[Pacemaker] Problem when relating it to colocation under placement-strategy environment

Yuusuke IIDA iidayuus at intellilink.co.jp
Fri Jul 15 03:30:16 EDT 2011


Hi, Andrew

(2011/07/15 15:09), Andrew Beekhof wrote:
> On Fri, Jul 15, 2011 at 3:16 PM, Gao,Yan<ygao at novell.com>  wrote:
>> On 07/15/11 10:55, Andrew Beekhof wrote:
>>> On Thu, Jul 14, 2011 at 4:28 PM, Gao,Yan<ygao at novell.com>  wrote:
>>>> Hi,
>>>> Sorry for the delay. I've been thinking about it...
>>>>
>>>> On 07/14/11 12:21, Andrew Beekhof wrote:
>>>>> This loop looks wrong
>>>>>
>>>>> +         for(gIter2 = resource1->rsc_cons; gIter2 != NULL; gIter2 = gIter2->next) {
>>>>>
>>>>> You're very dependant on the number and order of constraints because
>>>>> of the way resource1_weight is being updated.
>>>>> AFAICS, this only works if there is a single non INFINITY constraint.
>>>> Indeed. We can hardly tell what exactly the resources' scores are before
>>>> allocating resources. The scores would be merged/updated during
>>>> allocating. That means that we can hardly tell what the best allocating
>>>> order is before allocating resources. What "sort_rsc_process_order()"
>>>> does is just to predict a relatively ideal order.
>>>>
>>>>>
>>>>> I'll take a look at the before and after results tomorrow and see if
>>>>> there might be a better way to achieve the same results.
>>>> That would be great. Thanks!
>>>>
>>>
>>> Is there a bug I can reference in the commit message?
>> http://developerbugs.linux-foundation.org/show_bug.cgi?id=2613
>> http://developerbugs.linux-foundation.org/show_bug.cgi?id=2619
>
> Excellent, both are now fixed. I'll commit on Monday.
Because I thought that this was a difficult problem, I was surprised.
I look forward to Monday :-)

Thanks
Yuusuke
> Btw, you should join on irc - especially since we're in almost the
> same TZ now :-)
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>

-- 
----------------------------------------
METRO SYSTEMS CO., LTD

Yuusuke Iida
Mail: iidayuus at intellilink.co.jp
----------------------------------------




More information about the Pacemaker mailing list