[6.12.6] Does not start as USB Devices apparently switched ids?

Quote

 

A drive formerly identified as 
`Samsung_SSD_870_QVO_4TB_S5STNF0WA03436P -`

 

is now for some reason identified as
`Samsung_SSD_870_QVO_0000000000000002-0:0 `

 

 

USB bridges can be a problem with ID's, one of the reasons they are recommend for array or pool devices, though this one is kind of strange, since the ID is completely changing, can you post a couple of diags showing the same device with the correct and wrong ID?

7 hours ago, JorgeB said:

 can you post a couple of diags showing the same device with the correct and wrong ID?

 

There you go. Let me know if i could provide any other info.

 

tower-diagnostics-20240112-2054-id-is-weird.zip tower-diagnostics-20240112-2056-id-correct.zip

Edited January 12 by doc-unraid

The main difference I see between the diags, is that when it doesn't work, the USB enclosure is not being detected during the initial boot as it should, like right after the flash drive, it's only being detected in the end, after the plugins get loaded, and possibly you are disconnecting and reconnecting it?

 

In any case, can't really offer a solution, you can try a different USB port/controller if available, but ideally I would recommend not using USB for array or pool devices if possible, they can cause various issues, including ID issues, devices dropping, spin down and SMART issues, etc.

52 minutes ago, JorgeB said:

it's only being detected in the end, after the plugins get loaded, and possibly you are disconnecting and reconnecting it?

 

Nope, i tried different Cables, host ports etc. Same behaviour whatsoever. Between the two diags i did not touch any hardware at all, just did an init 6.

 

The thing i do not get is: i used a simple ubuntu with ZFS before and that worked every time. Even when booting unraid, and some IDs are messed up, zpool import works without any complains every time. (Sometimes the pool contains the disks identified by either their kernel provided IDs, some just via their scsi devices (/dev/sde)), so that i assume that zfs just scans all the available disks and then decides which disks are used as vdevs for which pool, based on their label data (or smartctl data). This is why i am wondering, if unraid would be able to circumvent the problem in a similar way.

 

Anyway, thanks for investigating.
As i am using a NUC for my NAS Host, avoiding USB is not an option for me.

27 minutes ago, doc-unraid said:

Even when booting unraid, and some IDs are messed up, zpool import works without any complains every time.

Yes, because they don't care about the device IDs, Unraid does, it uses them to keep track of all array/pool assignments, devices that don't have a consistent ID, or in some cases have duplicate IDs for different devices, should not be used with Unraid.

I understand. I was just wondering (given those IDs tend to be less stable with some hardware setup than the smartctl provided info is), if it would be an improvement for unraid (optionally) to use identifiers generated from the smartctl data instead of the kernel provided IDs to keep track of the disk membership.

As i understand this would expand the usable hardware configs.

 

Don't know if it was possible (no idea of unraid internals) or if it was worth the effort, taking into account that you hinted at other possible problems with USB devices.

No hard feelings - then unraid just isn't usable for me and i'll go back to a simple linux+zfs setup and sell my license

 

Thanks @JorgeB

Just wondering: Would this be something i could potentially work around via udev rules?
https://reactivated.net/writing_udev_rules.html#why

Not sure, that's beyond my knowledge.