I've just installed Centos 5 on a Dell P1950 system. I'm trying to set up a gfs filesystem (in preparation for creating a cluster). The disks are in an Equalogic box attached via ISCSI, and appear to be working fine. I can create the file system with no problems. However, I can't mount it. I get an error 19, which I think is "no such device". Here is a transcript:

Script started on Wed 18 Apr 2007 06:52:54 PM EDT

root@merge01 ~]# gfs_mkfs -p lock_nolock -j 10 /dev/sdd
This will destroy any data on /dev/sdd.
It appears to contain a gfs filesystem.

Are you sure you want to proceed? [y/n] y

Device: /dev/sdd
Blocksize: 4096
Filesystem Size: 2294792
Journals: 10
Resource Groups: 36
Locking Protocol: lock_nolock
Lock Table:

Syncing...
All Done
[root@merge01 ~]# gfs_fsck /dev/sdd
Initializing fsck
Clearing journals (this may take a while).....
Journals cleared.
Starting pass1
Pass1 complete
Starting pass1b
Pass1b complete
Starting pass1c
Pass1c complete
Starting pass2
100 percent complete.
Pass2 complete
Starting pass3
Pass3 complete
Starting pass4
Pass4 complete
Starting pass5
100 percent complete.
Pass5 complete
Writing changes to disk
[root@merge01 ~]# mount -t gfs /dev/sdd /opt/oracle/orahome
/sbin/mount.gfs: error 19 mounting /dev/sdd on /opt/oracle/orahome
[root@merge01 ~]# exit

Script done on Wed 18 Apr 2007 06:53:47 PM EDT

Can anyone help me get this working?

(Please let me know if this should be posted in another forum.)

Thanks.

Sorry -- forgot one additional piece of information. I was able to create and mount an ext3 file system on this same device with no problems, so I'm pretty sure this is a GFS issue of some kind.

Still not working, but maybe this is another piece of the puzzle. When I look in /sbin to see the various gfs files there, I find:

/sbin/mount.gfs -> mount.gfs2
/sbin/umount.gfs -> umount.gfs2

Every other program (fsck, mkfs, grow, etc.) has a version of gfs and gfs2. Where is the gfs mount? Is this why I can't mount?

Can anyone tell me where I can find a regular mount.gfs program so I can try it?

Yet more in the saga....

I've discovered that I can create and mount a gfs2 filesystem, but not a gfs filesystem.

Does anyone know why we can only use gfs2?

The docs say it is not ready for production yet. How do I get a gfs filesystem working instead?

The issue is resolved. I had numerous issues:

1) copy the gfs directory from 2.6.18-8.el5/extras to 2.6.18-8.1.1.el5/extras (it installs to the wrong directory)
2) depmod -a
3) modprobe gfs
4) chkconfig gfs on
5) service gfs start

I discovered that the kmod-gfs package was missing when I did a "yum groupinstall 'Cluster Storage'" and then couldn't find the gfs kernel module. A "find" located it in the wrong /lib/modules area.

This system was originally created by doing an Upgrade install via the DVD from CentOS 4.4 (without cluster storage). It's possible that some of these issues would go away if it was an original install or if Cluster Storage was installed on CentOS 4.4 first.

Who do I report the configuration problem to?

I did your steps and it seemed to help a little, but I still cannot get my cluster to work. I have created clusters in the past with RHEL4/5 and everything worked fine. I think something in C5 is broken.

Quote:

We too encountered "/sbin/mount.gfs: error 19" when trying to mount a GFS volume on an up-to-date CentOS 5 cluster we're trying to build. However the instructions you gave were for a standard kernel. Those with more then 3 GB of RAM must use the "2.6.18-8.1.1.el5PAE" kernel, which errors after copying the file and doing a modprobe:

FATAL: Error inserting gfs (/lib/modules/2.6.18-8.1.1.el5PAE/extra/gfs/gfs.ko): Invalid module format

We're trying a "yum install kmod-gfs-PAE", but it requires the base PAE kernel, which we don't have. Yum won't let us install it either (package kernel-PAE-2.6.18-8.1.1.el5 (which is newer than kernel-PAE-2.6.18-8.el5) is already installed).

Thanks for any tips,
Rich

It seems like this (broken cluster/gfs) would be a bigger deal than it appears to be.

install the latest gfs-kmod

see: http://rhn.redhat.com/errata/RHBA-2007-0372.html