SUMMARY: Wrong superblock state on a LUN filesystem

First of all, thanks to all who responded (not many), and a special thank
goes to Grzegorz Bakalarski for pointing me out that it was a known problem
with the newer fsck included within Solaris 10.

Looking for it in the bugs database at Sun Microsystems, I found out the
following, which is the problem I currently experienced:

---- cut here ----
Bug ID: 4775289
Synopsis: fsck reports wrong state in superblock if there once has existed a
largefile
Product: solaris
Category: utility
Subcategory: filesystem
Bug-RFE-EOU: Defect
Sub-Area: Functionality
State: 10-Fix Delivered
Keywords: SUPERBLOCK, files, fsck, large, 
Priority: 3-Medium
    Responsible Manager: lisa.chatham@sun.com
    Responsible Engineer: john.wren.kennedy@sun.com
Description: 
cust gets error messages from fsck even if the filesystem was cleanly
umounted:

FILE SYSTEM STATE IN SUPERBLOCK IS WRONG; FIX?

i could reproduce this with the folloeing steps:

1. create a fs that can contain largefiles ( fs > 2gb)
2. run fsck - this should be ok
3. mount the fs
4. create a largefile (mkfile -n 2500m testfile)
5. remove this largefile
6. umount the fs
7. run fsck - now fsck report the wrong state
---- cut here ----

I didn't find any fix yet (I'm still on search mode for a patch), but wanted
to let you know that this is known bug at Sun.

Thanks.

Ben Audet



----- Original post -----
Hi fellow managers,

I have a strange behavior here, and would like to know if some of you would
have an idea of a workaround on this.  We use two SunFire T2000, both linked
to a SAN.  On the SAN, we use two LUNs, one of 175gb and the other is 45gb.

On the first T2000, we put a container (zone) with Oracle 10g in it.  We
mounted the LUNs on the global zone of the first machine, and then we give
it access to the container with "lofs" settings, and it works great.

Now, we do some testing shutdowning the container on the first T2000, and
transfer it to the second T2000, and the second T2000 also sees both LUNs on
the SAN.  Everything but datafiles, archives and redo logs (which are on the
SAN LUNs) are on local disk, in the container.

When the container is shuted down, we as well "umount" both filesystems
associated with the LUNs on the SAN, and before powering up the container on
the 2nd T2000, we do an "fsck" on both raw devices of the LUNs and we
*always* get this error message, but only on the first LUN:

FILESYSTEM STATE IN SUPERBLOCK IS WRONG; FIX?

Of course, I say "yes" and it comes up fine: I'm able to mount both
filesystems on the 2nd T2000, the container comes up OK and Oracle is
starting up like if there was no problem...   ...but as long as we try to
mount the first LUN (the 175gb one), if we do a "fsck", it always comes with
the wrong superblock state error.  The 45gb LUN works flawless...

I've through regular log files, but found nothing.  I'm not controlling the
"SAN side" however, and wasn't able to access log files...  Any idea out
there?  I did a "newfs" on the LUN to make sure it was OK, put back data in,
but it still does this error.


Thanks in advance for any ideas!

Ben Audet
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers