View Issue Details

IDProjectCategoryView StatusLast Update
0001777OpenFOAMBugpublic2015-07-19 20:58
Reporteruser667Assigned Tohenry  
PrioritynormalSeverityminorReproducibilityalways
Status closedResolutionunable to reproduce 
PlatformGNU/LinuxOSUbuntuOS Version14.04
Summary0001777: PimpleDyMFoam / memory usage increases continously
DescriptionThe used memory increases continously with pimpleDyMFoam. If you let it run long enough, it goes up to the maximum installed RAM and ends up with a crash eg.:
"mpirun noticed that process rank 0 with PID 4513 on node CAEresNd1 exited on signal 9 (Killed)."
Could you please look over it. Many thanks in advance!
Steps To ReproduceRun the propeller tutorial for a longer solution endtime and check the allocated memory in the system monitor. The memory usage increases faster, if you enlarge the model mesh size and number of partitions.
Additional InformationI have another machine with CentOS 7.1 and OF 2.4.x. It's the same thing there.
TagsNo tags attached.

Activities

henry

2015-07-10 22:15

manager   ~0005067

Do you see the memory leak only when running in parallel or also serial?

henry

2015-07-12 13:04

manager   ~0005070

I have been tracing through the operation of this tutorial case and the memory leak relates to the surfaces functionObject on a moving mesh. Both the cuttingPlane and isoSurface leak memory. If you run without these functionObjects the case should run without memory problems.

It will take a bit of time working through these complex functionObjects to find and solve the memory leak.

user667

2015-07-13 08:40

  ~0005075

i ran the propeller case again without the functionObjects, but i still see the memory leak. I didn't check in serial mode so far.

user4

2015-07-13 09:07

  ~0005076

Could be related to #1487. They (sampledIsoSurface, sampledCuttingPlane) use (through volPointInterpolation) a cached version of the pointField which is fine except isoSurface holds a reference to it. So it all works as long as the pointField on the database does not get deleted.

henry

2015-07-13 11:25

manager   ~0005077

I am unable to reproduce the memory leak without functionObjects either in serial or parallel. My understanding is that the memory leak is caused by the cuttingPlane and isoSurface functionObjects which may relate to #1487 as Mattijs suggests. Either way it is likely that these functionObjects will need substantial re-development so solve this and other issues.

user667

2015-07-14 07:55

  ~0005083

Now i ran the propeller case in serial without functionObjects overnight. When i started, the system monitor showed me 1.7GB (5.5%) allocated memory. After 0.3 sec solution time this morning it said 2.6GB (8.2%).
If i understand you right, the functionObject "forces" shouldn't have a memleak?

henry

2015-07-14 08:05

manager   ~0005084

I don't believe the forces functionObject has a memory leak but I didn't check for this case, I simply removed all the functionObjects. If you ran without any functionObjects then this should not be an issue anyway.

I have not been able to reproduce the memory leak without functionObjects.

user667

2015-07-14 17:35

  ~0005085

One last question. Do you have any idea, why the memory usage increases in my case even without the functionObjects? Is there anything i could check?

I ran the propeller case again with a finer backround mesh (18 30 18), which leads to about 1.5 Mio cells. This is roughly the size of my actual model. After the solver start in parallel with 16 partitions (2 4 2), it uses 3.9GB and after 0.1 sec solution time it uses already 11.3GB. So it rises quickly.
Is there anything i could do or check?

henry

2015-07-14 17:49

manager   ~0005086

You could try running your case in OpenFOAM-dev.

I have tested for the memory leak in the propeller case in OpenFOAM-2.4.x and OpenFOAM-dev and see no problem with either without functionObjects.

Issue History

Date Modified Username Field Change
2015-07-07 20:31 user667 New Issue
2015-07-10 22:15 henry Note Added: 0005067
2015-07-12 13:04 henry Note Added: 0005070
2015-07-13 08:40 user667 Note Added: 0005075
2015-07-13 09:07 user4 Note Added: 0005076
2015-07-13 11:25 henry Note Added: 0005077
2015-07-14 07:55 user667 Note Added: 0005083
2015-07-14 08:05 henry Note Added: 0005084
2015-07-14 17:35 user667 Note Added: 0005085
2015-07-14 17:49 henry Note Added: 0005086
2015-07-19 20:58 henry Status new => closed
2015-07-19 20:58 henry Assigned To => henry
2015-07-19 20:58 henry Resolution open => unable to reproduce