TB is hot plug interface. As such, I'd sort of expect that HW and SW assumes that it can safely traverse power states and that connection resets may be OK.
If that is what is happening, it is of course pretty nasty thing to do to md arrays. Native higher level file systems btrfs/zfs may fare better, but I would be skeptical and expect similar problems. JBOD should be more resilient. In any case system logs, particularly power state changes should confirm what is actually happening. Hope that helps, Tomas On Mon, Nov 28, 2022, 04:02 Russell Senior <[email protected]> wrote: > Another good starting point is: cat /proc/mdstat > > On Sun, Nov 27, 2022 at 11:03 PM John Jason Jordan <[email protected]> wrote: > > > On Sun, 27 Nov 2022 22:39:48 -0800 > > Paul Goins <[email protected]> dijo: > > > > >When this occurs, have you tried "dmesg -T" or checked > > >/var/log/kern.log or /var/log/syslog for suspicious log entries around > > >the time of the error? Might be worth taking a peek as well. > > > > Your suspicion was correct - I had not tried either of those options, > > and because I was unaware of them. I have added your suggestions to my > > file for mdadm solutions. Next time this happens I will see what they > > produce. > > >
