[Lustre-discuss] mount.lustre: mount /dev/sdd1 at /hpcdata/ost1 failed: Protocol error

[Andreas Dilger]

On May 05, 2008  12:38 +0200, Christian Gajan wrote:
> my lustre configuration is RHEL5.1 (2.6.18-53.1.13 + lustre patch) + OFED 
> 1.2.5.5 + lustre 1.6.4.3
> my lnet seem to work correctly but when I try to configure my OSS I got a 
> Protocol error at the mount step
>
> Has someone already used lustre on infiniband (ofed stack) under redhat 5.1 
> (2.6.18-53.1.13) ?
> If Yes : with which version of OFED and which version of Lustre ?
>
> Thanks in advance for your help
>
> Here some detail
>
> OST server
> # modprobe lnet
> # lctl network up
> # lctl list_nids
> 192.168.1.16 at o2ib
> # lctl ping 192.168.1.17 at o2ib
> 12345 - 0 at lo
> 12345 - 192.168.1.17 at o2ib
> # mkfs.lustre --fsname hpcdata --ost --mgsnode=192.168.1.17 at o2ib /dev/sdd1
> # mkdir -p /hpcdata/ost1
> # mount -t lustre /dev/sdd1 /hpcdata/ost1
> mount.lustre: mount /dev/sdd1 at /hpcdata/ost1 failed: Protocol error   
> !!!!!!!!!
>
> I got in the /var/log/messages (OST side)
>
> Apr 24 07:17:55 s1206 kernel: LustreError: 
> 13483:0:(pack_generic.c:782:lustre_unpack_msg()) bad lustre msg magic: 
> 0X26F54000

This magic value is completely incorrect - it should be something like
0x0bd00bd3.

Are you using some strange CPU architecture on the MDS or OSS node by
some chance?

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.