View Issue Details

IDProjectCategoryView StatusLast Update
0002821OpenFOAMBugpublic2018-02-02 17:25
Reporterlars Assigned Tohenry  
PriorityhighSeverityfeatureReproducibilityalways
Status closedResolutionno change required 
PlatformGNU/LinuxOSUbuntuOS Version16.04
Summary0002821: foamJob stops without internet connection
DescriptionI realised that my simulation stops invariably when I execute a parallel descomposition of the case, I run it, and I unplug the ethernet Lan wire.

After a short while (about 2 min) the running stops.

NO error message, no notification. It stops as if I 'ctrl + c'.


Steps To ReproduceI execute:

foamJob -s -p rhoCentralFoam (with 8 cores running)

Then I unplug the ETHERNET LAN wire (pc loses internet connexion)

After short time (2 min) simulation stops


Additional InformationSo I assume that foamJob uses, for some reason, the internet connexion, and If disconnected, It stops.

I need this issue solved because I work in a place where the internet get down sometimes during the weekend, and so I get stopped my simulation trhough the weekend.
TagsNo tags attached.

Activities

henry

2018-01-29 09:09

manager   ~0009226

Try running with nohup

lars

2018-01-29 09:26

reporter   ~0009227

I did it. No changes at all. It stops anyway

henry

2018-01-29 09:41

manager   ~0009228

Is the issue with using foamJob? Try running without foamJob and use nohup and redirect input and output.

Can you reproduce the problem with other non OpenFOAM batch applications?

lars

2018-01-29 09:59

reporter   ~0009229

I think the issue is by using mpirun In the end foamJob uses mpirun
ordres comandes.png (445,181 bytes)

wyldckat

2018-01-29 10:51

updater   ~0009230

If Open-MPI is being used, then it's because it uses greedy connections by default and if the relevant connection is dropped, it gets utterly lost without it.

There is an option for Open-MPI's 'mpirun' for restricting to only using the local memory system, if a single machine is used:

  mpirun ... --mca btl self ...


A few more details are available on Open-MPI's FAQ pages, for example:

  - On TCP: https://www.open-mpi.org/faq/?category=tcp

  - On Shared Memory: https://www.open-mpi.org/faq/?category=sm

  - On how to set-up MCA parameters: https://www.open-mpi.org/faq/?category=tuning#setting-mca-params


In the last link, it tells how to set environment variables for the relevant MCA entry... if I'm not mistaken, the option above can be done automatically by running the following command before 'foamJob' or 'mpirun':

 export OMPI_MCA_btl=self



I don't know if it's a good idea to add this to the OpenFOAM User Guide, but at least pointing people to the Open-MPI FAQ would be nice ;)

lars

2018-02-02 07:31

reporter   ~0009239

Dear Bruno, thank you for your detailed answer.

I have tried and added a last command in the bashrc :

'
export OMPI_MCA_btl=self
'

But, nothing changes. It still gets down after 1 minute once internet is shutted down.

To me I find a huge mistake that, by default, all the simulation depends on whether there is internet or not. I appreciate it as a bug.

henry

2018-02-02 10:57

manager   ~0009240

Have you tested running serial? It appears that the problem is with OpenMPI and not OpenFOAM and running serial will check this.

If the problem is with OpenMPI you should report it to the maintainers of that package or you could try one of the alternative MPI implementations.

henry

2018-02-02 17:25

manager   ~0009249

This issue relates to the operation of OpenMPI rather than OpenFOAM and should be reported to the maintainers of OpenMPI.

Issue History

Date Modified Username Field Change
2018-01-29 09:04 lars New Issue
2018-01-29 09:09 henry Note Added: 0009226
2018-01-29 09:26 lars Note Added: 0009227
2018-01-29 09:41 henry Note Added: 0009228
2018-01-29 09:59 lars File Added: ordres comandes.png
2018-01-29 09:59 lars Note Added: 0009229
2018-01-29 10:51 wyldckat Note Added: 0009230
2018-02-02 07:31 lars Note Added: 0009239
2018-02-02 10:57 henry Note Added: 0009240
2018-02-02 17:25 henry Assigned To => henry
2018-02-02 17:25 henry Status new => closed
2018-02-02 17:25 henry Resolution open => no change required
2018-02-02 17:25 henry Note Added: 0009249