Possible memory leak in recorder

[Deniz]

Hi everyone,

Does anyone experienced an "out of memory" issue while performing a series of sequential (not the parallel verions of OpenSees) transient analyses. I think this is a memory leak problem associated with recorder command.

I have recieved the error when I re-run the same frame model with the same modeling parameters: Geometry, materials, connectivity, loading, but with different recorder.

The former recording scheme was:

recorder EnvelopeElement -file $outputPathName/Element-$objectId.Deformation1.txt -ele $objectId section 1 deformation;
recorder EnvelopeElement -file $outputPathName/Element-$objectId.Force1.txt -ele $objectId section 1 force;
recorder EnvelopeElement -file $outputPathName/Element-$objectId.Deformation4.txt -ele $objectId section 4 deformation;
recorder EnvelopeElement -file $outputPathName/Element-$objectId.Force4.txt -ele $objectId section 4 force;

It was changed to:

recorder Element -file $outputPathName/Element-$objectId.Deformation1.txt -ele $objectId section 1 deformation;
recorder Element -file $outputPathName/Element-$objectId.Force1.txt -ele $objectId section 1 force;
recorder Element -file $outputPathName/Element-$objectId.Deformation4.txt -ele $objectId section 4 deformation;
recorder Element -file $outputPathName/Element-$objectId.Force4.txt -ele $objectId section 4 force;

Somewhere in the middle of the analyses (ie 70 out of 100 eq. records) I always get an "out of memory" error.

The interesting part is that I checked "Mem Usage" and "commit charge memory" from "windows task manager/Processes" and
"windows task manager/performace" tabs, respectively. While using "recorder element" "Mem usage" displays a fluctuating pattern with constantly increasing memory commits. As a result I get the error! When using "recorder EnvelopeElement" there is no such problem.

The model is a very simple 2 story 3 bay steel frame with nonlinearBeamColumn elements with 4 integration points. It performs transient analysis after the application of gravity loads (as the OpenSees manual guides). Then, new analysis starts after wipe.

I will be very happy if anyone has experienced similar problem and help me to solve this issue.

Thanks in advance.

[avytin]

I remember I had a memory leak leading to the same problem some months ago. I will try to find the file to see if it was a recorder error, but I am not sure I still have it. I think I simply destroyed the recorders and created new ones after some timesteps to tweak around this problem.

[fmk]

are you using nonlinearBeamColumn or the Transformation constraint handler in your script? .. you might as well send me your script fmckenna AT berkeley DOT edu