[ofa-general] Re: OpenSM and fat tree

Yevgeny Kliteynik kliteyn at mellanox.co.il
Sat May 17 08:52:45 PDT 2008


Sasha Khapyorsky wrote:
> On 11:37 Thu 15 May     , Chris Worley wrote:
>   
>> On Thu, May 15, 2008 at 11:14 AM, Ira Weiny <weiny2 at llnl.gov> wrote:
>>     
>>> On Thu, 15 May 2008 10:26:37 -0600
>>> "Chris Worley" <worleys at gmail.com> wrote:
>>>       
>> <snip>
>>     
>>>> After an sm change (i.e. using the "-r" switch), nodes can't ping each
>>>> other over IPoIB (other protocols also can't communicate).
>>>>         
>>> Is it absolutely necessary to run with the "-r" switch?  Here we have not
>>> problems letting the SM attempt to use the same LID's for nodes.
>>>       
>> yes, especially when chaging routing algorithms between the default
>> and fat-tree.
>>     
>
> As Yevgeny said it looks like an error (or at least unexpected behavior)
> in fat-tree code. Could you send ibnetdiscover output and "old" guid2lid
> file for us?
>   

There's also an open bug on bugzilla for this: 
https://bugs.openfabrics.org/show_bug.cgi?id=1031
(which also lacks the details that would help me to reproduce it).

-- Yevgeny

> Sasha
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
>
>   



More information about the general mailing list