|
[Sponsors] |
February 6, 2007, 17:25 |
Importan Problem In Parallel Solver
|
#1 |
Guest
Posts: n/a
|
I cannot wonder the solution, and the people of ANSYS Support was not able to give me a solution after 10 mails.
When i run solver CFX 10.0 in parallel mode (boths MPICH or PVM), the proccesors freeze after 20 - 30 - 40 - 50 iterations. The RAM is 100% working and full. The simulation is medium - large size. All posible solution will be welcome, and would help me very much. If someone need more information to help me, please, ask me. My configuration is 4 machines Intel 64 bit Dual Core, with 8GB RAM each. I use Gbit Net to comunicate them. I have a Red hat Linux Enterprise Ed. in all machines. Thank you very much in advance. I look forward for a solution. |
|
February 7, 2007, 17:41 |
Re: Importan Problem In Parallel Solver
|
#2 |
Guest
Posts: n/a
|
Hi,
Does the machine just lock up or does it give an error message? Glenn Horrocks |
|
February 8, 2007, 03:55 |
Re: Importan Problem In Parallel Solver
|
#3 |
Guest
Posts: n/a
|
Hello Glenn,
only the proccessors stop, the RAM is completely operative, and i can run another programs, but CFX stops. Regards |
|
February 8, 2007, 09:23 |
Re: Importan Problem In Parallel Solver
|
#4 |
Guest
Posts: n/a
|
Couple of questions to you: 1. Do you have any problem to run your case with local parallel? 2. At the time processor is frozen, what is code doing? Is it writing backup file, during iteration, or at end of one iteration? 3. Have you sent your case to tech support so that they can try on their machine? You need to know whether your problem is machine-dependent or not.
|
|
February 8, 2007, 12:47 |
Re: Importan Problem In Parallel Solver
|
#5 |
Guest
Posts: n/a
|
Hello longbow,
1. In local parallel there is no problem, 2. Is during an iteration. And it doesn't depends of the phase of proccess. Sometimes in 10th iteration, sometimes in 60th iteration, in equals simulations 3. Yes, i sent the problem and the simulation that failed. They were not able to help me, and said me are studying the problem...... |
|
February 8, 2007, 17:41 |
Re: Importan Problem In Parallel Solver
|
#6 |
Guest
Posts: n/a
|
Hi,
How reliable is your network? If you network has glitches or is occasionally heavily loaded from another user this could cause havoc to a distributed CFX run. Glenn Horrocks |
|
February 8, 2007, 18:56 |
Re: Importan Problem In Parallel Solver
|
#7 |
Guest
Posts: n/a
|
I don't think in the net could be the problem. I have a GBit connection, and on weekends and nights, when the local is empty, fall down too.
Could be something wrong with software, in memory allocation for instance? Thanks, I look forward for more possible suggestions. |
|
February 9, 2007, 18:48 |
Re: Importan Problem In Parallel Solver
|
#8 |
Guest
Posts: n/a
|
Have seen somthing like this before when I had an expression with a reference to boundary. Something like:
InletVelocity = areaAve(velocity)@outlet Do you have something like that? If so, then check if each partition has a part of the boundary you are refering to, outlet in my example. I had the problem resolved by defining a user difined direction for the partitioner, such that my outlet was present in each of my four partitions. Another option can be to define a different expression.. Regs, Astrid |
|
February 10, 2007, 00:52 |
Re: Importan Problem In Parallel Solver
|
#9 |
Guest
Posts: n/a
|
Hi
Astrid's comment seems to refer to a bug. Astrid, have you reported this to CFX support as a bug? Or even better, have you tested it on CFX11 and see if it still does it? If it is still there in CFX11 then definitely report it to CFX supprot. Then the problem can be fixed for all of us. Inigo: Some networks have funny moments overnight as IT people do system backups and maintenance at night. Don't assume that because you are at home the network is working! Your problem will take some fault finding to isolate I suspect. If you can't isolate the fault then call an IT person in to check if they can find a network problem causing your error. Glenn Horrocks |
|
February 10, 2007, 16:52 |
Re: Importan Problem In Parallel Solver
|
#10 |
Guest
Posts: n/a
|
Get cfx 11 and try it again. There was a problem fixed late in 11.0 which might cause the behaviour you observe.
|
|
February 12, 2007, 19:14 |
Re: Importan Problem In Parallel Solver
|
#11 |
Guest
Posts: n/a
|
Yep, what I mentioned is recognized as a bug in CFX-10. Don't know if it is still present in CFX-11. Haven't tried it.
Regs, Astrid |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
problem of running parallel Fluent on linux cluster | ivanbuz | FLUENT | 15 | September 23, 2017 20:12 |
Solver problem in Oscillating Plate tutorial | vovogoal | CFX | 1 | November 22, 2011 10:54 |
Working directory via command line | Luiz | CFX | 4 | March 6, 2011 21:02 |
Divergence detected when using parallel solver | Phil | FLUENT | 3 | May 9, 2008 09:17 |
Parallel solver, divergence | Phil | FLUENT | 0 | April 30, 2008 10:42 |