HBA driver upgrade to 8.02.23

[ahmad_bakr]

I have a EVA 4400, and 2 linux bl680c servers with rhel 4.7 32-bit OS.
Whenever HBA driver is upgraded to 8.02.23 we loose connectivity with the EVA. However if we install rhel 4.7 from scratch, and install HBA driver 8.02.23 immediatly (but not upgrading) this driver works fine.

And one seperate thing, Once the rhel is reinstalled, we re-presented one lun, which was presented to the system before reinstallation. However, whenever we try to mount it, an error message shows up stating that the device is already mounted. But using the command: df -h shows nothing about that device.

Sorry for the long questions. And i am looking forward to find a solution. Thanks in advance.

[Matti_Kurkela]

> Whenever HBA driver is upgraded to 8.02.23 we lose connectivity

Upgraded to 8.02.23 _from which version_?

It might be helpful to read all the HBA driver release notes, beginning from (your old version + 1) to 8.02.23. Perhaps the upgrade has some special requirements.

Are there any error messages or other symptoms when the connectivity is lost? For example, any boot messages about incorrect kernel module parameters? (In that case, you should look into /etc/modprobe.conf and verify that all the old kernel module parameters are applicable to the new version too. After making any necessary changes, remember to use the "mkinitrd" command to re-create your initrd file.)

> [...] an error message shows up stating that the device is already mounted. But using the command: df -h shows nothing [...]

Compare /etc/mtab and /proc/mounts.

If you've e.g. restored /etc from backups, you may have accidentally restored an old version of /etc/mtab. That will cause the "mount" command to use out-of-date information.

Because /proc/mounts is generated by kernel in real time as it's accessed, it should always have up-to-date information. So if /etc/mtab and /proc/mounts disagree about whether something is mounted or not, you should trust /proc/mounts and treat /etc/mtab as out-of-sync.

On the other hand, in some situations (mostly with loop devices, and when root has given regular users permission to mount some filesystems) /etc/mtab may have more information than /proc/mounts about the mount options of a particular mounted filesystem.

What's the filesystem type on the problematic LUN?

If there are I/O errors when accessing the LUN, the behaviour is filesystem-specific. Ext2 and ext3 filesystems typically switch to read-only mode (and it can be seen in /proc/mounts only, not in /etc/mtab); GFS may completely stop accessing the LUN.

MK

[ahmad_bakr]

The model number of the HBA is QMH2462, the version which was running ok was:
Driver: 8.01.07.25

The customer followd the exact instructions in the release notes.

and 8.02.23 is compatible with rhel 4.7 32-bit.

for the second part ill have to check with the customer to get back to you.

[Andrew Cowan]

For most SAN upgrades you have to delete the LUNs, reboot, and then rediscover them.

This should solve your problem.