[Users] [node-devel] Help required to configure shared storage on RHEL+KVM

[Sanglikar, Amol (Serviceguard)]

Hi All,

Thanks for all your responses.

1/ we are using FC setup and not iSCSI. SO using SCSI is not an option for us right now.
2/ Yes you are correct NPIV will give us VHBA , but is it supported on guests as of now , both in case of RHEL6.2+ KVM and RHEV 3.0 ?
  If they have VHBA support then pass through helps me reach the LUNs and have SCSI PR working.

3/ Also does RHEV 3.0 support direct LUN access ? 

 What I got to read from redhat docs is as follows.

latest "BETA" release of RHEV 3.1 seems to be only having direct LUN access.

http://www.redhat.com/about/news/archive/2012/8/red-hat-delivers-first-beta-for-red-hate-enterprise-virtualization-3-1

New virtual machine storage features including hotplug support and direct LUN access


SO I guess only option I have os to wait for RHEV 3.1 ?

Again Thanks a lot for answering my queries.


Thanks & Regards,
-Amol



-----Original Message-----
From: Johan Kragsterman [mailto:johan.kragsterman at capvert.se] 
Sent: Thursday, August 30, 2012 12:46 PM
To: Itamar Heim
Cc: Sanglikar, Amol (Serviceguard); David Jaša; users at ovirt.org
Subject: Re: [Users] [node-devel] Help required to configure shared storage on RHEL+KVM

Hi!

I don't think this is an oVirt discussion, but I can respond anyway, since the storage part of oVirt, imho, is a bit after in development, or has a strategy I don't really understand, perhaps. Maybe oVirt storage development strategy is hard tied to gluster?

You don't tell us about the storage backend you got? If you got FC you can create an NPIV on the RHEL 6.2 host. I don't think(unsupported so far...) you can create NPIV on oVirt node or RHEL-V node, but I'm not sure about the latter...

 NPIV will give you the raw LUN directly to a VHBA, but I am not sure the oVirt/KVM VM can use it the VHBA...?

Or as Itamar writes, you can always use iScsi....

Rgrds Johan


 

-----users-bounces at ovirt.org wrote: -----
To: users at ovirt.org, "Sanglikar, Amol \(Serviceguard\)" <amols at hp.com>
From: Itamar Heim 
Sent by: users-bounces at ovirt.org
Date: 2012.08.29 13:53
Cc: David Ja&#353;a <djasa at redhat.com>
Subject: Re: [Users] [node-devel] Help required to configure shared storage on RHEL+KVM

On 08/29/2012 12:34 PM, David Ja&#353;a wrote:
> (moving to users@, node-devel is definitely a wrong list but I'm not
> sure about a correct one)
>
> hi Amol,
>
> Sanglikar, Amol (Serviceguard) pí&#353;e v Pá 24. 08. 2012 v 10:20 +0000:
>> Hi Ovirt dev-Community,
>>
>>
>>
>> We have a host running RHEL 6.2 + KVM , and created 2 Virtual machines
>> on the host which are running RHEL6.2.
>>
>>
>>
>> We are trying to configure shared storage for 2 VMs. This shared
>> storage on VMs should understand the SCSI requests.
>
> IIRC, if you need to pass through SCSI requests from the VM to the
> underlying block device, you need virtio-scsi qemu device that is
> included in RHEL as a tech-preview since 6.3 - again IIRC.

not sure this is about ovirt at all?
why not iscsi from guests to storage directly?

>
>>   The reason for this requirement is that we would be using SCSI-3 PR
>> (persistent reservation ) utils, which is sg3_utils to register /
>> reserve  on the storage/disks.
>>
>>
>>
>> We believe that exposing the raw LUNs from host to VMs as shared
>> storage (which is same as RDM - Raw Device Mapping in VMware ESX
>> environment )  would meet our above stated requirement.
>>
>>
>>
>> So can you please help us understand the following
>>
>>
>>
>> 1/  How to expose the Raw LUNs , from host ( RHEL 6.2 + KVM ) to
>> Virtual machines which are running RHEK 6.2.
>>
>
> You should create a shared disk but you can't get it really raw in terms
> of scsi commands unless you switch to virtio-scsi. It is raw in a sense
> of not using intermediate layers like qcow2 on top of the LUN.
>
> David
>
>> 2/  Also , if there is any alternate way to achieve the above
>> requirement in RHEL + KVM environment, apart from the mechanism we
>> stated , please let us know.

_______________________________________________
Users mailing list
Users at ovirt.org
http://lists.ovirt.org/mailman/listinfo/users