https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=297337
--- Comment #5 from Warner Losh <[email protected]> --- Thanks for the tests and extra information. This tells me we're not chasing a startup issue, but there's something going on that's triggering the drives to become unresponsive. After a few seconds is likely quickly enough that we can start a dtrace script and then let the scrub start. This should tell us if the drive has stopped responding (my best theory) or if there's some weird bug in the storage stack we're hitting. I'm on vacation this week, but I'll create a good nvme dtrace script to see what layer is going amiss and to gather some data about changing workloads. ZFS is different between 15 and 16, and it may be scheduling more / different I/O now that's triggering the drives. The fact that it happens to both either means we're triggering the same firmware bug, or that we've started scheduling more than 30 seconds of I/O at a time or some other weird thing. -- You are receiving this mail because: You are the assignee for the bug.
