Booting HP-UX server from disks on EVA 5000.

[Ronnie Doggart]

Hi All,

We have two EVA 5000's with Continuous Access running. On site 1 we have an HP-UX RX4640 up and running with its disks CA'ed to site 2. On site 2 we have an identical RX4640 which we would like to boot of the CA disks in the event of a disaster on site 1. Has anyone tried this, we cannot get the second RX4640 to boot after we have failed the disks over, it sees the disks, but fails when trying to start the securepath driver.

Ronnie

[Victor Fridyev]

Hi,
I'm not sure that it's possible to boot HPUX machine from a disk storage, which is not a JBOD, especially using securepath.

HTH

[Kent Ostby]

Per this site, it should work:

http://www.hp.com/products1/serverconnectivity/support_matrices.html

[Duncan Edmonstone]

Ronnie,

This should be possible (I've done it on EMC DMX using powerpath anyway), but will no doubt require some re-configuration as part of the failover - what error message are you getting? Can you at least boot to LVM maintenance mode?

HTH

Duncan

[Ronnie Doggart]

We can boot to lvm maintenance mode and have a procedure put together which involves exporting the old disk pathes and importing the new disk pathes to the various lv groups.

We also have windows, linux and tru64 servers and we basically failover the disks and boot the dr servers.

I was however hoping there was a simplier way to failover and boot the HP-UX boxes than the one we have.

[Duncan Edmonstone · Accepted solution]

Ronnie,

Nope - unfortunately you have to muck around in LVM maintenance mode - you could look at using VxVM as your volume managemnt tool instead - that might handle things better (I don't know enough about VxVM to comment)
As another approach have you thought about using HP Integrity Virtual Machines at both ends - use local storage for the VM hosts and then make your current systems guest OS's with their storage on the SAN. Then you could have the systems at the DR end adctually active, but with the guest OS's stopped until you failover?

HTH

Duncan

[Ronnie Doggart]

Duncan,

Good idea, will have to investigate Oracle's position on running the database in a HP-UX Virtual Environment.

Ronnie

[Zinky]

Ronnie,

Of course Boot Over SAN disks specially HP SAN arrays are fully supported. See:

http://search.hp.com/redirect.html?type=REG&qt=HP-UX+SAN+Boot+Sypport&url=http%3A//docs.hp.com/en/7408/HPUXSANBootWP_final_v2_update.pdf%3Fjumpid%3Dreg_R1002_USEN&pos=1

What you are attempting to do is essentially the best use out of your CA'd EVA5Ks. It should work but you will need to come up with a few interim steps (sorry not automagic) before you can boot off your cloned OS disk ar your remote site. And these steps are include booting in LVM maintenance mode and essentially recreating your IO tree... I already forgot the trick wiht SecurePath but I think once your IOCONFIG and network config at your failover server is fixed, SecurePath should be able to hurdle its problems.

[Zinky]

BTW..Best recipe though is to use VxVM as your volume manager as it is immune to cXtYdZ changes. I have done EVA5K boots (SecurePath always required btw IF you are at VCS 3.028 and below) and am currently testing SAN boots to XP12000 LUNs -- successfuly so far with LVM (alternate pathing path protection) and VxVM (DMP).

[Fabio Ettore]

Hi Ronnie and everybody,

I take advantage of this thread to underline a thing.
It is not the first time that I read about boot HP-UX from EVA5k; I just read that I cannot avoid my mind to remember troubles that I faced with Ignite-UX and EVA5k. As I consider Ignite-UX is a great and the best tool in order to recover my system in failover situations then I think about Ignite-UX

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=974994


and already posted all in the thread above.


Conclusion: if like I think people wants to backup vg00 by Ignite-UX and boot disks are on EVA5k in SAN keep in mind that Ignite-UX supports EVA5k once installed an exact version of VCS (4.01) which does EVA5k Active/Active.

Good luck...

HTH.

Best regards,
Fabio