[Linux-cluster] GFS2 on FC6 bonnie++ problem (lock_nolock)

Philippe Depouilly Philippe.Depouilly at math.u-bordeaux1.fr
Thu Nov 30 09:40:54 UTC 2006


Hi,

I have a repeatable failure with bonnie++ :

Testing a not clustered GFS2 filesystem on an iscsi filer from a single node directly attached 
(GFS2 formating with lock_nolock) with FC6 x86_64 (xeon 3GHz, 8GB RAM) through open-iscsi connection :

bonnie++ -d /mnt/ -b -s 16G -u root
Using uid:0, gid:0.
Writing with putc()...done
Writing intelligently...done
Rewriting...Segmentation fault

/var/log/messages :
Nov 29 19:19:37 stheno kernel: attempt to access beyond end of device
Nov 29 19:19:37 stheno kernel: dm-0: rw=1, want=3199534848, limit=1887436800
Nov 29 19:19:37 stheno kernel: Buffer I/O error on device dm-0, logical block 12498182
Nov 29 19:19:37 stheno kernel: lost page write due to I/O error on dm-0
Nov 29 19:19:37 stheno kernel: attempt to access beyond end of device
Nov 29 19:19:37 stheno kernel: dm-0: rw=1, want=3199534592, limit=1887436800
Nov 29 19:19:37 stheno kernel: Buffer I/O error on device dm-0, logical block 12498181
Nov 29 19:19:37 stheno kernel: lost page write due to I/O error on dm-0
etc.

Commands to create the filesystem on a 1To partition :
pvcreate /dev/sde
vgcreate lvol0GFSsde /dev/sde
lvcreate -L 800G lvol0GFSsde
mkfs.gfs2 -j1 -p lock_nolock lvol0GFSsde
mount /dev/lv0GFSsde/lvol0 /mnt

Note : on same filer accessing from same server with same OS, bonnie works fine with ext3 partitions.
This is why I try to post to linux-cluster and not FC6 mailing list.

If you have any advice or any pointers.

Thank-you very much.

-- 
Philippe




More information about the Linux-cluster mailing list