Alex43

Turns out the issues appeared to be related to AMD SATA support (adjusted power settings in mothberboard BIOS, changed cables and in the case of the Samsung SSD updated the firmware and played with NCQ settings and it seemed to resolve). I ended up not returning the drive and it is ok so far. Will mark as resolved.

I was able to locate a newer firmware for my Samsung drive and after a bit of fiddling finally got the USB to boot and flashed it. It seems to be stable now (no errors) but use of the drive today suggests the throughput is lower (like 60-65 MB/s for transfers from a machine that also has an SSD) so that requires further testing.

Looks like I'm running in to the Samsung + AMD chipset problem with CRC errors that many others have reported. Running a full Smart diagnostic and also turned off NCQ. Will report findings.

It would be nice to have human readable timestamps in the dmesg output when troubleshootiung. The perl script hdmesg.pl provides this: https://github.com/loki-dv/tools/blob/master/hdmesg.pl EDIT: Nevermind, dmesg -T is also supported already

I shut down the array and rebooted the system and the cache appears to be back (albeit things reorederd as the UBS drive was still plugged in). Will try a large copy now.

Odd, I don't have a /mnt/cache just: root@Alex:/mnt# ls disk1/ disk2/ disks/ remotes/ rootshare/ user/ user0/ root@Alex:/mnt# Uh oh, when I look in the UI it says the cache isn't mountable. "Unmountable: Unsupported partition layout " Not sure what happened here, it was working... Weirdly it is green in the UI but then on the far right says it isn't mountable. Shouldn't it be red if that is the case? It shows up under the devices in the Tools tab: 0:0:0:0]disk USB Flash Disk 1100 /dev/sda 16.0GB [1:0:0:0]disk ATA ST4000VN008-2DR1 SC60 /dev/sdb 4.00TB [2:0:0:0]disk ATA ST4000VN008-2DR1 SC60 /dev/sdc 4.00TB [5:0:0:0]cd/dvd ATAPI DVD A DH16A1L KH1A /dev/sr0 1.07GB [6:0:0:0]disk ATA ST4000VN008-2DR1 SC60 /dev/sdd 4.00TB [9:0:0:0]disk ATA Samsung SSD 840 BB6Q /dev/sde 250GB [11:0:0:0]disk WD My Book 1230 1050 /dev/sdf 3.00TB /dev/sde is the device. I can see the kernel barking about errors in dmesg: [158710.373776] sd 9:0:0:0: [sde] tag#4 UNKNOWN(0x2003) Result: hostbyte=0x04 driverbyte=DRIVER_OK cmd_age=0s [158710.373787] sd 9:0:0:0: [sde] tag#4 CDB: opcode=0x28 28 00 1d 1c 58 80 00 00 08 00 [158710.373791] I/O error, dev sde, sector 488396928 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 0 [158710.373827] sd 9:0:0:0: [sde] tag#5 UNKNOWN(0x2003) Result: hostbyte=0x04 driverbyte=DRIVER_OK cmd_age=0s [158710.373832] sd 9:0:0:0: [sde] tag#5 CDB: opcode=0x28 28 00 1d 1c 58 80 00 00 08 00 [158710.373834] I/O error, dev sde, sector 488396928 op 0x0:(READ) flags 0x0 phys_seg 1 prio class 0 [158710.373839] Buffer I/O error on dev sde1, logical block 61049360, async page read [158714.769852] program smartctl is using a deprecated SCSI ioctl, please convert it to SG_IO [158714.769869] program smartctl is using a deprecated SCSI ioctl, please convert it to SG_IO [158714.769877] program smartctl is using a deprecated SCSI ioctl, please convert it to SG_IO

Not sure where to make suggestions but turning on "safe delete" and "safe overwrite" in the default midnight commander config might be worth considering. That exposes the option for the various overwrite "yes/no/if different size/if zero size" etc.

Not sure where this belongs on the forum so feel free to tow it somewhere else. I have a large number of devices backed up on to an external USB 3.0 WD MyBook and I was having problems accessing them in Windows. Not sure why but the source directory would keep being reported as "missing" or "disappearing" or something as soon as I started a copy from Windows to the Unraid drive. Googling around many others reported the same issue with no clear resolution. I decided to plug the drive directly in to the Unraid system and I used the "Unassigned Devices" free plug-in to mount the drive from the UI and then I spawned a shell and I am using Midnight Commander (the command 'mc') to copy the files. Takes me back to the early DOS days. I had no idea how powerful it was until I tripped across this amazing reference page that I think folks here will appreciate: https://klimer.eu/2015/05/01/use-midnight-commander-like-a-pro/ One thing that isn't clear to me is what path I should use to copy the contents to the cache drive rather than directly to the disk? If I copy to the disk array then I get horrible throughput since it needs to do the parity work on each bit at the time. What is the symbolic path on the cache to use when working at the command line?

