UFS2 SU+J could not recover after power-off sgain (was: One process which would not die force me to power-cycle server and ALL UFS SUJ FSes are completely broken after that AGAIN!)

[Lev Serebryakov]

Hello, Freebsd-fs.
You wrote 12 апреля 2014 г., 17:09:53:

LS>  All UFS2 filesystems can not be recovered with using of automated fsck, due
LS> to journal/softupdate inconsistencies. I need to run "fsck -f" TWICE for
LS> each of them (as first run ask to re-run fsck).
 "shutdown -h" reboots system, UPS switch power off after that (with delay),
2 out of 5 FSes could not be checked with journal automatically. Manual full
"fsck" run didn't find any serious problems, only one or two unlinked files
(recovered to lost+found) free block bitmaps!

 WHY?! How could I trust to UFS2 now?!

 Both filesystems have same scenario:

/dev/ufs/tmp: Journal file sequence mismatch 233263 != 231707
/dev/ufs/tmp: UNEXPECTED SU+J INCONSISTENCY
/dev/ufs/tmp: INTERNAL ERROR: GOT TO reply()
/dev/ufs/tmp: UNEXPECTED SOFT UPDATE INCONSISTENCY. RUN fsck MANUALLY.

/dev/ufs/usr: Journal file sequence mismatch 287936 != 282572
/dev/ufs/usr: UNEXPECTED SU+J INCONSISTENCY
/dev/ufs/usr: INTERNAL ERROR: GOT TO reply()
/dev/ufs/usr: UNEXPECTED SOFT UPDATE INCONSISTENCY. RUN fsck MANUALLY.

 Again: these FSes were checked with full fsck two days ago. They reside at
 SATA HDD without any non-standard or complex geom modules (only
 geonm_part), and HDD is attahed to chipset SATA port, there is no any RAID
 controllers or things like that.

 EVERY non-clean reboot of server leads to "RUN fsck MANUALLY".

-- 
// Black Lion AKA Lev Serebryakov <lev at FreeBSD.org>