[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH 0/1] Merge DanPB's SCSI HBA pool code



On Mon, Feb 23, 2009 at 09:41:20PM -0500, Dave Allan wrote:
> Guido Günther wrote:
>> On Mon, Feb 23, 2009 at 07:53:07PM +0000, Daniel P. Berrange wrote:
>>> On Mon, Feb 23, 2009 at 06:03:07PM +0100, Guido G?nther wrote:
>>>> On Fri, Feb 20, 2009 at 05:14:54PM -0500, David Allan wrote:
>>>>> This patch contains the implementation Daniel Berrange did of storage
>>>>> pools using SCSI HBAs.  I have updated it for the current tree in
>>>>> preparation for implementing NPIV support.  Let me know what you
>>>>> think.
>>>> This looks like a great addition but I wonder if it is of any real use
>>>> without supporting multipath? On SANs issuing I/O to some paths might
>>>> cause severe performance penalties or the LUN might be visible but I/O
>>>> is rejected until an explicit switch over command is sent. This would be
>>>> handled transparently if multipath would be used on top.
>>> This is *not* just for FibreChannel based HBAs. Even single host SATA
>> Sure.
>>> controllers show up as SCSI HBAs, so its perfectly usable even in that
>>> case. So there's no need to block on multipath support
>> It's just that using it on fibre based HBAs could cause trouble so
>> thought I'd ask.
>
> Hi Guido,
>
> People can already use fibre channel block devices like any other block  
> devices with libvirt, so it's already possible for people to try to use  
> a passive path on an active/passive array, or use two active paths to  
> the same LU thinking they're different LUs, etc.  That's being the case,  
> I don't think there's anything in this patch that makes it more likely  
> that people will make those kinds of mistakes.  This patch just makes it  
> easy to list the LUs accessible by a particular host and provides a  
> foundation on which NPIV operations can be implemented.
You are of course right. My mail probably sounded overly "pessimistic".
Dan's SCSI HBA pool code is a great addition! I just can't wait until we
support active/passive arrays ;)
 -- Guido


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]