I have a two node TrueNAS CORE 13.0u6.1 setup with shared DAS and three 90 bay shelves with four multipath connections each.
As drives have failed and been replaced, I have been careful to use the multipath disk reference as the replacement device for rebuilding the respective vdevs to maintain the multipathing.
As viewed from the GUI, this is working as intended. Viewing the Pool status, all vdev devices are listed with the multipath disk references.
However, today I was checking on a resilver process from the CLI because according to the information in the GUI it was going to take ~16 days to finish. I don’t know what was going on there, but the resilver information from ‘zpool status -v’ matched what I was expecting to see from the GUI (It finished much quicker than the estimated 16 days). I also noticed however that a handful of the disks in the pool are listed by their gptid instead of their multipath references. This includes the just replaced disk that I was checking on the resilvering process of.
So, it appears that although the GUI has been instructed to use the multipath references for replacing drives, and displaying that this is what has been being done, it does not match what is actually happening. I believe I can export/re-import the pool to fix this without having to replace/resilver the drives again (at the next opportunity to take the pool offline). But, how can I avoid this situation in the future? I can just perform these tasks from the CLI in the future, but I know some operations are not supported from the CLI.
Rob