> i've been testing the 0.9 release of xfs for linux and the following error
> seems to come up randomly when i run `xfs_logprint` on my xfs disks.
> ---
>
> Header 0x4 wanted 0xfeedbabe
>
> ******************************************
> * ERROR: header cycle=4 block=705 *
> ******************************************
> Bad log record header
>
> ---
>
> is this a case where the most recent log overlaps the oldest log record as
> stated in the `xfs_logprint` manpage? does this command require that the
> disk be unmounted before it is run?
>
> -k
>
In general xfs_logprint on a live system is going to be iffy, it is looking
at a moving target. Also there are various modes of operation where logprint
will try to interpret all the data in the log even when it is not all valid.
If you build a filesystem and do not completely fill the log for example
you will get some bad headers.
My other comment is that the 0.9 release is a little old now, there is now
a 0.10 set of images out there. The quickest way to upgrade should be to
go to this directory:
ftp://oss.sgi.com/projects/xfs/download/PreRelease-0.10-test2/Wolverine-SGI-XFS-0.10/RedHat/RPMS/
download the kernel rpm and command rpms you need and do an rpm -U
Note that xfscmds got broken into several things
xfsprogs is the main package, the rest you do not need for basic stuff.
Steve
|