Unable to mount filesets

[Amanda Deer]

I had created a filedomain/filesets on a SAN. We lost connection to it, but now we have it back and I can't access my filedomain/filesets. If I try to mount them I get 'I/O Error'. The directory and link are still in /etc/fdmns. If I try to cd to the mount point directory I get 'bad directory', if I try to umount I get 'Device busy'.

I've tried using advscan and verify - any suggestions?

[Ivan Ferreira]

If you run hwmgr show scsi, and hwmgr v d, do you see the device?

Can you run:
disklabel -r dskX

If you get I/O error again, maybe is a know issue.

If this is an HSG80 controller, this may be a problem with the SCSI reservations. There are SCU commands to clean the SCSI reservation. As a last resource, the command cleanPR can be used to clean the reservations.

Currently, I can't find the document to clean the scsi reservations via SCU.

[Ivan Ferreira · Accepted solution]

As i couldn't find the link to the document, i attached a copy.

[Amanda Deer]

Yes I can see the device using hwmgr, and the disklabel.

[Ivan Ferreira]

From the results of hwmgr show scsi, obtain the HWID for the device. Then run

hwmgr get attr -id HWID |grep paths

Verify the status of the paths.

Run fuser -u /mountpoint to try obtain the lists of process that are using the device. Kill the process if you can and then try to umount it.

How did you lost connection to the SAN and how did you get it back?

[Michael Schulte zur Sur]

Amanda,

could you give us more details about the SAN involved and the os release and patch kit?
fuser -k /mountpoint includes the killing of any process having open files on that fileset. Could you post the links and the entry in fstab?

greetings,

Michael

[SlopeZ]

Hi Amanda,

can you check if in the /var/adm/messages
you have a Domain Panic?

When i get the same problem the only way is reboot the machine.

My SAN is a HSG80 and the SO can not work properly after a connection lost.

Anyway, post the information about SO, SAN requested by Michel.

Regards,
Santibon.

[Amanda Deer]

Sorry should have mentioned that we have v5.1A and Trucluster, with 2 nodes. I have managed to umount the filesystems using cfsmgr, but still get I/O error when trying to mount. Yet the paths to the disks appear to be there.

Looking into persistant reserves now, please let me know if you think I'm barking up the wrong tree. Thanks

[Amanda Deer]

The SAN is IBM ES800 and we go through a SVC.

[Amanda Deer]

Looks like it was a problem with persistant reserves. Once I removed them i could them mount the filesets. Cheers for all your suggestions.