[rhelv6-list] Problems with XFS leaving 0-length files in RHEL 6.2 ... ?

[Edmund White]

What are the names of these files? Can you see when these files are
created? This may not be an XFS-specific issue.

If you *do* move to EL6.3, there is a major change in XFS that can impact
your storage requirements/performance.

A feature named "XFS Dynamic Speculative EOF Preallocation" was committed
to the 2.6.38 kernel in late 2010/early 2011 and made its way into the
EL6.3 distribution this year.

I won't go into full detail here, but I wrote a comprehensive breakdown of
the feature and impact on ServerFault.com this summer.
See: http://serverfault.com/q/406069/13325

-- 
Edmund White
ewwhite at mac.com




On 10/13/12 12:19 AM, "Paul Smith" <paul at mad-scientist.net> wrote:

>Hi all.  We've been using standard RHEL 6.2 on our test systems (no
>updates, straight off of the Server DVD... however we load only a
>small-ish subset of the full OS).
>
>Lately we've been seeing issues where we have a number of 0-length files
>on our partitions.  These systems have not (as far as I'm aware) crashed
>or had any kind of power hit or anything like that, and the files that
>are 0-length don't appear to be modified much, if at all.  It seems
>bizarre that they should suddenly be empty.
>
>These filesystems are on 2.7T partitions (each), and are on rack-mounted
>servers with three harddrives RAIDed using an LSI MegaRAID SAS 2208
>hardware RAID controller.
>
>The filesystems are created using the XFS filesystem... mainly we chose
>XFS because it's MUCH faster to create large partitions like this than
>ext, which we're doing a good bit.  It was a supported filesystem in
>RHEL 6.2, so I thought it would be robust.  I can't say for sure that
>this is an XFS problem; I'm trying to get more rigorous about detecting
>this problem closer to when it happens so I can check logs, etc.  On the
>other hand I've heard rumors about issues like this in XFS, but I
>thought all the known ones were resolved in RHEL 6.2.
>
>So, I'm looking for information:
>      * Does this seem likely to be an XFS problem?  Should we move to
>        ext4?
>      * Would it help this problem to go to RHEL 6.3?  Any known XFS
>        issues,  esp. kernel issues, resolved between 6.2 and 6.3?  I
>        took a spin through the 6.3 release notes but didn't see
>        anything.
>
>
>_______________________________________________
>rhelv6-list mailing list
>rhelv6-list at redhat.com
>https://www.redhat.com/mailman/listinfo/rhelv6-list