[Linux-cluster] gfs mounted but not working

[romero.cl at gmail.com]

Hi.

I'm trying your method, but still have a problem:

Note: /dev/db2/ is a local partition on my second SCSI hard drive (no RAID)
runing on HP ProLiant

On node3:
# /usr/sbin/vgcreate vg01 /dev/sdb2
  Volume group "vg01" successfully created
# /usr/sbin/vgchange -cy vg01
  Volume group "vg01" is already clustered
# /usr/sbin/lvcreate -n node3_lv -L 67G vg01
  Error locking on node node4: Internal lvm error, check syslog
  Failed to activate new LV.

  --->On node4 log : lvm[6361]: Volume group for uuid not found:
sgufJEs53VJSJTKG0vA1dLHXTthjnFctmfjC6YddzZvY3LI6db300wqEp8H0H58H


Then I can mount /dev/vg01/node3_lv as gfs on node3, but node4 can't view
the new files.

What i'm trying to do is to mount 2 partitions (one on node3, the other on
node4) as one big shared drive using gfs and then expand this to 4 nodes.

Any help is well appreciated!!! (i'm a cluster newbie)
Thanks.


> Hi,
>
> When using GFS in a clustered environment, I strongly recommend you use
LVM
> rather than using the raw device for your GFS partition.  Without a
> clustered
> LVM of some sort, there is no locking coordination between the nodes.
> I'm assuming, of course, that device sdb is some kind of shared storage,
> like a SAN.
>
> For example, assuming that your /dev/sdb2 has no valuable data yet, I
> recommend
> doing something like this:
>
> pvcreate /dev/sdb2
> vgcreate your_vg /dev/sdb2  (where "your_vg" is the name you choose for
> your new vg)
> vgchange -cy your_vg (turn on the clustered bit)
> lvcreate -n your_lv -L 500G your_vg (where 500G is the size of your file
> system,
>                 and your_lv is the name you choose for your lv)
> gfs_mkfs -p lock_dlm -t node1_cluster:node1_gfs -j 8 /dev/your_vg/your_lv
> (on only one node)
> At this point you've got to bring up the cluster infrastructure, if it
> isn't already up.
> Next, mount the logical volume from both nodes:
> mount -tgfs /dev/your_vg/your_lv /users/home
>
> Now when you touch a file on one node, the other node should see it.
>
> I hope this helps.
>
> Regards,
>
> Bob Peterson
> Red Hat Cluster Suite
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster