View Issue Details

IDProjectCategoryView StatusLast Update
0000401OpenFOAMBugpublic2012-02-01 17:53
Reporteruser337Assigned Touser4 
PriorityhighSeveritycrashReproducibilityalways
Status resolvedResolutionfixed 
PlatformCore i7-2600K and Q9550OSKubuntu 11.10OS Version64Bit
Summary0000401: Floating Object tutorial crashes when run in parallel
DescriptionThe floating object tutorial crashes, when runned in parallel. This happens no matter which decomposition method is used. (simple scotch hierarchical).
In single core it completes. Is reproducable on two different machines. The cfd-online-forum-user wyldckat was also able to reproduce the problem.
The old workaround for OpenFOAM 1.7.x doesn't helps. ("There is a problem with the decomposition of point fields (i.e. pointDisplacement).Decompose the mesh before copying 0.org to 0 in each processor directory, so the pointDisplacement field is not operated on by the decomposition").


Forum-Link: http://www.cfd-online.com/Forums/openfoam-solving/96641-floating-object-tutorial.html#post341572
Steps To ReproducerunApplication blockMesh
runApplication topoSet
runApplication subsetMesh -overwrite c0 -patch floatingObject
cp -r 0.org 0 > /dev/null 2>&1
runApplication setFields
runApplication decomposePar
runApplication foamJob -screen -parallel interDyMFoam
runApplication reconstructPar

This is the allrun-script coming with the case, with added decomposition. (for 4 cores in decomposeParDict, method scotch)
Additional InformationThe interDyMFoam log is appended.
This crash also occurs for different interDyMFoam cases with floating objects. (User made)
TagsNo tags attached.

Activities

user337

2012-01-28 22:53

 

log.foamJob (696,042 bytes)

user337

2012-01-28 22:53

 

Allrun.txt (512 bytes)   
#!/bin/sh
cd ${0%/*} || exit 1    # run from this directory

# Source tutorial run functions
. $WM_PROJECT_DIR/bin/tools/RunFunctions


runApplication blockMesh
runApplication topoSet
runApplication subsetMesh -overwrite c0 -patch floatingObject
cp -r 0.org 0 > /dev/null 2>&1
runApplication setFields
runApplication decomposePar
runApplication foamJob -screen -parallel interDyMFoam
runApplication reconstructPar

# ----------------------------------------------------------------- end-of-file
Allrun.txt (512 bytes)   

user337

2012-01-28 23:08

  ~0000968

The workaround mentioned has the ID 0000115

user4

2012-02-01 15:50

  ~0000988

Fixed in 0fe28f9ed9cd87a0def41b847f2503457febac91.

This is quite a tricky issue - was a boundary condition triggering processor communication. Could you re-open if it still gives you problems?

Thanks for reporting,

Mattijs

Issue History

Date Modified Username Field Change
2012-01-28 22:53 user337 New Issue
2012-01-28 22:53 user337 File Added: log.foamJob
2012-01-28 22:53 user337 File Added: Allrun.txt
2012-01-28 22:54 user337 Tag Attached: Added the used Allrun script
2012-01-28 22:54 user337 Tag Detached: Added the used Allrun script
2012-01-28 23:08 user337 Note Added: 0000968
2012-02-01 15:50 user4 Note Added: 0000988
2012-02-01 15:50 user4 Status new => resolved
2012-02-01 15:50 user4 Fixed in Version => 2.1.x
2012-02-01 15:50 user4 Resolution open => fixed
2012-02-01 15:50 user4 Assigned To => user4