Note, Hemachandran has inadvertently given the wrong link for the 
updated design doc.

The correct link is:
http://opensolaris.org/os/project/ha-xvm/ha-ldoms-design-doc_v2.pdf


On 06/25/09 22:48, Hemachandran Namachivayam wrote:
> Hi Ellard,
>
> Thanks for the comments for the "Solaris Cluster agent for LDoms Guest 
> Domains" design review. Please see my replies below.
>
> The design document is uploaded at 
> http://opensolaris.org/os/project/ha-xvm/ha-xvm-req-spec-v1.2.pdf for 
> reference.
>
> -Thanks and Regards
> Hemachandran
>
> On 04/09/09 21:18, Ellard Roush wrote:
>>
>> To Hemachandran, CLARC
>> From Ellard
>> Subject Design Review LDom Guest Domain Agent 2008/1765
>>
>> ER-1 Page 1 4th paragraph
>>
>> The paragraph completely misses the real reason why migration does
>> not provide High Availability.
>>
>> The real reason that migration does not provide High Availability is
>> that migration requires that both the source and destination machines
>> be up. High Availability requires the ability to recover after failure.
>> Migration cannot be used in that case.
>>
> I have incorporated this comment.
>>
>> ER-2 page 1 5th paragraph (but also many other places
>>
>> Using the resource type name HA-xVM for xVM types of virtual machines is
>> fine.
>>
>> Using the resource type name HX-xVM for LDom types of virtual machines
>> is crazy !
>>
>> This is equivalent to using the resource type name HA-RAC for DB2 !
>>
>> Use an appropriate name. Either change the resource type name
>> to something generic, like HA-VM; or alternatively use
>> separate names, like HA-xVM and HA-LDom.
>> With both of these approaches the implementation code can be shared.
>> Changing the names of the implementation code is a minor item.
>>
> The agent would introduce a new RT SUNW.ldom for managing LDoms and 
> the existing SUNW.xvm will continue to handle xVM domains. I have 
> updated this information in the document.
>> ER-3 Figure page 4
>>
>> Many people do not have color printers.
>> When using color always ensure that the figure can be understood
>> when printed on a black and white printer.
>> This figure fails this test.
>>
>> Do NOT use dark colors behind dark letters ! This is a problem here.
>>
>> When using multiple colors always use colors that print out
>> as distinctly different on a black and white printer. This figure
>> fails here as well. There are other alternatives to color, such
>> as background patterns. The current colors make it MORE DIFFICULT
>> to understand this figure.
>>
> Thanks for pointing this. I have removed the colors from the diagrams.
>>
>> ER-5 CLARC does NOT do code reviews.
>>
>> The review of the resource properties is equivalent to code review.
>> The resource properties should be placed in an appendix.
>>
> The RTR properties are now populated in a table and the earlier format 
> is no longer present.
>> ER-6 page 9 2nd bullet
>>
>> What is "option TDB" ?
>>
>> An important goal of design document is to educate people.
>> Do not assume that everyone looking at the document is an expert.
>> People will read the document to learn about what the project did.
>>
> TDB is a spell mistake. I have corrected the sentence to "Ensure the 
> "failure-policy" option is set to ?reset? on the control domain for 
> the guest domain."
>
>> ER-7 page 10 3rd bullet
>>
>> When an admin is looking for information in the CCR about LDoms,
>> the admin is not going to look at "xvm_service". Change the name.
>> See point 2.
>>
> I have changed the name to "ldom_domain_config".
>>
>> ER-8 page 11 3rd bullet
>>
>> Explain this bullet.
>> It is not clear what this means.
>>
> I have rephrased the sentence to be more precise.
>>
>> ER-9 page 13 3rd bullet
>>
>> Need to explain more about when migration will be done.
>> Not all stop operations will result in a migration.
>>
> Incorporated.
>
> -Thanks and Regards
> Hemachandran
>>
>> _______________________________________________
>> ha-clusters-discuss mailing list
>> ha-clusters-discuss at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/ha-clusters-discuss
>
> _______________________________________________
> ha-clusters-discuss mailing list
> ha-clusters-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/ha-clusters-discuss

Reply via email to