On Sun, 27 Nov 2022 19:06:35 -0600 Bill Barry <[email protected]> dijo:
>On Sun, Nov 27, 2022 at 7:02 PM Bill Barry <[email protected]> wrote: >> >> On Sun, Nov 27, 2022 at 6:55 PM John Jason Jordan <[email protected]> >> wrote: >> > >> > On Sun, 27 Nov 2022 16:07:09 -0600 >> > Bill Barry <[email protected]> dijo: >> > >> > >So reassembling is what was needed, but you don't need to force >> > >it. Now if you just had some way to automate that. Have you >> > >connected your external enclosure to the UPS so that it is >> > >protected from temporary power disruptions? >> > >> > Yes, everything is on my three massive APCs (connected together), >> > even the stereo. >> > >> > I keep thinking that the problem lurks somewhere in the crazy >> > setup: >> > >> > 4 Intel 8TB NVMe drives, mounted two each on two long PCIe cards, >> > inside a 4-bay PCI enclosure, in turn connected to a Thunderbolt 3 >> > port on a Lenovo dock, in turn connected to my Lenovo P73 laptop >> > via one of its TB3 ports. The whole setup is close to two years >> > old, during which time this is the sixth case where the RAID has >> > failed. >> > >> > When I got up this morning everything was running, so it couldn't >> > have been caused by a power glitch. There are lots of other >> > electrical devices around the house that tell me when PGE hasn't >> > been playing nice >> > - e.g., the clocks on the range and microwave will be flashing, the >> > bedroom television that runs all night will be off, among >> > others. >> >> If it's not a power problem then maybe cabling. How long are your >> Thunderbolt cables? >> >> Bill > >I see lots of posts out there claiming that their Thunderbolt >connection randomly disconnects. That could certainly cause your >problem. > >Bill Ah! Now that is news I was not aware of! I will do some searching and see what additional information I can come up with. And (crossing fingers) a solution. Regarding the cables, I have several TB3 cables, and after the first couple of failures I swapped them, with no change in the problem. Of course, there are lots more TB3 connection suspects - the three ports, for example. But I should have mentioned before that the two PCIe cards and their connections in the enclosure, and the connections of the four drives to the cards are the least likely suspects, because after each of the failures all four drives appear correctly and 'mdadm --query' and 'mdadm --detail' both list them as available without a hint of a problem.
