Hello,
I have a question regarding running PhotoScan in headless mode on Linux. Currently we us PBS scheduler submit jobs where we are required to request memory for the node we are requesting to use and the job we are submitting starts PhotoScan and runs a python script that creates various network tasks and submits them to an array of PhotoScan processing nodes and those nodes are also setup through the same scheduler. We are running into a memory issue while the script is running on the requested node with 50 GB of memory requested. From what I was able to monitor it appears that the processed data is being stored on a stack to be dumped later on to their respective directories and what I want to know is if PhotoScan running in headless mode uses the stack to store the data or if its just how out scheduler works. I also believe that since the data is being temporarily stored on a stack if that is also causing the point_cloud.zip file to be saved with a .tmp extension. I'm just wondering if you have any suggestions on whether we should request more memory because that's how PhotoScan works in headless mode or if we should divide up the network tasks in our current python script into individual scripts and submit each of those as separate jobs. Either way, some sort of insight would be helpful.
-Justin
Topic: Headless Mode Memory Usage (Read 1038 times)
