xfs
[Top] [All Lists]

Re: Xfs partition info seems to be lost on reboot

To: Melissa Terwilliger <techess@xxxxxxxxx>
Subject: Re: Xfs partition info seems to be lost on reboot
From: Mike Young <myoung@xxxxxxxxxxxxxxxxxxx>
Date: Tue, 9 Nov 2004 15:29:11 -0800
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <BDB6B0AF.2299%techess@xxxxxxxxx>
References: <BDB6B0AF.2299%techess@xxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: KMail/1.7
Melissa,

I too had problems related to this configuration.  In my case I was using a 
12-channel 3Ware 9500 series controller along with 12 WD 250GB drives.  XFS 
has been my default FS for some time now.  Also, I am running SLES9 SP1 
(2.6.5-7.111).

My 3Ware-based RAID began to complain about bad data in the scatter gather 
lists.  Soon my system began oopsing after being up for only a few minutes.  
Soon after, my controller flat out died.  I tried hooking up the 12 WD drives 
to another controller to use with MD RAID.  All 12 of the drives have failed 
to be readable by any of the SATA controllers I have.  They all return seek 
errors when trying to mount the drives.

I have another RAID, based on 15 fibre-channel drives and using MD RAID5 with 
XFS.  This combination has also been great.  My partitions are configured as 
RAID partitions.  This gets autodetected by SLES when you reboot and the 
array gets reassembled without a hiccup.  I have had this array in operation 
for more than 6 months.  

In all fairness, this 15 disk array was created under a different distribution 
and the migrated over to SLES9.  However, I picked up several more Hitachi 
SATA drives to replace my 3Ware/WD RAID.  It too has had no problem on reboot 
with XFS, though it's only a small amount of time.

-Mike

On Tuesday 09 November 2004 14:36, Melissa Terwilliger wrote:
> I have a dual opteron SuSE 9 SLES machine running 2.6.9 kernel.  The
> machine has 2 3ware 9500-12 PCI-X SATA raid cards installed with 12
> 250 GB drives on each card.
>
> I can create the partitions using parted & when I mount the two
> partitions  they show up correctly as 2.6 TB each and they seem to be
> working perfectly.  When I reboot the system I get the error:
> kernel:XFS:size check 2 failed in the kernel log and the mount command
> came back bad superblock on /dev/sdb1.  I can no longer mount the
> drives.  I tried doing a xfs_repair to fix the drive and I can mount
> it again, but it only shows up as 513GB after that.  Then I have to go
> into parted again to delete the partition and recreate it.  Again it
> works until I reboot whether or not I have it set to mount in the fstab
> or to mount by hand later.
>
> When I format the drive as ext3 I don't have this problem and it is 100%
> stable.  It only looses the information when I use xfs.  I'm sure I'm just
> missing something but I'm not certain what.
>
> Basically this is what I am doing to create the partitions:
>
> #parted /dev/sdb
> #mkpart primary xfs 0.0 2622488.000
> #print
> Disk geometry for /dev/sdc: 0.000-2622488.000 megabytes
> Disk label type: msdos
> Minor    Start       End     Type      Filesystem  Flags
> 1          0.031 525333.742  primary   xfs         type=83
> #quit
>
> then after the partition was created I did
>
> # mkfs.xfs -f /dev/sdc1
> meta-data=/dev/sdc1              isize=256    agcount=32,
> agsize=20979885 blks
>              =                       sectsz=512
> data      =                       bsize=4096   blocks=671356320,
> imaxpct=25
>              =                       sunit=0      swidth=0 blks,
> unwritten=1
> naming  =version 2              bsize=4096
> log        =internal log           bsize=4096   blocks=32768,
> version=1
>              =                       sectsz=512   sunit=0 blks
> realtime =none                   extsz=65536  blocks=0, rtextents=0
>
> #mount /dev/sdc1 /scr3
> #df -h
> /dev/sdc1             2.6T  528K  2.6T   1% /scr3
>
> I can mount and unmount this device repeatedly with no problems until
> I reboot.
>
> Thanks,
>
> Melissa Terwilliger
> Space Physics Research Lab
> AOSS Support
> techess@xxxxxxxxx
>
>
> -- Don't anthropomorphize computers.  They hate that.


<Prev in Thread] Current Thread [Next in Thread>