[Linux-cluster] GFS volume already mounted or /mnt busy?

[Robert Peterson]

bigendian+gfs at gmail.com wrote:
> I had a curios thing happen last night. I have a two-node GFS cluster 
> configuration that currently has only one node.  After shutting down 
> and restarting the one node, the I couldn't mount my GFS volume 
> because it was no longer visible.
>
> The pvdisplay, lvdisplay, and vgdisplay all came up blank.  I was able 
> to use pvcreate --restorefile and vgcfgrestore to get the volume 
> back.  I then got the following message when trying to mount the volume:
>
> mount: /dev/etherd/e1.1 already mounted or /gfs busy
>
> I was able to gfs_fsck /dev/etherd/e1.1, but I continue to get this 
> error.  Running strace on the mount command turns up this error:
>
> mount("/dev/etherd/e1.1", "/gfs", "gfs", 
> MS_MGC_VAL|MS_NOATIME|MS_NODIRATIME, "") = -1 EBUSY (Device or 
> resource busy)
>
> What could be happening here?
>
> Thanks,
> Tom
Hi Tom,

Hm.  Sounds like something bad happened to the logical volume (i.e. LVM).

Out of curiosity, what was happening on the other node?  It wasn't, by 
chance, doing
an install was it?  In the past, I've seen where some versions of the 
Anaconda installer
loads the QLogic driver, detects my SAN and offers to automatically 
reformat it as
part of the installation.  I hope that didn't happen to you, or if it 
did, that you
unchecked the box for your SAN where the eligible drives were listed.

I'd check all the systems that are attached to the SAN, regardless of 
whether or
not they're part of the cluster.  See if one of them has done something 
unexpected
to the device.

Regards,

Bob Peterson
Red Hat Cluster Suite