[YARN-4016] docker container is still running when app is killed

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Component/s: nodemanager
• Labels: Docker
• Assignee: Hong Zhiguo
• Reporter: Hong Zhiguo
• Created: 04/Aug/15 09:49
• Updated: 16/May/18 19:58
• Resolved: 07/Aug/15 00:22

Description

The docker_container_executor_session.sh is generated like below:

### get the pid of docker container by "docker inspect"
echo `/usr/bin/docker inspect --format {{.State.Pid}} container_1438681002528_0001_01_000002` > .../container_1438681002528_0001_01_000002.pid.tmp

### rename *.pid.tmp to *.pid
/bin/mv -f .../container_    1438681002528_0001_01_000002.pid.tmp .../container_1438681002528_0001_01_000002.pid

### launch the docker container
/usr/bin/docker run  --rm  --net=host --name container_1438681002528_0001_01_000002 -v ... library/mysql /container_1438681002528_0001_01_000002/launch_container.sh" 

This is obviously wrong because you can not get the pid of a docker container before starting it. When NodeManager try to kill the container, pid zero is always read from the pid file.