that will also give an idea if it was previously formatted ort not.

Yes, only two devices are in the pool, suggest doing: -stop array -unassign nvme2n1 -start array -stop array -wipe that device with blkdiscard -f /dev/nvme2n1 -assign it back to the pool -start array If it still doesn't work post new diags.

Please post the diagnostics.