[PVFS2-users] setting up multihome with same protocol

Phil Carns pcarns at wastedcycles.org
Wed Oct 12 16:12:44 EDT 2005


If this works as is, then it will be by pure luck :)  BMI should be 
getting both TCP addresses passed to it (this much is already in place 
to support using multiple methods), but it isn't doing anything clever 
with them.

-Phil

Rob Ross wrote:
> I'm not surprised that this doesn't work quite right at the moment.  Are 
> we at least capturing the two addresses and passing them to BMI so that 
> it *could* try both of them?
> 
> Thanks,
> 
> Rob
> 
> Robert Latham wrote:
> 
>> On Tue, Oct 11, 2005 at 11:17:05AM -0500, Robert Latham wrote:
>>
>>> (5) Verifying that all servers are responding...
>>>
>>>   meta servers:
>>
>>
>>
>> sorry, left off the rest of the information:
>>
>> (5) Verifying that all servers are responding...
>>
>>    meta servers:
>> [E 12:26:58.290061] *** msgpairarray_completion_fn: msgpair failed, no 
>> retry: Co
>> nnection timed out
>>    tcp://172.30.1.202:3334,tcp://140.221.8.226:3335 Failure!
>> noop_all_servers: Connection timed out
>> Failure: could not communicate with one of the servers.
>>
>> i.e. the client is getting both addresses for the server.  strace(2)
>> further tells me the client is connecting to the first addresss.  That
>> first address is unreachable from this client.  What i thought would
>> happen is the client would get an error from trying to connect to the
>> first address, then try to connect to the second address. 
>> ==rob
>>
> _______________________________________________
> PVFS2-users mailing list
> PVFS2-users at beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users



More information about the PVFS2-users mailing list