PaddleCloud icon indicating copy to clipboard operation
PaddleCloud copied to clipboard

Can not stop docker container and too much `kworker` process on host

Open Yancey0623 opened this issue 8 years ago • 1 comments

The problem describetion:

  • GPU can not be released normally, The Docker Continaer which used GPU can not be stop.
  • The command nvidia-smi will be hanged.
  • Too high for the system load(3k+).
  • Too much kworker process(3k+).

Environment

  • OS: CentOS 7.2
  • Kernel: 4.4.79-1.el7.elrepo.x86_64
  • GPU: P40, Driver: 375.26
  • Docker: 1.12
  • How to use GPU in Docker: mount nvidia libraries and nvidia device in container.

Yancey0623 avatar Aug 24 '17 01:08 Yancey0623

Thought kubernetes will use --device argument to attach GPU device to the container, and this issue is very possible that has something to do with docker.

typhoonzero avatar Aug 24 '17 02:08 typhoonzero