search

mfrey / baltimore

baltimore is an analysis script for the libARA framework
GNU General Public License v3.0
0 stars 0 forks source link

Memory leak in analysis #18

Closed fgrosse closed 11 years ago

fgrosse commented 11 years ago

When I run huge simulation runs (like bigStatic 50 nodes) with 30 repetitions and the analysis starts, my memory slowly fills until swap is full and then everything gets so slow that I usually abort the script. Maybe I have introduced this with the parsing of the binary routing table data.

We need to check if we free all no longer needed resources and make the analysis executable on systems that don't have 4GB RAM available.

Currently 2 GB free RAM + SWAP memory are not enough

mfrey commented 11 years ago

I have never used it, but maybe it is worth to try the python profiler?

fgrosse commented 11 years ago

Ah thanks for the hint, I will try this out. Unfortunately, form reading the manual I understand that this is no memory profiler so we will see if it can help us in this case.

mfrey commented 11 years ago

The missing plot save really solves the issue with the memory leak?

fgrosse commented 11 years ago

Yep I just ran a full analysis and watched my memory usage. It does no longer rise significantly

Edit: The files where saved before that but pyplot seems to have internal handles for the currently open figure. If those are not closed they accumulate in memory