[Bob]

Dear,

I am trying to mesh the void of a box, in which a sphere resides, with snappyHexMesh (OF1.5). After a couple of interations with the castellatedMeshControls the process is killed without any message (see below). The result is the same when I change the parameters. Does someone has experienced this too? And how to solve this problem?

The parameters used for the castellatedMeshControls are given below as well...


Many thanks!

Regards,
Bob

....

Shell refinement iteration 2
----------------------------

Marked for refinement due to refinement shells : 840 cells.
Determined cells to refine in = 0.7 s
Selected for internal refinement : 1192 cells (out of 328040)
Edge intersection testing:
Number of edges : 1029440
Number of edges to retest : 90128
Number of intersected edges : 0
Refined mesh in = 5.51 s
After refinement shell refinement iteration 2 : cells:336384 faces:1029440 points:356997
Cells per refinement level:
0 58448
1 11056
2 266880

Shell refinement iteration 3
----------------------------

Marked for refinement due to refinement shells : 0 cells.
Determined cells to refine in = 0.69 s
Selected for internal refinement : 224 cells (out of 336384)
Killed

Used parameters:

geometry
{
ball.stl
{
type triSurfaceMesh;
name ball;
}
};


castellatedMeshControls
{
maxLocalCells 1000000;
maxGlobalCells 2000000;
minRefinementCells 10;
nCellsBetweenLevels 2;

features
(
);

refinementSurfaces
{
}

resolveFeatureAngle 30;


refinementRegions
{
ball
{
mode distance;
levels ((2 2));
}
}


locationInMesh (0.1 0.1 2.6);
}

[wikstrom]

Seems to me that your are out of memory and the system kills the memory consuming process?

Secondly, use OF-1.5.x if you want to run snappyHexMesh.

/Niklas

[0.1 watts]

Hey, I also have the problem of SHM being consistently killed quite close to writing the final iteration, unless I specify a very small mesh. I monitored the memory usage and it wasn't that which caused it. Is it possible that the system is terminating the process for overheat protection - my CPU peaked at 90C? It would be odd if it is because I have previously used SHM with a core temperature peaking at 97C without any problems. Of course this is just speculation so if anyone has a better suggestion I'd be really grateful.

[wyldckat]

Greetings 0.1 watts and welcome to the forum!

OK, something is not right there... how is your machine still alive? CPUs aren't meant to run at temperatures higher than 80ºC! GPUs can handle the stress up to 90-100ºC, but CPUs will start getting severe damage when running above 80-85ºC!!

It's very possible that your CPU has been damaged already from running at such high temperatures!
I suggest that you run stressapptest on your machine to verify if there are any specific problems going on! On Ubuntu you can install it directly with the apt-get package manager:

sudo apt-get install stressapptest

On Fedora and openSUSE it shouldn't be much more different. The project page is here: http://code.google.com/p/stressapptest/

To do a good test, run it like this:

stressapptest -W --cc_test

It'll run for 20 seconds and if anything wrong happens, it will report to you something like "memory failure" or something like that. This program has been known to detect problems that memtest86+ can't detect, simply because sometimes the problems only occur when data is being transported in certain ways between CPU and RAM.

If by any chance it doesn't report any failures... then buy a new and (very) good CPU cooler for your machine before anything bad does happen, because it deserves a proper and nice cooling solution!

Best regards,
Bruno

[0.1 watts]

Originally Posted by wyldckat

Greetings 0.1 watts and welcome to the forum!

OK, something is not right there... how is your machine still alive? CPUs aren't meant to run at temperatures higher than 80ºC! GPUs can handle the stress up to 90-100ºC, but CPUs will start getting severe damage when running above 80-85ºC!!

It's very possible that your CPU has been damaged already from running at such high temperatures!
I suggest that you run stressapptest on your machine to verify if there are any specific problems going on! On Ubuntu you can install it directly with the apt-get package manager:

sudo apt-get install stressapptest

On Fedora and openSUSE it shouldn't be much more different. The project page is here: http://code.google.com/p/stressapptest/

To do a good test, run it like this:

stressapptest -W --cc_test

It'll run for 20 seconds and if anything wrong happens, it will report to you something like "memory failure" or something like that. This program has been known to detect problems that memtest86+ can't detect, simply because sometimes the problems only occur when data is being transported in certain ways between CPU and RAM.

If by any chance it doesn't report any failures... then buy a new and (very) good CPU cooler for your machine before anything bad does happen, because it deserves a proper and nice cooling solution!

Best regards,
Bruno

Hi Bruno, thanks for your welcome and assistance!

This explanation would fit in with my problems, although my laptop has passed some high load checks. I'll try meshing the same case tomorrow on the cluster at work and see if the error re-occurs. If I get problems there I'll update this thread.

Thanks for the tip on my CPU temperature; I broke my own rules by letting it get above 87! You've spurred me on to apply some new thermal paste this weekend, my laptop is in a bad way with overheating (it actually had a cooling pad at the time it approached 100C).

Incidentally I hear mixed messages about safe temperatures in notebooks; some say that any temperature up to the manufacturers Tj_max won't cause damage, and the safe temperature is lower in desktops. Anyone have some opinions on this?