View Issue Details
|ID||Project||Category||View Status||Date Submitted||Last Update|
|0003817||OpenFOAM||Bug||public||2022-03-15 14:50||2022-03-15 14:50|
|Platform||GNU/Linux||OS||centOS||OS Version||(please specify)|
|Fixed in Version|
|Summary||0003817: 3D parallel simulation using cyclicAMI, decomposed using scotch; randomly runs or gives fatal error|
|Description||When using cyclicAMI patches in a rotational symmetric 3D case, it may happen that |
(a) the case runs fine in parallel (e.g. np=128, np=256), and
(b) the case fails to run with "FOAM FATAL ERROR: Unable to set sources and target faces" (e.g. np=32, np=192)
depending on the number of processors.
Serial simulation always runs fine, thus, there is not an error related to the cyclicAMI specification in polyMesh/boundary.
For decomposition, scotch method was used and no constraints are defined in order to avoid imbalanced mesh distribution over processors.
Case geometry is rather complex with sth. like 10-100M cells and a high number of cells connected to cyclicAMI patches. (The case should run on some 100s of processors.)
I don't know if that is an actual bug or if there is a workaround to avoid the random behaviour.
Anyway, the random behaviour in terms of at which no. of processors it still works and at which it throws an fatal error appears to be quite strange.
|Tags||No tags attached.|