[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference decwet::ultrix-disk-shadow

Title:ULTRIX-DISK-SHADOW
Moderator:DECWET::MORRIS
Created:Wed Oct 24 1990
Last Modified:Wed Apr 23 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:250
Total number of notes:947

249.0. "fsck can't read block" by NETRIX::"collins@pylon.aqo.dec.com" (William Collins) Thu Mar 27 1997 17:11

I'm at a site which shadows 3 partitions of a disk on a few (> 5) duplicate
hosts.  I've discovered a rather curious behavior, always on a d partition for
each host.  This site has peculiar habits(cf, note 248) in which a mistake
corrupts the filesystem which resides on a shadow member.  So...

While we can sometimes assemble a suspect shadowset, we can always fail to
fsck
the same partition on each host.  (Even hosts which we have not had a failed
disk or corrupted filesystem.)  On every failed case, fsck can't read a block
near the end of the partition and exits, all without producing an I/O error
in uerf.  (It's definately not a media problem.)

I believe that fsck is reading a filsystem block (8k in this case), which
would read into the Metadata, given the failed block number.  If this is so,
I suppose the read error (sans I/O error) is the kernel protecting the
metadata.

Why, as we (I believe) follow all the directions(use shadnewfs AND NOT
newfs, etc...) does the filesystem cause fsck to fail?  shadnewfs should not
create a filesystem that would point fsck beyond the boundry(ie, into the
meta data.)  Right?

Not a shadfsck anywhere in sight. :-)

						Bill
------
collins@denver.enet.dec.com (DENVER::COLLINS)
[Posted by WWW Notes gateway]
T.RTitleUserPersonal
Name
DateLines
249.1NABETH::alanDr. File System's Home for Wayward Inodes.Mon Mar 31 1997 21:218
	If the customer is paying for a support contract, you might
	want to start an IPMT on the odd chance that the group
	responsible for maintaining the shadowing driver has forgotten
	this conference exists...

	Does fsck give the LBN of the area with the problem?  If
	so, then it should be easy to reproduce outside fsck and
	see what errno is set to.