[mpich-discuss] Volatile nodes

[Hiatt, Dave M]

Thank you Dorian, I will then have to make a monitor process on node 0 to handle incoming and outgoing requests and find some other way to move message traffic in and out of node 0 to the workstations that are requesting the runs.

-----Original Message-----
From: mpich-discuss-bounces at mcs.anl.gov
[mailto:mpich-discuss-bounces at mcs.anl.gov]On Behalf Of Dorian Domenic
Krause
Sent: Thursday, December 04, 2008 9:54 AM
To: mpich-discuss at mcs.anl.gov
Subject: Re: [mpich-discuss] Volatile nodes


On Thu, 4 Dec 2008 10:24:07 -0500
  "Hiatt, Dave M " <dave.m.hiatt at citi.com> wrote:
> No, I was thinking more of a case where a node does 
>MPI::Init, communicates information, does an 
>MPI::Finalize, then returns later with another MPI::Init 
>and MPI::Finalize pair on and off.  The node would be 
>part of a communicator that isn't part of the primary 
>calculation but more on the idea of a monitor.


I can't say whether MPICH supports this kind of 
application but as far as I know the standard says that 
the only function safely callable after MPI_Finalize is 
MPI_Finalized. In this respect you would at least get a 
non-portable program.

Dorian


> 
> -----Original Message-----
>From: mpich-discuss-bounces at mcs.anl.gov
> [mailto:mpich-discuss-bounces at mcs.anl.gov]On Behalf Of 
>Rajeev Thakur
> Sent: Wednesday, December 03, 2008 7:44 PM
> To: mpich-discuss at mcs.anl.gov
> Subject: Re: [mpich-discuss] Volatile nodes
> 
> 
> Do you mean what happens if one of the nodes crashes? 
>The current version of
> MPICH2 will abort in that case, but within the next year 
>we plan to have a
> version that is more resilient to such failure; i.e., it 
>won't abort, but
> any attempt to communicate with that node will return an 
>error.
> 
> Rajeev
> 
> 
>> -----Original Message-----
>> From: mpich-discuss-bounces at mcs.anl.gov 
>> [mailto:mpich-discuss-bounces at mcs.anl.gov] On Behalf Of 
>>Hiatt, Dave M 
>> Sent: Wednesday, December 03, 2008 6:06 PM
>> To: mpich-discuss at mcs.anl.gov
>> Subject: [mpich-discuss] Volatile nodes
>> 
>> Is there any problem with a node being volatile in 
>>MPICH2
>> 
>> If you lived here you'd be home by now
>> Dave Hiatt
>> Manager, Market Risk Systems Integration CitiMortgage, 
>>Inc.
>> 1000 Technology Dr.
>> Third Floor East, M.S. 55
>> O'Fallon, MO 63368-2240
>> 
>> Phone:	636-261-1408
>> Mobile:	314-452-9165
>> FAX:	636-261-1312
>> Email:     Dave.M.Hiatt at citigroup.com
>> 
>> 
>> 
>>