[Ocfs-users] Mouting Filesystem problems

[Sunil Mushran]

mkfs.ocfs does not build the root directory. That is upto the module
to do and it does so at the very first mount. To prevent races, we require
node 0 to mount the volume first. Once done, the root dir will be created
and parallel mounts can happen.

However, you have specified an optional node_number parameter in
ocfs.conf. And it is set to 2. Hence the problem.

Either remove the node_number parameter from ocfs.conf. You will
need to reload the module as that file is only read during load.

Or, mount the volume from another node which either has
no node_number set or set to 0.

David Sharples wrote:

>Hi, thanks for he reply.  Here are the messages from the time I tried
>to mount it on the second node
>
>Mar 10 17:41:47 LNCSTRTMIT03 kernel: (30369) ERROR: The volume must be
>mounted by node 0 before it can be used and you are node 1,
>Linux/ocfsmount.c, 289
>Mar 10 17:41:47 LNCSTRTMIT03 kernel: (30369) ERROR: Error mounting
>volume, Linux/ocfsmain.c, 314
>Mar 10 17:46:58 LNCSTRTMIT03 sshd(pam_unix)[27281]: session closed for user root
>Mar 10 18:01:07 LNCSTRTMIT03 sshd(pam_unix)[31795]: session opened for
>user root by (uid=0)
>Mar 10 18:01:08 LNCSTRTMIT03 sshd(pam_unix)[31795]: session closed for user root
>Mar 10 18:01:09 LNCSTRTMIT03 sshd(pam_unix)[31804]: session opened for
>user root by (uid=0)
>Mar 10 18:01:10 LNCSTRTMIT03 sshd(pam_unix)[31804]: session closed for user root
>Mar 10 18:05:17 LNCSTRTMIT03 kernel: (32120) ERROR: Invalid volume
>signature, Common/ocfsgenmisc.c, 820
>Mar 10 18:05:17 LNCSTRTMIT03 kernel: (32120) ERROR: Device (8,0)
>failed verification, Linux/ocfsmount.c, 232
>Mar 10 18:05:17 LNCSTRTMIT03 kernel: (32120) ERROR: Error mounting
>volume, Linux/ocfsmain.c, 314
>
>The node part interests me, do I need a node 0 in a /etc/ocfs.conf
>(the users guide I have doesnt have that parameter mentioned)
>
>
>On Thu, 10 Mar 2005 10:32:56 -0800, Sunil Mushran
><Sunil.Mushran at oracle.com> wrote:
>  
>
>>When you do the mount, the actual errors are dumped in /var/log/messages.
>>What are those messages?
>>
>>David Sharples wrote:
>>
>>    
>>
>>>Hi,
>>>
>>>I have blatantly done something wrong here but I dont know what.
>>>
>>>Two nodes and my SA presented the same SAN space to both nodes
>>>
>>>on node1 the device is /dev/sda and on node2 the device is /dev/sdb
>>>
>>>So on node I did this
>>>
>>>mkfs.ocfs -F -b 1024 -g dba -u oracle -L OCFS1 -m /ocfs1 -p 755 /dev/sdb
>>>
>>>and the output was
>>>
>>>Cleared volume header sectors
>>>Cleared node config sectors
>>>Cleared publish sectors
>>>Cleared vote sectors
>>>Cleared bitmap sectors
>>>Cleared data block
>>>Wrote volume header
>>>
>>>
>>>All returned OK with no problems, when I try to mount that file system I get
>>>
>>>mount -t ocfs /dev/sdb /ocfs1
>>>mount: wrong fs type, bad option, bad superblock on /dev/sdb, or too
>>>many mounted filesystems
>>>
>>>So ok, I did fsck.ocfs
>>>
>>>fsck.ocfs /dec/sdb
>>>
>>>fsck.ocfs 1.1.2-PROD1 Fri Apr  2 13:59:23 PST 2004 (build
>>>2df841d19c031db220f8cfb73439339d)
>>>Checking Volume Header...
>>>Volume has never been mounted on any node. Exiting
>>>/dev/sdb: clean, 0 objects, 0/20478 blocks
>>>
>>>I tried the exact same thing on the other node and got the same
>>>errors.  These partitions has already been mounted as linux
>>>filesystems before I did the mkfs.ocfs stuff.
>>>
>>>ocfs.conf
>>>
>>>
>>>       node_name = mit04
>>>       node_number = 2
>>>       ip_address = 10.85.151.11
>>>       ip_port = 7000
>>>       comm_voting = 1
>>>       guid = 6EFDD701A21BAEED6A36000E7FAE352A
>>>
>>>ocfs-support-1.1.2-1
>>>ocfs-tools-1.1.2-1
>>>ocfs-2.4.9-e-smp-1.0.13-1
>>>ocfs-2.4.9-e-enterprise-1.0.13-1
>>>
>>>
>>>ocfs module is loaded
>>>
>>>lsmod | grep ocfs
>>>ocfs                  306016   0
>>>
>>>Any ideas?
>>>
>>>Thanks
>>>_______________________________________________
>>>Ocfs-users mailing list
>>>Ocfs-users at oss.oracle.com
>>>http://oss.oracle.com/mailman/listinfo/ocfs-users
>>>
>>>
>>>      
>>>