I successfully (at least the script completed successfully, haven't tried it yet, will remove this note when I do) dumped the BIOS from an Nvidia 670GTX 2GB Gigabyte Windforce card. I will attach the ROM and update with more deails assuming it works. I had previously bound the GPU and Audio together to the VFIO with ACS (when testing, prior to script) but that was causing the script to fail. in order for the script to work I had to remove that binding and reboot. Then the script ran fine and produced the attached ROM file. I ended up not using the VBIOS (so far) because I was able to enable ACS in my motherboard BIOS of my AM350 Pro4 Ryzen 2400G system (different than the Unraid ACS config referenced in the first paragraph) and that allowed the NVidia card and sound device to move in to their own IOMMU group and then be passed directly through to the VM. I am testing with Linux Mint 21 using the OVMF bios type and no VBIOS set. 670gtx-vbios-dump.rom

Not to thread hijack, but is it possible in this scenario to replace it with a drive of slightly different size? I have a 240GB and a 250GB SSD and I was thinking of putting them in as RAID1 cache drives. Will that work or is some extra adjustment required?

Recognizing that not all data is of the same value, I'd like to do something extra to protect the archive of family photos. Losing those would really suck. I have a friend in another location who wishes to do the same thing for the same reason. Is there a mechanism in Unraid to replicate certain trees over a VPN or other method to a remote system?

What is your motherboard and CPU? If it is AMD/Ryzen, I had a panic as well. There are details in the FAQ about some power settings in the BIOS that you may have that seem to have stabilized it for me (although I'm only a few days in, to be fair)

Is there a mechanism to create a second array with the 16 TB as parity and another 16TB as the only disk (for now) then move all the files over then add the 4TB drivers under the new array?

I have a new Unraid build, my first one. I would like to do a VM with a GPU passthru of my NVidia 670 GTX. The board is an ASRock AM350 Pro 4 with a Ryzen 5 2400G Processor and 16 MB of RAM. The NAS side of things seems to be set up and working. I have tried both x16 slots and both result in the Nvidia card appearing in IOMMU group 0, so I don't seem to be able to isolate it. Am I going to be able to pass it through given this situation? My IOMMU groups are:

I've made the change to the Idle Power option in the BIOS (can't recall the setting, basically don't have it reduce at all) and I've turned off any suggestion of overclocking in the bios (the AUTO settings) per the Unraid FAQ in the post above this one. I also changed all the SATA cables and I made sure I have no drive on a shared SATA port that is shared automatically with an unused M2 slot. One of these changes appears to have resolved the issue as I am almost all the way through a rebuild of the drive and I haven't had a single CRC error.

Am discussing in the Discord with someone as well. Currently running a Memtest (had to make sure I booted in CSM compatibility mode, not UEFI, for Memtest to load) Have replaced the cable completely to that drive. If Memtest is clean, will try another port on the board. There are two available ports on an ASMedia chip, whereas the three HDD and SSD are currentnly on the SATA ports directly off the CPU. Was hoping to avoid having to transit to the other bus.

I guess I need to get used to a full parity check being required on each power loss/crash event. Hopefully they are infrequent, of course. Will the data that was in the process of being moved when the kernel crashed still be on the cache drive or is that a situation where something could have been corrupted in process?

New build, Ryzen 5 2400G CPU, 8GB RAM, GigE NIC, 3 x 4TB Ironwolf drives and 1 x Samsung SSD 256GB Cache drive. Running trial of 6.11.1. Did the initial format and parity check last night. Came up clean. Temps around 32C the whole time. All good. Created a few shares and copied about 10GB of data to the cache as a test. Initiated a 'move' from the UI and lost access to the webgui. Thought maybe it was just temporary IO load. After a few minutes realized something was wrong. Turned on the console screen and could see a stack trace from a kernel fault. What would have caused that? What should I do?

Just built my first Unraid system. System has 3 x 4TB drives and a 256MB Cache pool drive. Right now the parity disk is being built, looks like that will take all night. If I have one of my data drives fail, is it possible to replace it with a smaller drive (say a 2TB drive) if it is large enough to support all the data? I guess I'm asking how the data is distributed across multiple disks in an array and if a failed drive must be replaced with an identical drive? I didn't buy spares but I'm starting to think maybe I need to?