[mpich-discuss] Fwd: Can't receive messages

Rajeev Thakur thakur at mcs.anl.gov
Mon Dec 30 08:23:39 CST 2013


There must be a firewall issue. A process on one machine cannot open a TCP connection on the other machine. Only known services like ssh are working.


On Dec 30, 2013, at 6:43 AM, Matthias Neuer <mneuer at web.de> wrote:

> On 12/30/2013 12:12 PM, Rajeev Thakur wrote:
>> There may be some other connectivity issue between the machines. Does the cpi example from the MPICH examples directory run across the two machines?
> 
> The cpi-example hangs too. Output:
> 
> Process 1 of 4 is on matze-debian
> Process 3 of 4 is on matze-debian
> Process 2 of 4 is on notebook
> Process 0 of 4 is on notebook
> 
> Then it hangs.
> Seems like a configuration issue but why are the printf messages transfered correctly?
> 
> Matthias
> 
>> On Dec 30, 2013, at 3:58 AM, Matthias Neuer <mneuer at web.de> wrote:
>> 
>>> Hi.
>>> 
>>> Unfortunately, this does not solve the problem.
>>> 
>>> Maybe I should have written that the program hangs after the output in my last post. It waits until MPI_Recv returns which does not happen.
>>> 
>>> By the way I can login from one machine to the other using ssh, no problem.
>>> 
>>> Thanks
>>> 
>>> On 12/30/2013 06:39 AM, Rajeev Thakur wrote:
>>>> Try adding an fflush(stdout) after the printf after the recv.
>>>> 
>>>> On Dec 29, 2013, at 12:50 PM, Matthias Neuer <mneuer at web.de> wrote:
>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> -------- Original Message --------
>>>>> Subject: Can't receive messages
>>>>> Date: Sun, 29 Dec 2013 19:00:05 +0100
>>>>> From: Matthias Neuer <mneuer at web.de>
>>>>> To: discuss at mpich.org
>>>>> 
>>>>> Hi.
>>>>> 
>>>>> I installed MPICH on 2 machines (called notebook and matze-debian) using
>>>>> the same sourcecode. The output of mpichversion is both times:
>>>>> 
>>>>> MPICH Version:          3.0.4
>>>>> MPICH Release date:     Wed Apr 24 10:08:10 CDT 2013
>>>>> MPICH Device:           ch3:nemesis
>>>>> MPICH configure:        --disable-f77 --disable-fc
>>>>> --prefix=/home/matze/mpich-install
>>>>> MPICH CC:       cc    -O2
>>>>> MPICH CXX:      c++   -O2
>>>>> MPICH F77:      no
>>>>> MPICH FC:       no
>>>>> 
>>>>> I wrote a small test program which I send as an attachment.
>>>>> On execution on the machine matze-debian I get the following:
>>>>> 
>>>>> Process 0 is on matze-debian
>>>>> Process 2 is on matze-debian
>>>>> Process 1 is on notebook
>>>>> received: Hello, from number 2 on matze-debian
>>>>> Process 2 has sent the message
>>>>> Process 1 has sent the message
>>>>> 
>>>>> I don't receive the message from process number 1, but the output
>>>>> suggests that the message was succesfully sent. Actually I checked the
>>>>> network traffic using wireshark and matze-debian received a package in
>>>>> which I found the sent message. So somehow the program does not
>>>>> recognize the arrival of the message.
>>>>> 
>>>>> When I run the program on a single machine it works.
>>>>> 
>>>>> I don't know if this is a programming error or a configuration error.
>>>>> 
>>>>> Thanks for your help
>>>>> 
>>>>> 
>>>>> 
>>>>> <ex1.c>_______________________________________________
>>>>> discuss mailing list     discuss at mpich.org
>>>>> To manage subscription options or unsubscribe:
>>>>> https://lists.mpich.org/mailman/listinfo/discuss
>>>> 
>>>> _______________________________________________
>>>> discuss mailing list     discuss at mpich.org
>>>> To manage subscription options or unsubscribe:
>>>> https://lists.mpich.org/mailman/listinfo/discuss
>>>> 
>>> 
>>> _______________________________________________
>>> discuss mailing list     discuss at mpich.org
>>> To manage subscription options or unsubscribe:
>>> https://lists.mpich.org/mailman/listinfo/discuss
>> 
>> _______________________________________________
>> discuss mailing list     discuss at mpich.org
>> To manage subscription options or unsubscribe:
>> https://lists.mpich.org/mailman/listinfo/discuss
>> 
> 
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss




More information about the discuss mailing list