Russell Hires wrote:
>Hello all...
>
>I just switched over to XFS a couple of months ago, and my system has actually
>de-stabilized. I get kernel oopses all the time, and I know it's related to
>the filesystem. I didn't check the integrity of the disk before I put the XFS
>on it, so I'm sure that's part of it.
>
>
XFS, as with every filesystem requires that the underlying block device
be reliable. XFS will not correct problems with the underlying block
device. If you even remotely suspect the underlying block device of
being faulty, this is why your filesystem is corrupt.
The only thing I'm aware of that will even attempt a block level sector
auto-remap is the EVMS "bad block relocation" metadevice, which sits
below a filesystem. All modern harddrives do this internal to the drive
anyway (look into the SMART tools: learn how to use "smartctl" to kick
off online/offline media tests).
>So...I'm going to move everything again to a different portion of my disk
>(after doing a thorough check of it first), and then redo everything. I found
>that there are some nifty tools related to repairing XFS partitions, but they
>can't seem to fix everything. I have, for instance, some things
>in /lost+found to sort out what they are...
>
>
Filesystem checkers (fsck/xfs_repair/etc) put orphaned files and file
fragments into the lost+found directory for you to go through and
hopefully identify after a filesystem repair. These checkers are more
interested in cleaning your filesystem and making metadata "sane" than
they are about recovering lost data. Some filesystems have repair tools
that are better at this than others (reisierfs has always been abysmal
for me. I've had roughly the same luck with e2fsck as I have with
xfs_repair).
>I guess my question is, I don't understand exactly journaling filesystems. I
>thought that they would save me this kind of trouble. I wonder if I set up
>XFS the "right" way. I don't know if I want the journal to be on the same
>disk, or on another, or any of the other options that are available to me.
>
>
You generally want an "internal" journal, on the same block device as
the file data it is journalling writes for. For speed reasons, you may
consider splitting off an "external" journal to another block device to
save platter head reads, particularly if you are doing frequent writes
(think relational database). Honestly, you probably don't want to
consider anything but an internal journal for your home installation.
>I also don't know if I'm dealing with bugs in the XFS code, which is another
>thing. I'm using kernel 2.4.26...
>
>
I generally try to patch up to the latest XFS stable release I can apply
to a kernel at the time I'm compiling it. Using development code,
particularly straight out of CVS, will generally get you into trouble.
Verify the underlying block device. Consider using at least RAID1 to
protect against drive death.
And the #1 rule to remember regarding data storage: Always backup your data.
- Ian C. Blenke <icblenke@nks.net>
Networked Knowledge Systems
-----------------------------------------------------------------------
This list is provided as an unmoderated internet service by Networked
Knowledge Systems (NKS). Views and opinions expressed in messages
posted are those of the author and do not necessarily reflect the
official policy or position of NKS or any of its employees.
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 14:22:40 EDT