Hi!

We are two students working on a wind-turbine cfd project. The mesh is about 1.5 Mio Hexas and we are running distributed parallel on two Windows XP machines (PVM). The Solver quits randomly (sometimes 3 iterations, sometimes 253 iterations) with Return Code 15.

Here are the last lines from the solver output:

OUTER LOOP ITERATION = 187 CPU SECONDS = 4.81E+04

----------------------------------------------------------------------

| Equation | Rate | RMS Res | Max Res | Linear Solution |

+----------------------+------+---------+---------+------------------+

| U-Mom | 0.99 | 2.9E-03 | 3.5E-02 | 3.0E-04 OK|

| V-Mom | 1.00 | 3.6E-03 | 1.2E-01 | 2.3E-04 OK|

| W-Mom | 1.00 | 3.1E-03 | 9.1E-02 | 2.5E-04 OK|

| P-Mass | 0.99 | 2.2E-05 | 1.7E-03 | 9.5 3.5E-02 OK|

+----------------------+------+---------+---------+------------------+

| K-TurbKE | 0.98 | 1.3E-02 | 3.2E-01 | 5.8 7.7E-08 OK|

| E-Diss.K | 0.97 | 4.6E-03 | 2.9E-01 | 11.7 3.5E-08 OK

+----------------------+------+---------+---------+------------------+

================================================== ====================

OUTER LOOP ITERATION = 188 CPU SECONDS = 4.84E+04

----------------------------------------------------------------------

| Equation | Rate | RMS Res | Max Res | Linear Solution |

+----------------------+------+---------+---------+------------------+

+--------------------------------------------------------------------+

| An error has occurred in cfx5solve: |

| |

| The CFX-5 solver exited with return code 15. No results file has |

| been created.

|

+--------------------------------------------------------------------+

End of solution stage.

Does anyone know what this means??

Kind Regards.

Hi,

Looks strange. Are you sure you have not run out of disk space, some other process not gobbled up a heap of memory and your network is reliable? I assume there are not signs of the simulation diverging.

Also you will probably find the MPICH distributed parallel much faster than PVM.

Regards, Glenn Horrocks

Hi,

Thanks for the reply. Yes it could be a network problem. We have an old bnc ring network her at the university, which seems very instable. Any clue what network problem it could be? You recommend to run with MPICH? I heard its not as stable as PVM or am i wrong. Do i have to make specific settings or to install something to use MPICH (besides choosing it in solver control)?

Thanks in advance.

Regards Kramer

Hi,

If your network is dodgy then forget running CFX in parallel until you fix it. Parallel simulation needs a reliable network to function.

PVM is more stable than MPICH, but on a decent network with a good model it is fine. As you discovered, even PVM cannot work properly if the network has poor reliablility.

Installation of MPICH is described in the documentation.

regards, Glenn Horrocks