What is the issue you are having? Do you have a USB attached boot pool with mirroring? If you trying to use multiple MYBook drives attached by USB, that will still be considered ‘unsupported’. You can look into shucking the drives from the cases and attaching them directly using SATA.
NAS-139947 is showing the problem with USB attached boot-pool in mirror as fixed in 25.10.3 and the 26.0.0 Beta1. That specific setup was causing issues with SATA or SAS pool creation at the time.
I had a different make/model USB drive attached and it was causing the problem. I removed it and its working. I didn’t realize I couldn’t have multiple attached. I guess that’s the “bug” or whatever?
I don’t know if some code was changed in the 25.10 series but that is when we started to see more issues with ‘duplicate serial numbers’ If the Bug report showed USB, Thunderbolt, or a few known, troublesome SAS shelfs, they were closed as ‘unsupported’.
You should be okay if you stick with SATA or SAS.
I’ve repeated some of your work and additionally dug into the kernel source. IMHO a workaround is within reach.
Hi @richardm,
Thanks for doing the work, unfortunately after isolating the related code as a standalone piece of code, life happened and I was unable to find the time required to continue working on it.
I also believe that the problem can be solved fairly easily.
I looked into your comment, but the problem is not only how kernel looks for disks’ serial numbers, but how the enclosures are exposing these serials and the topology of the enclosure itself. So, from my experience, the Kernel has no glaring problems, but the code for looking disks’ serial numbers needs some more work to try other paths and possibilities, somewhat in a smarter way.
That core library I isolated needs some love and informed development IMHO.
Hiya, after trying the latest beta release and still having the same issue, I decided to create a patch that users can apply themselves using a script.
It’s been written with AI, but the patch is small, targeted, and borrows from the already working logic used by the UI to get the serials.
It requires enabling developer mode to apply unfortunately. I have included some mitigations to re-enable read only on the root dataset and switch apt and dpkg pack to not being executable, but obviously this is not the same as never having enabled dev mode in the first place.
Tested on my system. All working.
It’s on github at 1gr1g/truenas-usb-serial-patch
I think the solution here is for ixsystems to use some of their can-do to leverage their own page 80 detection logic instead of letting Jesus^H^H^H^H^HLinus take the wheel. There’s clearly a gap when I can use three different tools to pull a GD serial out of my disks but this middleware cannot.
