[illumos-Advocates] RTI: RFE #937 - WRITE_SAME support for COMSTAR

[Gordon Ross]

The only thing that was not clear to me was whether the _illumos_ code
had been tested, or only the NCP code, and how much they differ.
(If they're the same, then I'll accept NCP tests.)

I don't mean to stand in the way.  I just wasn't sure I wanted to be
listed as the approver if it goes badly (not that I think it will:)

Gordon

On Wed, Jun 1, 2011 at 11:21 AM, Garrett D'Amore <garrett at damore.org> wrote:
> I recall that he did quite a lot of perf testing.  Was there something specifically you were looking for?   (I'm happy with the testing, but I don't want to overrule you if you felt there was something specific you wanted to see.)
>
>  -- Garrett D'Amore
>
> On Jun 1, 2011, at 7:13 PM, Gordon Ross <gordon.w.ross at gmail.com> wrote:
>
>> Sorry for the lack of action on this, Dan.
>>
>> I remember now, I wasn't sure whether the testing was adequate.
>> Nobody else commented either way.
>>
>> So, other advocates:  Anyone want more testing?
>> Otherwise, I'm inclined to approve Dan's RTI.
>> Let me know today, please, if you want more time.
>>
>> Thanks,
>> Gordon
>>
>>
>> On Fri, May 20, 2011 at 3:19 PM, Dan McDonald <danmcd at nexenta.com> wrote:
>>> On Fri, May 20, 2011 at 03:12:58PM -0400, Gordon Ross wrote:
>>>> Hi Dan,
>>>>
>>>> What testing is appropriate for this?
>>>> and what has been done?
>>>
>>> In-house, I've done write-same testing on the NCP3 version of this change.  I
>>> used VMware to iSCSI-initiate to an NS box running these changes.  I wiped
>>> 9GB of disk clean using write-same with multiple parameters and even one
>>> implementation difference.  I was told the beta (not yet released) VMware we
>>> have in-house is the only one that speaks write-same.  OTOH, Google searching
>>> indicates release VMware ESX4.2 does it.  I found it hard to get even basic
>>> tests done.
>>>
>>> If you want some Illumos-specific testing or further tests on the NCP3
>>> version, please let me know.  Off the top of my head, I can see maybe:
>>>
>>>        - Finding more VMware instances to do parallel testing (though
>>>          multiple VMwares sharing the same LU... is that done in the real
>>>          world?  COMSTAR already separates LUs rather well).
>>>
>>>        - Something I haven't yet imagined.
>>>
>>> Hope this helps, and I appreciate any further suggestions,
>>> Dan
>>>
>>
>> _______________________________________________
>> Advocates mailing list
>> Advocates at lists.illumos.org
>> http://lists.illumos.org/m/listinfo/advocates
>