Input/Output error

[MorgothCreator]

Quote

 

I know that here are some topics with the same issue, but this issue need to never happen on a server system.

So, the story is that some time I want to move data around from a disk to another, to compress some data, to decompress it etc.

But on large files some time is getting me Input/Output read error and I am tired of this error on a paid OS.

The btrfs check is with no errors, and this Input/Output error is getting me on both , on array and on unassigned devices.

Mention is that I virtualized a Synology DSM with pass through and that system is working without any errors, so I know that is not a device fault, but something with FS implementation.

Another mention is that the machine is on a semiprofessional UPS, so, can not be a power fault.

 

Last time I try to convert a 70GB image and this is the result:

 

qemu-img convert -f vmdk -O qcow2 SomeFile-disk1.vmdk SomeFile.qcow2 
qemu-img: error while reading sector 7978112: Invalid argument

 

And the conversion can not pass this sector.

 

The same error some time give me tar application on large files.

 

The mv and cp never fail or report IO issues but I see sometime artifacts on videos, so the issue is present even in mv and cp, so in one sentence, about all my large files are corrupted to put it like that.

 

I see it on transmission when is reporting disk error and I need to put it to check the file multiple times, and transmission use a virtualized docker disk.

 

When I recreate the source file and try the same above command, will give error on another sector or some time no error.

 

One time an entire disk was corrupted when I powered down a Windows7 VM, and that disk was unrecoverable, needed format.

 

The system is: Thredripper 1950X liquid cooled, with 64GB matched RAM kit on a ASROCK Taichi X399 motherboard, Power source 1200W, UPS APC Smart-UPS C 1000VA LCD the OS is 1.7.2.

 

All rotated disks are from WD, The SSD's are from Samsung and are PRO or Evo, a single SSD is from Intel and is a 120GB and never give me a error because, I suppose that I never transferred large files on it.

 

What is happening???????????

Edited October 20, 2019 by MorgothCreator

[John_M]

It sounds like a hardware problem - bad RAM or disks or power supply - but without diagnostics it's just a guess.

[MorgothCreator]

8 minutes ago, John_M said:

It sounds like a hardware problem - bad RAM or disks or power supply - but without diagnostics it's just a guess.

I see that the problem is on writing to the disk, not on reading, because the error on the file is happening at the same sector when I read the same write file, and the sector change if I rewrite the file and try to read it again.

I mentioned that I have a VM with 4 pass-through disks that never had problems, so all components are working properly.

The system does not report any other problem, does not freeze, or have strange behavior, being up and running for several weeks in a road.

The problem is only when working with big files, like image disks and large videos.

On some rare occasions when VM virtual disk is edited by the VM machine the entire disk is corrupted ( happened one time ).

When the Windows7 VM has corrupted the disk, when has been shut down by me, the Unraid web page stop responding because has stuck on waiting the disk to respond, but the other VM's and docker run normally ( working on other disks ), when I open the web terminal, has opened normally and I can send commands, the disk was mounted and was navigable, only after Unraid (unclean off course, because after several hours the system refuse to restart ) restart the corruption of the disk was observable.

I think that is something with FS arbitration between multiple app's writting on the same disk or something like that.

[MorgothCreator]

The problem seems to be from tar application.

I decompress the ova archive using tar from Synology DSM on the VM, after that, I copied the vmdk image to the Unraid machine and convert it to qcow2 with qemu-img on the Unraid, and the conversion was done successfully.

So, something is wrong with tar application working with large files on the Unraid OS.