We are attempting to use recon-all -all on T1w MRI image analysis. We tested it on AWS EC2 in a docker container successfully. The uname -a output the EC2 is Linux ip-10-193-42-20 5.11.0-1019-aws #20~20.04.1-Ubuntu SMP Tue Sep 21 10:40:39 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux. The docker file entrypoint is a job_script.py python code to call a subprocess of bash script, like recon-all using a python library called subprocess. 

When we used the same docker image with AWS ECS or batch processing, from the files created inside the processed subject's mri folder, the recon-all stuck at "-fill" step because the input files (wm.mgz, aseg.auto_noCCseg.mgz, and transforms/talairach.lta) are created, but NO filled.mgz was created as the output file of "-fill " step in recon-all -autorecon2. 

I have searched the list and no similar errors have been reported. Does anyone have any thoughts on how to troubleshoot this? Also I have attached the reon-all.log in cas it is of any use. 

1). FreeSurfer version: 7.2.0

2). Platform: AWS EC2 batch or ECS Fargate

Looking forward to your feedback.

Best Regards

Xin