one HBA for each node of an MC/ServiceGuard Cluster

[Kouider]

We would like to build an two node hp-ux cluster but we are limited with only on FC HBA on each node.

1 Is this configuration supported by HP?
2 If no, why do I need to have two HBA's?
3 Is it a question of the setup being certified by HP?

thanks in advance
Ali

P.S. This thread has been moevd from general to HP-UX > Serviceguard. - Hp Forum moderator

[Luk Vandenbussche]

Ali,

It will perfectly work.
Supported I don't know. But I also would advise you two HBA for redundancy reasons.

1 HBA : single point off failure
Connected to one FC switch : single point off failure

[Kouider]

Thanks Luk,

Can somebody else contribute regarding the supportability of the setup and the certification of the cluster?

TIA
Ali

[Duncan Edmonstone · Accepted solution]

Ali,

The short answer is no, this configuration is not supported. See the following section of the Serviceguard manual:

http://docs.hp.com/en/B3936-90100/ch02s03.html

The key points in here are:

-------------------------------

It is required that you provide data protection for your highly available system, using one of two methods:

* Disk Mirroring

* Disk Arrays using RAID Levels and Multiple Data Paths

-------------------------------

It then goes on to say:

-------------------------------

To protect against Fibre Channel or SCSI bus failures, each copy of the data must be accessed by a separate bus; that is, you cannot have all copies of the data on disk drives connected to the same bus.

-------------------------------

The point is, Serviceguard doesn't provide any protection against disk IO failure. This is provided by alternate paths to a disk array, or disk mirroring down different paths.

So whilst you could set this up, it wouldn't be a 'supported configuration' as defined by HP. What would happen if your HBA or SCSI/FC cable failed? The node effected would hang, and there certainly wouldn't be a failover. Whilst there are ways around this, they are fairly crude and harsh (using EMS resources to monitor the disks, and setting up node fail fast to force the system to TOC if a disk is inaccessible).

As pointed out - you *can* build this config without any problem, but I wouldn't consider doing this for anything except a sandpit/lab environment

Why can't you use 2 HBAs?

HTH

Duncan

[Kouider]

Thanks Ducan for your reply.

The question was asked by the management. I presume they want to save money.

it happens :)