[Gmsh] Error due to many volumes
quey at emse.fr
Mon Jun 16 09:10:26 CEST 2008
Christophe - Yes, it only happens with Netgen.
Quoting Christophe Geuzaine <cgeuzaine at ulg.ac.be>:
> Werner, Romain - I think I might understand where the problem comes
> from: does it only happen with the Netgen algorithm?
> magpar wrote:
>> Dear Romain,
>> Dear Gmsh developers and users,
>> I can confirm Romain's problem and his solution:
>>>> Romain Quey wrote:
>>>>> Dear Gmsh developers, dear all,
>>>>> Using gmsh 2.2.0 on Linux, I'm having the following error:
>>>>> $ gmsh -3 n1100-id1.geo
>>>>> Info : Parsing file 'n1100-id1.geo'
>>>>> Error : [on processor 0] Unable to open file 'n1100-id1.msh'
>>>>> My geo file contains many 'Volume' definitions (1100) -- see the
>>>>> attached example.
>>> You are right, the bug does not occur here, but when writing the mesh.
>>> However, I think the reason for the bug is really this one: when
>>> increasing the max number of openable files per process on my system
>>> (from 1024 to 1000000), I can successfully mesh the file. I think that
>>> if you run 'ulimit -n 1024' on your system, gmsh will crash too.
>> I observed similar problems with models with many volumes a while
>> ago and reported them on the gmsh mailing list:
>> Based on Romain's observation I tested his solution to increase the
>> resources (max. number of open files) on the shell and I can
>> confirm that this fixes my problems, too.
>> Here is an example (from one of my earlier posts) which exhibits
>> the problem (mesh generation is successful, but the msh file cannot
>> be saved):
>> To make this increase in the resources permanent one has to modify
>> the limits in /etc/security/limits.conf by adding or modifying the
>> * hard nofile 10000
>> After that it is necessary to login on a new shell or even reboot
>> the machine because new processes inherit this setting from the
>> process/shell which launches it.
>> So, finally the question is why gmsh keeps so many files open (one
>> for each volume!?) and how this could be fixed.
>> Thanks for the useful discussion and help on this mailing list
>> which fixed my problem, too!
> Prof. Christophe Geuzaine
> University of Liege, Electrical Engineering and Computer Science
Romain QUEY, Doctorant (bureau H3-05)
Ecole des Mines de Saint-Etienne
Laboratoire PECM - UMR CNRS 5146
158 cours Fauriel
42023 Saint-Etienne, cedex 2
Tel. 04.77.49.97.38 - Fax. 04.77.42.66.78
E-mail : quey at emse.fr
This message was sent using IMP, the Internet Messaging Program.
More information about the gmsh