Date: Wed, 29 Jul 2009 01:36:52 -0700 From: "Edward Sanford Sutton, III" <mirror176@cox.net> To: freebsd-questions@freebsd.org Subject: fsck and dump freeze freebsd. any ideas? Message-ID: <200907290136.52614.mirror176@cox.net>
next in thread | raw e-mail | index | archive | help
The smartctl tests on the three drives came back saying they were checked 100% without error; bad sectors have always caused an error and an aborted scan at the point of trouble for me in the past. I 'thought' I recalled seeing a panic along the lines of ffs_free or something related to free blocks or something at one point; maybe that was an error with fsck on the live system when manually ran. I will try to reproduce. Otherwise I do not know that fsck crashes; the entire system locks. I will see if I can reproduce the problem as a panic. There were some errors fsck seems to think it cleaned. A couple still exist indicating wrong counts about '12 should be 4' or '4 should be 0' type stuff. One error had a ridiculously large seven digit or so number that fsck said should be reduced. I had made a many terabyte file with a dd write to the end of a file of the largest size it would let me create and have since deleted the file; maybe that was what the reference was to. After fixing some errors, the fsck still causes a freeze and at what seems to be about the same point. The filesystem was unmounted for the dump and fsck has been attempted both mounted and unmounted. The filesystem has (ufs, NFS exported, local, soft-updates) reported for features by mount under normal operation. The freeze with dump did occur during a snapshot creation, which I found I cannot kill with a -9 (kill attempted minutes before the crash which still would not stop the creation). The powerdown question is more for how to handle an unsafe powerdown/crash on an active system. If I need to read from a partition, would read only leave it completely clean? Is there a way to operate on a file system which is treated as more of a ramdisk of changes and keeps the real partition unmodified (giving results like Faronics deepfreeze software or qemu disks in snapshot mode)? Would a zfs mirror configuration handle the unexpected crash/powerdown? Would it just report and fix the corruption, mention what files/structures weror impacted, offer restore of that data from a recent snapshot, or just say it is time to restore from a backup? Thanks again for the feedback, Ed Sutton
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200907290136.52614.mirror176>