We created a docker image as stated in the nvflare doc - Getting Started — NVIDIA FLARE 2.4.0 documentation with nvflare package in that image. Then we are trying to start the nvflare application in poc mode and starting the server process.
We are facing 2 issues here.
- we initialized the poc mode and tried starting the server process by placing the commands in Dockerfile , then built the image . When deployed the pod is not coming up and going to crashloopbackoff error.
How to start the process in a running state? - We added a dummy sleep command to the Dockerfile and deployed that image and the pod is now up. when we try to start the server process application inside the pod manually, then it is getting started. But now , the process is spawning multiple child processes and as a result it is utilizing 200% of CPU. How should we address this to ensure the process doesn’t spawn too many child processes.