Bug 198228

Summary:	zfs loader failed to load kernel from degraded mirror (disk with SMART errors)
Product:	Base System	Reporter:	Bjoern A. Zeeb <bz>
Component:	kern	Assignee:	Bugmeister <bugmeister>
Status:	Closed Overcome By Events
Severity:	Affects Only Me
Priority:	---
Version:	10.1-RELEASE
Hardware:	Any
OS:	Any

Description Bjoern A. Zeeb freebsd_committer

2015-03-03 15:20:09 UTC

After a power glitch one of the zfsroot machines in the lab did not come up anymore.

The machine in questions had
1 SSD with freebsd-boot on it + l2arc cache
2 HDD in a ZFS mirror setup.

The first of the two HDDs (according to BIOS sort order at least), had SMART event errors (as notified by BIOS upon boot).

While loading the loader still worked fine, loading the kernel was not possible anymore and the twiddle went very slowly on the text segment.  We aborted after about 8 hours.

Removing the disk with SMART error notification and the system booted back up just fine with a degraded mirror.


Further diagnosis is no longer possible as the machine in question was needed and the disk is going to RMA.  However ZFS loader should be able to deal with this situation (better).  Someone with enough loader + zfs knowledge could go and have a look and either report read errors (checksum mismatches, ...) or if detecting a mirror trying the other source instead or providing a possibly to more easily select which source to use (abort, print message, let user set the variable and retry).


PS: There is also the point that ^c does not work during the time we are trying to load entities so breaking into the loader shell was not possible, which is somewhat unhelpful.

Comment 1 Mark Linimon freebsd_committer

2025-01-19 05:37:28 UTC

^Triage: I'm sorry that this PR did not get addressed in a timely fashion.

By now, the version that it was created against is long out of support.
As well, many newer versions of ZFS have been imported.

Please re-open if it is still a problem on a supported version.