solaris7: ufs logging goes bezerk

solaris7: ufs logging goes bezerk

Post by Aleksandar Milivojevi » Wed, 02 Jun 1999 04:00:00



I used to have a directory named q on partition with ufs logging
turned on.  Just a while ago, I typed ls -l in it's parent dir and got
something like this:

pc-alex$ ls -la
total 17953794
drwxr-xr-x   3 alex     staff        512 Jun  1 11:18 .
drwxrwxrwt   8 root     root         512 Apr 21 15:16 ..
?-wxrwxrw- 254 254      254      71495735022846206 Jul 12  1970 q
-rw-r--r--   1 alex     staff    669044736 May 14 15:09 redhat6.iso

"q" used to be a directory, and now it is "something"?!?!?

There was no disk crash, no system crash or anything else that could
explain this (well, ufs logging was supposed to protect me from system
crashes, right?).  After I saw this, I unmounted partition, and did
manual "fsck -y".  fsck reported a lot of "PARTIALLY ALLOCATED INODE
I=nnnn" (nnnn is some number) and copule of really wiered things like:

DIRECTORY CORRUPTED  I=29440  OWNER=hniksic MODE=40755
SIZE=512 MTIME=Apr  2 11:54 1999
DIR=/hniksic

SALVAGE?  yes

MISSING '.'  I=29440  OWNER=hniksic MODE=40755
SIZE=512 MTIME=Apr  2 11:54 1999
DIR=/hniksic

FIX?  yes

MISSING '..'  I=29440  OWNER=hniksic MODE=40755
SIZE=512 MTIME=Apr  2 11:54 1999
DIR=/hniksic

FIX?  yes

For my nice "q" directory, I got:

UNALLOCATED  I=35328  OWNER=root MODE=0
SIZE=0 MTIME=Jan  1 01:00 1970
NAME=/alex/q

REMOVE?  yes

This was all on Solaris 7 x86 system, patched to MU2.  Did anybody
else noticed data corruption like this when using ufs logging on
Solaris 7?

Luckilly this disk is used only to temporarly store CD images.  But
I'm now a bit considered how safe is my data on other partitions?

Ah, yes...  on console, I also got message:


Jun  1 11:20:27 pc-alex.srce.hr         Error setting up next portionof DMA transfer

????

--

                  Opinions expressed herein are my own.
================================ooooO=Ooooo==============================
  Real Users never know what they want, but they always know when your
                       program doesn't deliver it.

 
 
 

1. UFS logging VS Solstice DiskSuite's Trans metadevice "UFS logging"

Does anyone know the difference from mounting a UFS file system with
the options "logging" and using Solstice DiskSuite's Trans metadevice
to do UFS logging.

Is it the same, or using DiskSuite more reliable?   I've been using
DiskSuite (ufs logging) for a few years now and I've never have a
problem with any file system from a hard crash or "power off".  I
noticed the other day when glancing over the mount_ufs man page that
there was a "logging" options that says the same stuff of what a trans
metadevice does, but was curious if it is proven to have the same
reliability.

Any info is appreciated.

-- Shawn

2. openbsd raid1 bootloader support

3. fcntl file locking goes bezerk

4. default route instead of static

5. UFS logging - how much space can be logged ?

6. Mousey probs

7. UFS logging vs meta device logging

8. AIX 4.3.3 conks out under heavy load

9. Sharing Logs with ODS + UFS Logging

10. how large is the log area on a UFS filesystem with logging.

11. VRTS RAID 1+0 volume, DRL Log and UFS log

12. mount freebsd's ufs in solaris7 x86

13. System logging going to console not log file