Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] Spawn_multiple with tight integration to SGE grid engine
From: Ralph Castain (rhc_at_[hidden])
Date: 2012-02-01 15:14:12


FWIW: I have fixed this on the developer's trunk, and Jeff has scheduled it for release in the upcoming 1.6 release (when 1.5 series rolls over). I don't expect we'll backport it to 1.4 unless someone really needs it there.

Thanks!
Ralph

On Feb 1, 2012, at 9:31 AM, Ralph Castain wrote:

> Ah - crud. Looks like the default-hostfile mca param isn't getting set to the default value. Will resolve - thanks!
>
> On Feb 1, 2012, at 9:28 AM, Reuti wrote:
>
>> Am 01.02.2012 um 17:16 schrieb Ralph Castain:
>>
>>> Could you add --display-allocation to your cmd line? This will tell us if it found/read the default hostfile, or if the problem is with the mapper.
>>
>> Sure:
>>
>> reuti_at_pc15370:~> mpiexec --display-allocation -np 4 ./mpihello
>>
>> ====================== ALLOCATED NODES ======================
>>
>> Data for node: Name: pc15370 Num slots: 1 Max slots: 0
>>
>> =================================================================
>> Hello World from Node 0.
>> Hello World from Node 1.
>> Hello World from Node 2.
>> Hello World from Node 3.
>>
>> (Nothing in `strace` about accessing someting with "default")
>>
>>
>> reuti_at_pc15370:~> mpiexec --default-hostfile local/openmpi-1.4.4-thread/etc/openmpi-default-hostfile --display-allocation -np 4 ./mpihello
>>
>> ====================== ALLOCATED NODES ======================
>>
>> Data for node: Name: pc15370 Num slots: 2 Max slots: 0
>> Data for node: Name: pc15381 Num slots: 2 Max slots: 0
>>
>> =================================================================
>> Hello World from Node 0.
>> Hello World from Node 3.
>> Hello World from Node 2.
>> Hello World from Node 1.
>>
>> Specifying it works fine with correct distribution in `ps`.
>>
>> -- Reuti
>>
>>
>>> On Feb 1, 2012, at 7:58 AM, Reuti wrote:
>>>
>>>> Am 01.02.2012 um 15:38 schrieb Ralph Castain:
>>>>
>>>>> On Feb 1, 2012, at 3:49 AM, Reuti wrote:
>>>>>
>>>>>> Am 31.01.2012 um 21:25 schrieb Ralph Castain:
>>>>>>
>>>>>>> On Jan 31, 2012, at 12:58 PM, Reuti wrote:
>>>>>>
>>>>>> BTW: is there any default for a hostfile for Open MPI - I mean any in my home directory or /etc? When I check `man orte_hosts`, and all possible optiions are unset (like in a singleton run), it will only run local (Job is co-located with mpirun).
>>>>>
>>>>> Yep - it is <prefix>/etc/openmpi-default-hostfile
>>>>
>>>> Thx for replying Ralph.
>>>>
>>>> I spotted it too, but this is not working for me. Neither for mpiexec from the command line, nor any singleton. I also tried a plain /etc as location of this file as well.
>>>>
>>>> reuti_at_pc15370:~> which mpicc
>>>> /home/reuti/local/openmpi-1.4.4-thread/bin/mpicc
>>>> reuti_at_pc15370:~> cat /home/reuti/local/openmpi-1.4.4-thread/etc/openmpi-default-hostfile
>>>> pc15370 slots=2
>>>> pc15381 slots=2
>>>> reuti_at_pc15370:~> mpicc -o mpihello mpihello.c
>>>> reuti_at_pc15370:~> mpiexec -np 4 ./mpihello
>>>> Hello World from Node 0.
>>>> Hello World from Node 1.
>>>> Hello World from Node 2.
>>>> Hello World from Node 3.
>>>>
>>>> But all is local (no spawn here, traditional mpihello):
>>>>
>>>> 19503 ? Ss 0:00 /usr/sbin/sshd -o PidFile=/var/run/sshd.init.pid
>>>> 11583 ? Ss 0:00 \_ sshd: reuti [priv]
>>>> 11585 ? S 0:00 | \_ sshd: reuti_at_pts/6
>>>> 11587 pts/6 Ss 0:00 | \_ -bash
>>>> 13470 pts/6 S+ 0:00 | \_ mpiexec -np 4 ./mpihello
>>>> 13471 pts/6 R+ 0:00 | \_ ./mpihello
>>>> 13472 pts/6 R+ 0:00 | \_ ./mpihello
>>>> 13473 pts/6 R+ 0:00 | \_ ./mpihello
>>>> 13474 pts/6 R+ 0:00 | \_ ./mpihello
>>>>
>>>> -- Reuti
>>>>
>>>>
>>>>>>> We probably aren't correctly marking the original singleton on that node, and so the mapper thinks there are still two slots available on the original node.
>>>>>>
>>>>>> Okay. There is something to discuss/fix. BTW: if started as singleton I get an error at the end with the program the OP provided:
>>>>>>
>>>>>> [pc15381:25502] [[12435,0],1] routed:binomial: Connection to lifeline [[12435,0],0] lost
>>>>>
>>>>> Okay, I'll take a look at it - but it may take awhile before I can address either issue as other priorities loom.
>>>>>
>>>>>>
>>>>>> It's not the case if run by mpiexec.
>>>>>>
>>>>>> -- Reuti
>>>>>> _______________________________________________
>>>>>> users mailing list
>>>>>> users_at_[hidden]
>>>>>> http://www.open-mpi.org/mailman/listinfo.cgi/users
>>>>>
>>>>> _______________________________________________
>>>>> users mailing list
>>>>> users_at_[hidden]
>>>>> http://www.open-mpi.org/mailman/listinfo.cgi/users
>>>>
>>>>
>>>> _______________________________________________
>>>> users mailing list
>>>> users_at_[hidden]
>>>> http://www.open-mpi.org/mailman/listinfo.cgi/users
>>>
>>>
>>> _______________________________________________
>>> users mailing list
>>> users_at_[hidden]
>>> http://www.open-mpi.org/mailman/listinfo.cgi/users
>>
>>
>> _______________________________________________
>> users mailing list
>> users_at_[hidden]
>> http://www.open-mpi.org/mailman/listinfo.cgi/users
>