EVA 4400 Redhat Linux 5.1

[Mike Elliott_1]

I have an EVA 4400 that is presenting luns to two Redhat Linux 5.1 machines. I am experiencing a problem when moving luns from machine to machine.

Here are the details of the EVA 4400

xcs 090006000
Management Module firmware mmp-0001-1000-000117R

The machines are

DL380 G5
Redhat Linux 5.1
Qlogic Failover Driver v 8.02.11-2
Lifekeeper is installed on both machines

Description of problem:

Server A and Server B are both presented 2 luns each. Both machines have SCSI2 reservations on their respective luns.

If Server A experiences a full disconnection from the fabric (i.e. dual fiber break) and Server B puts a SCSI2 reservation on the luns (i.e. lifekeeper software fails the devices over to server B). Then server A will not see the devices when it's reconnected to the fabric.

The only way to clear the SCSI2 reservations is to neuter the failover software (lifekeeper), reload the qlogic driver, and do a hp_rescan for the luns.

Has anyone else experienced such a thing? Is there a better way of doing this (i.e. Device Mapper?)

Thanks in advance,

Mike

[Mike Elliott_1]

I have some more info, the engineers at Steeleye believe that the EVA has both the paths marked as down, or in a state that the Qlogic driver sees both paths as down.

When I view /proc/scsi/scsi on a reboot of the machine, there is nothing listed at all.

If I do an hp_rescan -a, it brings back the luns, however the partitions created are not listed in /proc/partitions

It's only when I use Lifekeeper to remove the scsi2 reservations, that I can rescan the luns, and get the proper partition information.

[Sivakumar MJ._1]

Mike, Just check if this helps u a little..

https://bugzilla.redhat.com/show_bug.cgi?id=491774

[Mike Elliott_1]

Thanks,

That is good news that someone with a slightly different architecture is experiencing the same symptoms.

I am wondering if his case is possibly based on our case as we opened a case with HP from two angles (one from our office and one from steeleye)