[imagineaspen]

Hi all,

I am trying to run a simulation on a Linux cluster but getting this error immediately after starting the simulation (also attached the output file):

Details of error:-
----------------
Error detected by routine PEEKI
CDANAM = /FLOW/GEOMETRY/KZnBcp
CRESLT = ADRS

Current Directory : /FLOW/PHYSICS

+--------------------------------------------------------------------+
ERROR #001100279 has occurred in subroutine ErrAction.
Message:
Stopped in routine MEMERR

An error has occurred in cfx5solve:
The ANSYS CFX partitioner exited with return code 1.


Does anyone know what this error means? It seems to have something to do with memory? I have tried increasing the topology estimate factor but I still get the same error. Also tried changing the partitioning method since it seems like the simulation fails during partitioning. The simulation is able to run on another computer, it is only when I run it on the cluster that this occurs.

[ghorrocks]

It is failing in the partitioning run, not the solver.

In these cases the first thing to try is to use a different partitioning algorithm. The default (METIS) does give memory problems occasionally. Try using recursive bisection or one of the other algorithms as they tend to have much less memory problems (but do not produce as good partitions in general).

[imagineaspen]

Thank you for the advice Glenn. I've tried different partitioning algorithms but unfortunately I am still getting the same error.

[ghorrocks]

Then it is possibel that you are running out of memory in the partitioner. Have a look at the memory settings in the partitioner, also try the large problem partitioner.

Alternately, you could partition this on a large computer with lots of memory and write the partitions to a partition file. Then you can use that partition file on this cluster so you do not have to do the partitioning at all on this cluster.

[Opaque]

Originally Posted by imagineaspen

Hi all,

I am trying to run a simulation on a Linux cluster but getting this error immediately after starting the simulation (also attached the output file):

Details of error:-
----------------
Error detected by routine PEEKI
CDANAM = /FLOW/GEOMETRY/KZnBcp
CRESLT = ADRS

Current Directory : /FLOW/PHYSICS

+--------------------------------------------------------------------+
ERROR #001100279 has occurred in subroutine ErrAction.
Message:
Stopped in routine MEMERR

An error has occurred in cfx5solve:
The ANSYS CFX partitioner exited with return code 1.


Does anyone know what this error means? It seems to have something to do with memory? I have tried increasing the topology estimate factor but I still get the same error. Also tried changing the partitioning method since it seems like the simulation fails during partitioning. The simulation is able to run on another computer, it is only when I run it on the cluster that this occurs.

something is off with your physics-to-mesh connectivity. ADRS is the error. The array KZnBcp is being accessed beyond its bounds, either too low, or too high.

[imagineaspen]

Originally Posted by ghorrocks

Then it is possibel that you are running out of memory in the partitioner. Have a look at the memory settings in the partitioner, also try the large problem partitioner.

Alternately, you could partition this on a large computer with lots of memory and write the partitions to a partition file. Then you can use that partition file on this cluster so you do not have to do the partitioning at all on this cluster.

Thanks Glenn. I tried modifying the partitioner memory settings but it all resulted in the same error. I also tried partitioning on a different computer and inputting that partitioning file to the cluster but then I get a memory error immediately when the solver starts. I think I will have to bring this issue my school’s technical support; the simulation is not large enough that memory should be an issue.

One thing I forgot to put in the initial post is that I ran a steady state simulation using the same mesh to generate an initial values file before running this simulation. The steady state simulation runs properly on the cluster, it is only the transient simulation that produces this error. My set-up includes a rotating domain which seems to be part of the issue; I was testing simpler geometries to see if I could replicate the error and I am able to run simulations with large numbers of elements but as soon as I have a rotating domain I start getting memory errors. I think this must be why the steady state simulation is able to run but the transient does not.

[imagineaspen]

Originally Posted by Opaque

something is off with your physics-to-mesh connectivity. ADRS is the error. The array KZnBcp is being accessed beyond its bounds, either too low, or too high.

Thanks Opaque. Could you clarify what you mean by physics-to-mesh connectivity? Does this imply that it is a problem with my mesh itself or with my CFX-Pre set-up?

[ghorrocks]

Opaque's comment is suggesting that some form of internal CFX error is occurring, and it is manifesting as a failure to find some data initially, but as a memory error later. I suspect this is correct, which would explain why my suggestions on the partitioner settings did not help.

In this case I would have a look at your setup and check it is OK. In particular look to see if you have missed something - is there an essential variable you have not defined; such as a material property or definition of the rotating domain.

[imagineaspen]

Thanks Glenn and Opaque. I will take another look at my set-up.