Replies: 2 comments 1 reply
-
Hi @mgd-hin , This is both a docker problem and an NVFlare problem. So the "start.sh" of NVFlare is actually running in background as daemon process. If you use docker to run a background process it will exit right away. |
Beta Was this translation helpful? Give feedback.
0 replies
-
you can use production model ( non-HA, secure, localhost) as well. our internal testing team is using this mode. client/servers are docker containers |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
I have a project that runs well in simulator and POC mode (on one machine). I would now like to run this project on several different containers to try a more real-life scenario for my FL project. However, I am having trouble starting my clients. This is the docker-compose that I use:
The image I use is from the standard Dockerfile (i.e. what is created when I run the provision mode with the Docker builder component).
Paths and file names to the
start.sh
scripts are correct. The volume mounts to the container workspace folders are also successful. When I usedocker-compose up
, all containers start but exit immediately with exit code 0.The POC Readme says to use
startup/start.sh server_host(name or IP)
. I have tried providinglocalhost:8002:8003
,[server-container-name]: 8002: 8003
, and assigning IP addresses to every container in the compose network. However, I am not sure if this is a connectivity issue first and foremost, because even if a client cannot find a server to connect to, I assume the service would keep running on that client for more than a few seconds.Is there anything obvious that I am missing or doing wrong from NVFlare side (if this is only a docker struggle, I understand this is the wrong place to ask) ?
Is there an example or a tutorial for running POC on different containers (ideally with docker-compose)?
Beta Was this translation helpful? Give feedback.
All reactions